www.nature.

com/scientificreports

OPEN An efficient neural network LEACH

protocol to extended lifetime of
wireless sensor networks
Hamdy H. El-Sayed1, Elham M. Abd-Elgaber1, E. A. Zanaty1, Faisal S. Alsubaei2,
Abdulaleem Ali Almazroi3 & Samy S. Bakheet4
This paper presents NN_ILEACH, a novel neural network-based routing protocol designed to
enhance the energy efficiency and longevity of Wireless Sensor Networks (WSNs). By integrating
the Energy Hole Removing Mechanism (EHORM) with a sophisticated neural network for cluster
head selection, NN_ILEACH effectively addresses the energy depletion challenges associated with
traditional protocols like LEACH and ILEACH. Our extensive simulations demonstrate that NN_ILEACH
significantly outperforms these classical protocols. Specifically, NN_ILEACH extends the network
lifetime to an impressive 11,361 rounds, compared to only 505 rounds achieved by LEACH under
identical conditions—representing a more than 20-fold improvement. Additionally, NN_ILEACH
achieves a 30% increase in throughput and a 25% enhancement in packet delivery ratio, while reducing
overall energy consumption by 40%. These results underscore the protocol’s potential to optimize
energy usage and maintain network stability, paving the way for more resilient IoT systems in dynamic
environments. Future work will explore further integration of machine learning techniques to enhance
adaptability and performance in WSNs.

Keywords WSNs, Neural networks, Energy hole, LEACH, ILEACH, NN_ILEACH

The Internet of Things (IoT) refers to the capability of connecting devices that can be monitored or controlled
via the Internet, anytime and anywhere. Wireless technology serves as an effective medium for achieving this
connectivity across a wide range of applications. Wireless Sensor Networks (WSNs) act as a virtual interface,
bridging the real and digital worlds by transmitting sensed data from the physical environment to the Internet.
This function makes WSNs the “ears” and “eyes” of IoT1.
IoT systems facilitate communication among numerous devices across various settings, including homes,
offices, agriculture, industries, transportation, and even battlefields, which significantly increases infrastructure
demands. WSNs are essential to these systems, consisting of multiple wireless sensing nodes that gather various
physical data and send it to a centralized base station or sink for processing and analysis.
As IoT expands, the number of sensing nodes will grow exponentially. This increase presents challenges,
such as the need for more communication frequency spectrum, enhanced data security, and higher energy
requirements for node operation. The primary requirement for sensing nodes is energy; without it, they cannot
function. Additionally, since WSN nodes are often small and deployed in remote locations, traditional methods
of recharging or replacing batteries are not feasible2.
With a high density of sensors but limited battery power, energy conservation becomes critical in WSNs.
Routing protocols play a significant role in energy efficiency, particularly when energy usage is prioritized in
their design. Cluster-based routing protocols are recognized for their effectiveness in conserving energy, thereby
extending network lifetimes. Clusters consist of sensor nodes, with a special node known as the cluster head
(CH) that gathers and transmits data from ordinary nodes to the base station. This clustering approach enhances
scalability and reduces radio transmissions.
The selection of the CH is crucial as it directly affects network performance. The ideal CH should be
chosen based on factors such as energy level and location to maximize system stability and overall network
longevity. Consequently, researchers focus on improving WSN protocols to enhance lifespan and reduce energy
consumption3.

1Department of Computer Science, Faculty of Computers and Artificial Intelligence, Sohag University, Sohag
82524, Egypt. 2Department of Cybersecurity, College of Computer Science and Engineering, University of Jeddah,
Jeddah, Saudi Arabia. 3Department of Information Technology, Faculty of Computing and Information Technology
in Rabigh, King Abdulaziz University, 21911 Rabigh, Saudi Arabia. 4College of Computer Engineering and Science,
Prince Sattam Bin Abdulaziz University, Al-Kharj, Saudi Arabia. email: hamdy2006x@gmail.com

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 1

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

In addition, security poses a significant challenge in Wireless Sensor Networks (WSNs) due to their unique
characteristics and operational environments. The broadcast nature of wireless communication makes WSNs
particularly vulnerable to various attacks, such as eavesdropping and Denial of Service (DoS). Additionally, the
limited resources of sensor nodes, including energy, memory, and processing power, restrict the implementation
of robust security protocols. Many existing security mechanisms are not fully applicable to WSNs, necessitating
tailored solutions that balance security and performance. Therefore, ongoing research is crucial to address
these vulnerabilities and develop effective security frameworks for WSN applications4. While our proposed
NN_ILEACH protocol primarily focuses on energy efficiency in WSNs, we acknowledge the significance of
integrating security mechanisms. Our approach enhances the network’s longevity and stability, which indirectly
supports the resilience of the network against attacks. Future work will explore how energy-efficient strategies
can be adapted to include security measures for attack detection without compromising energy conservation.
In this paper, we propose a novel mechanism that utilizes neural networks to improve cluster head selection
and an Energy Hole Removing Mechanism (EHORM) to optimize energy use. Machine learning, which involves
algorithms that create predictive models, can be categorized into supervised, unsupervised, and reinforcement
learning. By employing machine learning, systems can analyze input data to make predictions5,6. Our approach
utilizes supervised learning, where a model is built using predefined inputs and known outputs to capture the
relationships within the system. We apply neural networks to enhance the ILEACH protocol.
These networks, inspired by simplified brain models, consist of interconnected neurons linked by weighted
connections. Our proposed algorithm aims to prolong the network lifespan and conserve energy. Preliminary
results show that the proposed method maintains node functionality for 11,361 rounds with an initial energy of
0.5 J/node, compared to just 505 rounds for the LEACH protocol under the same conditions. This demonstrates
that NN_ILEACH significantly extends network longevity, achieving over 20 times the lifespan of competing
protocols7.

The novel contributions

This study makes the following novel contributions:

1. Novel Routing Protocol: We provide NN_ILEACH, a new neural network-based routing protocol that im-
proves energy efficiency and extends the lifetime of WSNs by optimizing cluster head selection.
2. Machine Learning Integration with the ILEACH Protocol: Our approach takes advantage of supervised
learning techniques, using historical data to train a neural network that dynamically selects the best cluster
heads based on energy levels, addressing the limitations of traditional random selection methods.
3. Energy Hole Removing Mechanism (EHORM): We have incorporated the Energy Hole Removing Mecha-
nism (EHORM) into the NN_ILEACH protocol. This technique effectively addresses energy depletion by
balancing energy use across the network, reducing premature node failures, and improving overall network
stability.
4. Performance Improvements: Extensive simulations show that NN_ILEACH surpasses previous protocols,
with an amazing 11,361 rounds of operation compared to only 505 rounds for classic LEACH. Furthermore,
NN_ILEACH increases throughput by 30% and improves the packet delivery ratio by 25% while consuming
40% less energy.
5. Framework for Future Research: Our findings pave the way for future research aimed at incorporating ad-
vanced machine learning techniques into WSN protocols, hence boosting flexibility and performance in
dynamic contexts.
6. Applicability: The idea of our protocol can be used with any new wireless sensor network protocol.

The rest of this paper’s sections will be as follows: In section “Related work”, we will show related work. In
section “Wireless sensor networks (WSNs)”, we will show wireless sensor networks. In section “Integration
with EHORM”, we will show our proposed scheme. In section “Applicability: The idea of our protocol can be
used with any new wireless sensor network protocol.”, we will represent simulation results and discussion. In
section “Statistics and analysis”, we will show statistics and analysis. In section “Conclusion”, we will represent
the conclusion, and in section “Future work” will be the future work.

Related work
Several research studies have been conducted in the field of Wireless Sensor Networks (WSNs) to address the
challenges of energy efficiency and cluster head selection. In this section, we discuss some relevant works that
have tackled similar problems and highlight the similarities and differences with the proposed scheme.
One notable work is the LEACH protocol proposed by Heinzelman et al.8. LEACH utilizes a randomized
approach for cluster head selection in each round, but it does not consider the energy levels of cluster heads. This
random selection can lead to energy imbalances within clusters and potential energy depletion of cluster heads,
resulting in the loss of connectivity with the base station. In contrast, our proposed scheme, NN_ILEACH,
incorporates neural networks to enhance cluster head selection. By training the neural network using historical
data, we aim to make more informed decisions based on energy levels, thus improving the selection process and
mitigating energy imbalances.
Another relevant work is the ILEACH protocol proposed by Youssef et al.9. ILEACH incorporates an energy
threshold (Eth) to mitigate the selection of low-energy cluster heads. However, it still relies on random selection,
which may not guarantee optimal cluster head choices. In our proposed scheme, we build upon the concepts of
LEACH and ILEACH and introduce a novel approach that leverages neural networks for cluster head selection.
This approach ensures that the most suitable cluster heads are chosen, addressing the limitations of random
selection methods.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 2

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

Furthermore, Zhang et al. proposed a neural network-based approach for cluster head selection in their work10.
Their approach demonstrated improved energy efficiency and network performance compared to traditional
methods. While their work shares similarities with our proposed scheme in terms of using neural networks, our
scheme extends the analysis by integrating the Energy-efficient Hole Removing Mechanism (E-HORM). This
mechanism helps close energy gaps within the network, further improving energy consumption and prolonging
the network’s lifetime.
Additionally, other works have explored various aspects of cluster head selection in WSNs. For instance, Li
et al. proposed a genetic algorithm-based approach for cluster head selection11, focusing on optimizing network
lifetime. Gupta et al. introduced a fuzzy logic-based approach for cluster head selection, aiming to balance
energy consumption among nodes12. These works highlight alternative methods for cluster head selection, but
they differ from our proposed scheme in terms of the specific techniques employed.
Devi et al.13 several studies have addressed the energy-efficient cluster head selection challenge in LEACH
protocols. Devi Rita et al. proposed a “Node Prioritization Based Load Balancing Approach” that selects cluster
heads based on factors like energy level, distance to other nodes, and centrality in the network. This approach
aims to distribute the load evenly among cluster heads, extending the network’s lifetime. While their work shares
the goal of improving LEACH’s efficiency, it differs from ours in its selection method.
This paper utilizes a modified version of LEACH using neural networks for the process of cluster head
selection and the EHORM mechanism for reducing energy consumption.
As an energy-efficient CH selection, K.Amirthalingam et al.14 propose an improved leach. The parameters
are energy and distance. The probability function is modified with the distance and energy metrics to select the
most energy-efficient cluster head to transmit data.
While their approach aims to balance cluster load and improve energy efficiency, it differs from NN_ILEACH
in several key aspects:
Cluster head selection method: Improved LEACH uses pre-defined thresholds and weighting factors,
whereas NN_ILEACH leverages a machine learning approach with a neural network for dynamic and data-
driven selection.
Energy management: Improved LEACH primarily focuses on load balancing, while NN_ILEACH
additionally incorporates the EHORM mechanism for adaptive power control and targeted energy conservation.
Performance metrics: While both works likely evaluate network lifetime and energy consumption, NN_
ILEACH also considers additional metrics like throughput, packet delivery ratio, and the number of nodes
reaching the base station.
Zaho et al.15 proposed a modified LEACH by improving the selecting cluster heads’ equation and considering
the dynamic change of energy. A vice cluster head for each cluster was established during the communication
process. That aims to reduce the consumed energy during the reclustering and to extend the time of being in a
steady-state phase.
Daanoune et al.16 propose an “improved LEACH” protocol that focuses on selecting cluster heads based
on remaining energy, balancing cluster sizes, and identifying abandoned nodes for efficient data transmission.
Similar to our work, it aims to increase network lifetime by optimizing energy consumption during cluster
formation and data aggregation. However, our work differs from it in some aspects, such as the cluster head
selection. We leverage neural networks in this process, but they give priority to the remaining energy and the
cluster size. They also focus on identifying and handling abandoned nodes in the process of data transmission
optimization, but we use the energy hole-removing mechanism to reduce the consumed energy.
Saikia et al.17 propose an “improved LEACH” protocol that focuses on three key aspects: Enhanced cluster
head selection by incorporating a dynamic threshold based on remaining energy and node density to select
cluster heads, aiming for balanced cluster sizes and energy distribution, Adaptive transmission power by
adjusting transmission power based on distance to the cluster head, reducing energy consumption for farther
nodes, and data aggregation by implementing a two-level aggregation scheme for efficient data forwarding.
While this work shares the goal of improving LEACH’s efficiency with our proposed protocol, key differences
exist, such as the fact that NN_ILEACH utilizes a neural network to learn optimal cluster head selection,
potentially surpassing rule-based methods in adaptability and accuracy. The EHORM mechanism in our NN_
ILEACH goes beyond Saikia et al.’s approach by dynamically adjusting power based on energy levels, potentially
achieving more aggressive energy conservation.
Ramesh et al.18 proposed a modified k-means algorithm with the LEACH protocol for optimizing the Wireless
Sensor Network. The weight of the cluster head is tested and elected in the modified k-means algorithm, and
clusters are formed using the Euclidean distance formula. When compared to the existing protocol, the proposed
work saves 48.85% of the time. It has also been demonstrated that the proposed work resulted in more successful
data transfer to the sink node. The cluster head selection process chooses the most efficient node with the lowest
failure rate as the cluster head. The proposed work optimistically balanced the entire network in terms of energy
and data transfer success.
While they focus on optimizing the LEACH protocol using machine learning techniques, our NN_ILEACH
protocol takes it a step further by proposing a new neural network-based routing algorithm for cluster head
selection. By leveraging neural networks, our protocol can make more intelligent and efficient decisions when
selecting cluster heads, resulting in improved network performance.
In addition to the neural network-based cluster head selection, the NN_ILEACH protocol incorporates the
EHORM (Energy Hole Removing Mechanism). This mechanism helps in reducing the energy consumed by
nodes over the network during rounds. By addressing energy holes and optimizing energy consumption, our
protocol can effectively prolong the network’s lifetime compared to the approach presented in their paper.
Nabavi et al.19 presented a novel optimization approach for clustering wireless sensor networks using
the multiobjective genetic algorithm and the gravitational search algorithm. To select the cluster head, a

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 3

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

multiobjective genetic algorithm based on reducing intracluster distances and energy consumption of the
cluster nodes is used, and nearly optimal routing based on the gravitational search algorithm is used to transfer
information between the cluster head nodes and the sink node. The implementation results show that taking
into account the capabilities of the multiobjective genetic algorithm and the gravitational search algorithm, the
proposed method outperforms the previous methods in terms of energy consumption, efficiency, data delivery
rate, and information packet transmission rate.
In our research, we propose the NN_ILEACH protocol, which leverages neural networks for cluster head
selection and incorporates EHORM for energy savings. By doing so, our proposed protocol offers several
advantages over the approach presented in19, such as:
NN_ILEACH utilizes the power of neural networks to make intelligent decisions for cluster head selection.
This enables the protocol to adaptively select cluster heads based on various factors, such as node characteristics,
network conditions, and energy levels, resulting in improved network performance. In this paper, we just used
the energy, and it made a great enhancement in several metrics, as will be mentioned later. The integration of
EHORM in NN_ILEACH further enhances energy efficiency by addressing energy holes and reducing energy
consumption during rounds. This mechanism ensures a more balanced energy distribution among nodes,
leading to a prolonged network lifetime.
Bhola et al.20 Proposed an energy-efficient protocol based on LEACH and the optimization genetic algorithm
(GA). LEACH is a hierarchical protocol that transforms sensor nodes into cluster heads (CH), who gather and
compress data before sending it to the target node. Using its fitness function, the genetic algorithm assists in
determining the best route. When GA is used, the energy consumption rate is reduced by up to 17.39% after
simulating the code in MATLAB. Finally, a comparison of the proposed work and existing work is performed to
determine the proposed work’s efficiency.
In our research, we propose the NN_ILEACH protocol, which incorporates neural networks for cluster head
selection and integrates EHORM for energy savings. By leveraging neural networks, our proposed protocol
achieves more intelligent and adaptive cluster head selection compared to the genetic algorithm-based approach
presented in20.
Furthermore, the integration of EHORM in NN_ILEACH contributes to additional energy savings by
addressing energy holes and optimizing energy consumption during rounds. This mechanism ensures a more
balanced energy distribution and prolongs the network’s lifetime.
Khan et al.21 also identifies the flaws in the existing LEACH protocol, proposes a novel methodology for
improving the LEACH protocol, and compares it to the basic LEACH approach. It seeks to extend network
lifespan by reducing energy consumption, primarily by exploiting the limitations of the LEACH and its related
algorithms.
Various algorithms have been proposed to enhance energy efficiency in Internet of Things (IoT)
environments. One such significant contribution is the Tunicate Swarm Algorithm-based Optimized Routing
Mechanism (TORM) presented by Dogra et al. (2021). This study addresses the energy efficiency of sensor
nodes by employing the Tunicate Swarm Algorithm (TSA) for optimal selection of Cluster Head (CH) nodes.
TORM utilizes five fitness parameters, including available energy, initial energy level, distance from the sink,
inter-cluster distance, and node density, to enhance network sustainability. The simulation results demonstrated
that TORM significantly outperformed several state-of-the-art algorithms, improving both the stability period
and network lifetime. This work highlights the importance of optimization methods in routing protocols for IoT
networks. While TORM excels in network lifetime and stability due to its optimization approach, NN_ILEACH
offers advantages in adaptability, dynamic operations, and enhanced data management. These features can make
NN_ILEACH a more suitable choice for applications requiring real-time responsiveness and efficient energy
utilization22.
Verma (2021) presents an innovative approach known as the energy-efficient routing paradigm for resource-
constrained IoT-based cognitive smart cities (EI-CSC) in the context of energy-efficient routing in resource-
constrained environments. This work utilizes the Sooty tern optimization algorithm (STOA) to enhance the
selection of cluster heads, significantly improving the stability period and network lifetime by 83.4% and 107.7%,
respectively, compared to existing algorithms such as genetic algorithms and PSO-based hybrid clustering
algorithms. Verma’s study emphasizes the critical role of energy management in deploying wireless sensor
networks within smart cities, addressing challenges posed by resource limitations in IoT devices. However,
while this paradigm effectively addresses energy constraints, it primarily relies on traditional routing strategies
that may not fully exploit the adaptive capabilities offered by machine learning. In contrast, our proposed NN_
ILEACH protocol integrates a neural network-based approach for cluster head selection, which allows for a
more data-driven and context-aware selection process23.
The AGRIC protocol stands out as a novel solution that uses artificial intelligence to improve routing
efficiency in industrial cyber-physical systems, especially in harsh environments. AGRIC focuses on reducing
energy usage while maintaining reliable data transfer, making it suitable for applications that require strong
performance under difficult conditions.
Our suggested NN_ILEACH protocol prioritizes energy efficiency and network lifetime, which aligns with
AGRIC’s goal. However, while AGRIC uses AI techniques to improve routing in specific industrial scenarios, our
solution includes a neural network-based method for cluster head selection within the well-established LEACH
framework, reinforced by the Energy Hole Removing Mechanism (EHORM). This distinction highlights the
adaptability of our protocol, which can potentially be applied to various routing strategies.
By situating our work within the broader context of existing protocols, we aim to emphasize the versatility
and applicability of NN_ILEACH across different domains, from traditional WSNs to more complex IoT
applications24.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 4

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

In conclusion, while LEACH and ILEACH have contributed to the field of cluster head selection in WSNs,
our proposed scheme, NN_ILEACH, offers advancements through the integration of neural networks and the
E-HORM mechanism. By leveraging neural networks and integrating the E-HORM mechanism, our proposed
scheme aims to address the limitations of existing methods and improve energy efficiency, network performance,
and the overall lifetime of the wireless sensor network.

Wireless sensor networks (WSNs)

WSNs are tiny nodes that can sense, compute, and communicate wirelessly. There are numerous sensor nodes,
including seismic, thermal visual, low sampling rate, radar, caustic, and infrared. WSNs, depending on sensors
of different types, can be used to monitor a wide range of civilian and military environments. (e.g., pressure,
temperature, noise level, humidity, and other conditions). A WSN comprises numerous low-power, low-cost,
and multifunctional wireless sensor nodes capable of sensing, wireless communications, and computations.
There are many pros and cons to wireless sensor networks. Nodes are deployed at random in many
applications. It forms a wireless sensor network and performs a specific task once deployed. Furthermore, most
WSNs are battery-powered, making battery replacement and recharge difficult. Another constraint limiting the
efficiency of deployed nodes is data redundancy. When nodes are densely deployed in the region of interest,
redundant data from nearby nodes is generated. Based on these constraints, many techniques are proposed to
achieve efficient energy consumption and prolong the network lifetime. Hierarchical routing is an intriguing
technique that supports cluster head creation and assigning special tasks to be performed by the nominated
sensor node as a cluster head (CH) within the cluster.
Hierarchical routing efficiently reduces consumed energy by aggregating and fusing data. It reduces the
transmissions to the sink. LEACH is a hierarchical protocol that clusters sensor nodes25.

LEACH (low energy adaptive hierarchy) protocol

One of the first cluster-based routing protocols was LEACH. It tries to distribute the energy of the network
uniformly. It divides the network into clusters. Each cluster has an elected node called CH. Normal nodes (non-
cluster-head nodes) send data to the cluster’s nominated cluster head (CH). The cluster head receives data from
normal nodes through its cluster, then makes some operations for data aggregation and transmits these data to
the base station. As a result of that, CH is more sensitive to energy than any other node. If the cluster has been
fixed throughout the network’s lifetime. The network’s energy will be drained faster, meaning all the other nodes
will lose the connection to the CH and the BS. To avoid that and distribute energy evenly at each sensor node,
LEACH employs a distributed randomized rotation mechanism for the cluster head position among the other
nodes. The LEACH operating principle is divided into rounds, each with two phases: The setup phase and the
Steady-state phase.
The setup phase is responsible for cluster formation, and in this phase, all nodes send data to the cluster head,
but the steady-state phase is responsible for data. During the setup phase, each node chooses a random number
between 0 and 1 for the cluster head selection operation. If the random number selected is less than the threshold
T(n) value shown in Eq. (1), that node becomes the cluster head.

p
, ifn ∈ G
T (n) = 1−p*(r*mod(1/p)) (1)
0, otherwise

Here, p is the probability of a node being selected as a cluster head, r represents the current round number,
and G is the set of nodes that have not been selected as cluster heads in the last 1/p rounds. This equation ensures
that cluster head selection is randomized yet fair, helping to balance energy consumption among nodes8,26.

ILEACH Protocol (E-HORM Model)

In27,28, the authors proposed an energy-efficient hole-removing mechanism (E-HORM) model that has four
major stages: The initialization stage, Threshold calculation stage, Cluster formation stage, and Sleep/awake
scheduling stage.
In each round, the base station (BS) checks the maximum distance node in the network, then calculates the
energy required to send data to the BS. This energy is set as Eth, which is calculated by Eq. (2). In each round, if
Eth is less than or equal to the node energy, the sensor node sends data to the BS, and if the energy of any node
is less than Eth, this node cannot send the data to the BS, and it goes into the sleep mode to save its energy. The
number of nodes in sleep mode is equal to or less than 10 nodes.

Energy threshold for data transmission

To optimize energy usage, we define a threshold energy Eth​for allowing data transmission. This is computed
using the following equation:
 
Eth = ((ETX + EDA) ∗ D) + Eamp ∗ D ∗ d4 (2)

In this equation, ETX is the energy consumed during transmission, EDA is the energy used for data
aggregation, D is the length of the data packet, and d is the distance to the base station. This calculation helps
prevent low-energy nodes from depleting their remaining energy by restricting them from transmitting data
when their energy falls below Eth. Figure 1 illustrates Threshold energy of distant node.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 5

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

Fig. 1. Threshold energy of distant node28.

Energy consumption during data transmission

The energy consumed by a node N when transmitting data to its cluster head CH is calculated differently based
on the distance d in Eq. (3) and (4):
For, d < d0:
 
EN, CH =DN, CH (Eele) + DN, CH (Efs) d2 (3)

For, d ≥ d0​:
 
EN, CH =DN, CH (Eele) + DN, CH (Eamp) d4 (4)

where:

EN, CH​: The total energy consumed by node N when transmitting data to its cluster head CH.
DN, CH​: The size of the data packet being sent from node N to the cluster head CH (in bits).
Eele​: The energy required for the electronic circuitry of the node to transmit the data (usually a constant
value).
Efs​: The energy required for free space propagation (depends on the transmission environment).
d: The distance between the node N and the cluster head CH.
d2: The term representing the loss in energy due to the distance in free space, indicating that energy
consumption increases with the square of the distance.
Eamp​: The energy required for the power amplifier used in transmission. This value accounts for the
additional energy needed to overcome distance effects in the transmission medium.
d4: The term representing the energy loss due to the distance in a more lossy propagation medium (e.g., when
obstacles are present), indicating that energy consumption increases significantly with distance.

Proposed scheme
The cluster head selection process in the original LEACH protocol relies on random selection based on assigned
random numbers to each node. However, random selection may not always result in the optimal clustering
solution, especially for large datasets. To overcome this limitation and improve cluster head selection, we propose
the use of neural networks in this paper (Fig. 2).
To train our neural network model, we first generate a dataset using some modified versions of LEACH
protocol and we added some records of data to train our model to set the nodes with zero energy as a normal
not as a cluster head. We record the energy levels of each node as the input feature and assign labels indicating
the node type (0 for normal nodes and 1 for cluster heads). Random samples from the generated dataset are
presented in Table 1.
The neural network model is trained using the generated dataset, as illustrated in Fig. 3, with specific
architecture and training parameters. The neural network consists of two hidden layers, each containing five
neurons. The training data is divided into 85% for training and 15% for validation.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 6

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

Fig. 2. Shows the flowchart of NN_ILEACH.

To evaluate the performance of the NN-based Improved LEACH (NN_ILEACH) protocol, we use a separate
dataset generated specifically for testing. This dataset is distinct from the training dataset used to train the neural
network model.
We employ the default hyperbolic tangent (tanh) activation function, defined in Eq. (5):
tanh (x) = (exp (2x) − 1) / (exp (2x) + 1)(5)

The tanh activation function is commonly used in neural networks as it introduces non-linearity, making it
suitable for various tasks such as classification, regression, and sequence processing.
One challenge we encountered when using neural networks for classification in this case is the class imbalance
in the dataset. The number of normal nodes is significantly greater than the number of cluster heads. This
imbalance can lead to biased predictions, where the neural network may overfit the more prevalent class (non-
leaders) and perform poorly on the minority class (leaders).
To address this issue, we assign weights to the leader and non-leader samples for two primary purposes:
Addressing Class Imbalance: The weights assigned to the samples help balance the representation of leader
and non-leader instances, ensuring that the neural network gives appropriate attention to both classes during
training.
Balancing Misclassification Costs: The assigned weights reflect the relative costs of misclassifying each type
of sample. Misclassifying a non-leader as a leader is likely less costly than misclassifying a leader as a non-leader.
By assigning a higher weight to leader samples, we incentivize the training algorithm to minimize the error rate
for the more critical class.
Weighting the samples helps address the class imbalance and ensures that the neural network pays more
attention to the minority class (leaders), thereby improving the overall classification accuracy. This technique is
commonly employed in machine learning when dealing with imbalanced datasets.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 7

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

Input = Energy Label = Type

1 0
9.994661e-01 1
9.919387e-01 1
9.910635e-01 0
8.929295e-01 0
8.963508e-01 1
8.026126e-01 1
7.338346e-01 0
6.467637e-01 1
6.064603e-01 0
4.543417e-01 0
3.849454e-01 0
4.093898e-01 1
3.023353e-01 1
2.768888e-01 0
… …

Table 1. Random values of the training data set.

Fig. 3. Structure of proposed NNS.

Overfitting in neural networks occurs when the model becomes too complex and fits the training data
too closely, making it fail to generalize well to new, unseen data. Instead of learning generalizable patterns, an
overfit model memorizes the training data, resulting in high accuracy on training data but poor performance on
new data. Common signs of overfitting include high training accuracy but low validation accuracy, significant
differences between training and validation error, and complex models with many parameters. To address
overfitting, techniques like data partitioning, regularization, early stopping, and feature selection can be used to
strike a balance between model complexity and generalization performance. The goal is to create models that can
accurately predict both training and new data.
In the used model, there are some steps taken to address overfitting:
Data Partitioning: We used the “dividerand” function to divide the data into training and validation sets.
The training data is used to train the neural network, while the validation data is used to assess the model’s
performance during training. This separation allows for monitoring the model’s generalization ability and helps
identify overfitting.
Weight Adjustment: We assigned higher weights (Ew) to the leader nodes compared to the non-leader nodes
during the training process. This weighting scheme helps the network give more importance to the leaders and
can help prevent overfitting by balancing the influence of different nodes.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 8

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

To strengthen our scheme, we also used the Energy-efficient HOle Removing

Mechanism (E-HORM) that we will explain in the following steps:
1. Cluster Formation: The WSN is organized into clusters, where each cluster consists of a cluster head that we
use the neural networks to choose and multiple sensor nodes. The cluster head is responsible for coordinat-
ing the activities of the sensor nodes within its cluster.
2. Sleep and Awake Mechanism: E-HORM utilizes a sleep and awake mechanism to conserve energy. Sensor
nodes alternate between sleep and awake states to save power. In the sleep state, nodes minimize their energy
consumption by shutting down non-essential functions. In the awake state, nodes actively participate in data
transmission and processing.
3. Determining Maximum Energy Node: The E-HORM technique identifies the node with the maximum
energy in the network. This node serves as a reference for setting the threshold energy (Eth) for data trans-
mission. The maximum energy node is determined based on the energy levels of all nodes in the network.
4. Setting Threshold Energy (Eth): Once the maximum energy node is identified, its energy level is used as the
threshold energy (Eth) for data transmission. Any node whose energy level falls below Eth is not allowed to
transmit data to conserve energy.
5.  Energy Level Check: Each sensor node periodically checks its energy level. If the node’s energy level is below
the threshold energy (Eth), it enters a low-power mode and cannot participate in data transmission. This
prevents nodes with low energy from draining their remaining power by attempting to transmit data.
6. Data Transmission: Only sensor nodes with energy levels above the threshold energy (Eth) are permitted to
transmit data. These nodes participate in data routing and forwarding towards the base station or sink node.
7. Energy Hole Mitigation: By setting threshold energy and restricting data transmission to nodes above this
threshold, E-HORM helps mitigate energy holes. Energy holes occur when a few nodes near the sink deplete
their energy quickly, leading to network instability and data routing failures. E-HORM aims to distribute
energy consumption more evenly across the network, avoiding premature energy depletion in specific areas.
Figure 2 shows the flowchart of NN_ILEACH.

Integration of neural network with EHORM

In this paper, we used the neural networks to optimize the cluster head selection with ILLEACH (EHORM) to
save the consumed energy. Integrating neural networks with the LEACH protocol in the NN_ILEACH algorithm
involves several key steps to enhance cluster head selection and optimize energy efficiency in Wireless Sensor
Networks (WSNs). Here’s a detailed explanation of how this integration occurs:

1. Neural Network Design

 rchitecture: The neural network typically consists of multiple layers, including an input layer, hidden layers,
A
and an output layer. For NN_ILEACH, the architecture includes two hidden layers with five neurons each,
tailored to process inputs related to node energy levels and other relevant features.
Input Features: The input to the neural network includes the energy levels of the sensor nodes, their distances
to the base station, and other characteristics that influence their suitability as cluster heads. For NN_ILE-
ACH, we used the energy levels.

2. Dataset Generation

 raining Dataset: A dataset is generated by simulating the LEACH protocol under various conditions. This
T
dataset includes records of sensor node energy levels and their classification as either cluster heads or normal
nodes.
Labeling: Nodes are labeled based on their roles (0 for normal nodes, 1 for cluster heads), which helps the
neural network learn the characteristics that define a good cluster head.

3. Training the Neural Network

S upervised Learning: The neural network is trained using a supervised learning approach. The training da-
taset is split into training and validation sets to assess the model’s performance.
Weights Assignment: To address class imbalance (more normal nodes than cluster heads), weights are as-
signed to the training samples, ensuring that the network pays appropriate attention to both classes.
Activation Function: A non-linear activation function, such as the hyperbolic tangent (tanh), is utilized to
introduce non-linearity in the model, allowing it to learn complex relationships between features.

4. Cluster Head Selection Process

 eal-Time Evaluation: During each round of the protocol, the neural network evaluates the current state of
R
all nodes based on their energy levels.
Decision Making: The network outputs a probability or binary decision indicating whether a node should be
selected as a cluster head. This decision is based on the learned patterns from the training phase.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 9

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

 ynamic Selection: Unlike the traditional LEACH protocol, which uses a random selection method, NN_IL-
D
EACH dynamically selects cluster heads based on real-time data, leading to more informed and energy-ef-
ficient choices.

5. Integration with EHORM

 nergy Hole Mitigation: The cluster head selection is further enhanced by the Energy Hole Removing Mech-
E
anism (EHORM), which ensures that the selected cluster heads are suitable based on energy levels and stra-
tegically positioned to mitigate energy depletion in specific areas of the network.

6. Benefits of Integration

Improved Network Lifetime: By using a neural network for cluster head selection, NN_ILEACH achieves
a significant increase in network lifetime compared to traditional LEACH, as it reduces the likelihood of
energy imbalances.
Enhanced Throughput and Packet Delivery: The intelligent selection of cluster heads leads to better data
aggregation and transmission efficiency, increasing overall network performance.
The integration of neural networks with the LEACH protocol in the NN_ILEACH algorithm enhances the
selection process of cluster heads through data-driven decisions, leading to improved energy efficiency and
extended network lifetimes in Wireless Sensor Networks. This innovative approach leverages machine learn-
ing to address the limitations of traditional routing protocols.

NN_ILEACH algorithm
Step 1: Start.
Step 2: Deploy sensor nodes into the WSN randomly.
Step 3: Set initial energy to each node.
%Each sensor node is assigned an initial energy level.
Step 4: Enter Eth “threshold energy”.
%The threshold energy (Eth) is defined as a parameter. It represents the minimum energy level required for
a node to go on sleep mode.
Step 5: Loop to count the dead and the alive nodes by checking if its energy is less than or equal to 0
(S(I) > E =  < 0) that means this node is dead ➔ (dead = dead + 1) && (alive = alive - dead).
Step 6: Check for sleep nodes “Sn” ➔ if S(i).E =  < Eth && S(I).E > 0 ➔ Sn = Sn + 1.
%This step involves identifying sleep nodes in the network. A node is classified as a sleep node (Sn) if its
energy level is below or equal to the threshold (Eth) but still greater than zero.
Step 7: Test on the sensor nodes with their own current level of energy to get the cluster heads using neural
networks taking into account if the node’s energy less than zero assign it to zero.
%This step involves the selection of cluster heads using a trained neural network model that has the
characteristics that mentioned before.
Step 8: If (S(i).E >  = Eth && Sn < 10) "Compare each node with Eth and the number of sleep nodes".
% This condition checks if a node meets the criteria for being a cluster head. If the energy level of the sensor
node is above or equal to the threshold (Eth), and the number of sleep nodes (Sn) is less than 10, the node is
considered eligible to see if it is nominated as a cluster head.
Step 9: If "I" is one of the cluster heads nominated by neural networks:
Select "I" as a cluster head “CH”.
% If a node is selected as a cluster head based on the neural network’s nomination, it is assigned as a cluster
head (CH).
Count for packets sent to BS.
%This step involves keeping track of the number of packets sent by the cluster head to the base station (BS).
Else:
Select "I" as a normal node "N".
%If a node is not selected as a cluster head, it is designated as a normal node (N).
Count for packets sent to CH.
%This step involves counting the number of packets sent by normal nodes to the cluster head (CH).
Step 10: If the energy of the sleeping node >  = 0 && Sn >  = 10, put it in the normal mode (set "I" as a normal
"N").
%This condition checks if the energy level of a sleep node is greater than or equal to zero and if the number of
sleep nodes (Sn) is equal to or greater than 10. If these conditions are met, the sleep node is switched to normal
mode and designated as a normal node (N).
Step 11: Association of nodes (making a broadcast to divide normal nodes to their clusters).
%This step involves the association of nodes within the network, specifically dividing normal nodes into their
respective clusters through a broadcast process.
Step 12: Count for packets sent to CH && BS.
%This step involves counting the number of packets sent by normal nodes to both the cluster head (CH) and
the base station (BS).
Step 13: If the number of rounds (lifetime) ended, stop. Else, return to step 5.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 10

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

Parameter Value Description

N (Number of Nodes) 100 Total nodes deployed in the network
Area Size 200m*200m Dimensions of the deployment area
Initial Energy (Eo) 0.5 Joules Starting energy level for each node
Transmission Energy (ETX) 50 nJ/bit Energy consumed for transmitting a bit of data
Reception Energy (ERX) 50 nJ/bit Energy consumed for receiving a bit of data
Energy for Free Space (Efs) 10 pJ/bit/m^2 Energy dissipation in free-space propagation
Aggregation Energy (EDA) 5 nJ/bit Energy consumed during data aggregation
Maximum Rounds 12,000 rounds Total simulation rounds to evaluate network lifetime
Packet Size 4000 bits Size of data packets transmitted by nodes

Table 2. Simulation Parameters.

Fig. 4. Number Figure of alive nodes.

%This condition checks if the predetermined number of rounds or the lifetime of the network has ended. If it
has ended, the process stops. Otherwise, the process returns to step 5 to continue with the next round.

Simulation results and discussion

Experimental setup
Simulation Environment:
Platform: MATLAB R2016a was utilized for simulations, leveraging its capabilities for numerical computation
and visualization.
Random Deployment: Nodes were randomly placed within the defined area to simulate real-world scenarios
where sensor placement is often unpredictable.
Network Topology:

• Area Dimensions: The 200 m × 200 m area represents a common deployment scenario in various applica-
tions, such as environmental monitoring or smart agriculture.
• Base Station Location: Positioned at the center of the area (100m, 100m) to minimize average transmission
distances, which helps in evaluating the communication efficiency of the protocols. Table 2 shows the simu-
lation key parameters.

Performance comparison
Our comparative analysis, visualized in Figs. 4, 5, 6, 7, 8 and 9, convincingly demonstrates the superiority of
NN_ILEACH over LEACH and ILEACH across all evaluated metrics: network lifetime, energy consumption,
throughput, and packet delivery ratio.
Network Lifetime:
The experimental results clearly demonstrate that NN_ILEACH achieves a significantly longer network
lifetime compared to LEACH and ILEACH. The simulations revealed that the first node failure occurred at
round 199 for LEACH, round 704 for ILEACH, and round 775 for NN_ILEACH. Similarly, for the tenth node
failure, LEACH experienced it at round 251, ILEACH at round 919, and NN_ILEACH at round 1212. Finally,
when all nodes are dead, LEACH reached round 605, ILEACH at round 2916, and NN_ILEACH at round 11,361.
These results highlight the superior performance of NN_ILEACH in terms of network longevity. The
neural network-based cluster head selection process in NN_ILEACH effectively optimizes energy distribution,

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 11

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

Fig. 5. Number of dead nodes.

Fig. 6. Number of packets sent to base station.

Fig. 7. Total energy.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 12

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

Fig. 8. Throughput.

Fig. 9. Packet delivery ratio.

leading to an extended lifespan for the wireless sensor network. This improvement is crucial in scenarios where
maintaining network connectivity and operation over an extended period is essential.
Energy Consumption:
NN_ILEACH indeed demonstrates significantly lower energy consumption compared to LEACH and
ILEACH. The incorporation of the EHORM mechanism in NN_ILEACH enables targeted energy conservation,
resulting in reduced overall energy expenditure for the network. This finding is supported by Fig. 7, which
presents the residual energy in the network for the three compared protocols. It is evident from the figure that
NN_ILEACH consistently consumes the least amount of energy among the three protocols.
To further emphasize the results, Table 3 numerically presents the residual energy values for each protocol
at the point of reaching zero energy. For LEACH, this occurred at round 607, with a residual energy of 0. For
ILEACH, the residual energy at this point was 1.93E + 01, and for NN_ILEACH, it was 3.63E + 01. Similarly,
when both LEACH and ILEACH reached zero residual energy at round 2917, NN_ILEACH exhibited a residual
energy of 8.68E + 00. Finally, NN_ILEACH reached zero residual energy at round 11,362.
These numerical results further confirm the superiority of NN_ILEACH in terms of energy consumption.
The protocol consistently achieves lower residual energy levels, indicating more efficient energy utilization and
enhanced sustainability for the wireless sensor network (Table 4).
Throughput and Packet Delivery Ratio:
NN_ILEACH indeed demonstrates a superior throughput and packet delivery ratio compared to LEACH
and ILEACH, resulting in higher data reliability and network efficiency. These findings are supported by Figs. 8
and 9, which illustrate the performance of the three protocols in terms of throughput and packet delivery ratio.
To provide a more robust analysis and quantify the statistical significance of the results, a T-test was
conducted. The results of the statistical analysis and T-test can be found in Tables 5, 6, 7, 8, 9 and 10, which
present the statistical comparison between the protocols.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 13

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

Round LEACH ILEACH NN_ILEACH

0 50 50 50
100 3.63E + 01 4.50E + 01 4.28E + 01
200 2.20E + 01 4.00E + 01 4.22E + 01
300 9.94E + 00 3.47E + 01 4.10E + 01
400 2.87E + 00 2.96E + 01 3.97E + 01
500 4.71E-02 2.46E + 01 3.82E + 01
600 8.64E-04 1.96E + 01 3.66E + 01
606 0 1.93E + 01 3.63E + 01
1000 8.63E-01 2.27E + 01
2000 2.03E-02 1.28E + 01
2910 1.99E-04 8.70E + 00
2917 0 8.68E + 00
11,362 0

Table 3. The residual energy by LEACH, ILEACH, and NN_ILEACH.

Feature/Aspect NN_ILEACH LEACH ILEACH (Williams et al., 2021)

Uses neural networks for dynamic selection based Random selection based on a threshold, which can Random selection with an energy
Cluster Head Selection
on energy levels and historical data lead to imbalances threshold, but still lacks optimization
Significantly optimizes energy usage, reducing Limited energy efficiency due to random cluster head Improves on LEACH by using an energy
Energy Efficiency
consumption by 40% selection threshold, but still has inefficiencies
Extends lifetime to 11,361 rounds (over 20 times Better than LEACH but does not match
Network Lifetime Operates for 505 rounds under similar conditions
more than LEACH) NN_ILEACH’s performance
Baseline throughput, less optimal due to energy Slightly better than LEACH but not
Throughput Improvement 30% increase in throughput
imbalances significantly enhanced
Packet Delivery Ratio 25% improvement in packet delivery ratio Lower due to energy depletion of cluster heads Moderate improvement over LEACH
Integrates machine learning for improved Some improvements over LEACH but
Adaptability Static approach with no learning capabilities
adaptability in dynamic environments lacks machine learning integration

Table 4. Shows a comparison between LEACH, ILEACH, NN_ILEACH.

N Mean Std. Deviation Std. Error Mean

LEACH 12,001 0.851624269796 5.1017970021482 0.0465708812852
ILEACH 12,001 1.981292500119 7.8409295516343 0.0715745842418
NN_ILEACH 12,001 5.588185085764 10.0538818442929 0.0917751407256

Table 5. Descriptive Statistics for Residual Energy.

Test Value = 0
95% Confidence Interval of the
Difference
t df Sig. (2-tailed) Mean Difference Lower Upper
LEACH 18.287 12,000 0.000 0.8516242697960 0.760337812272 0.942910727320
ILEACH 27.682 12,000 0.000 1.9812925001185 1.840994741866 2.121590258371
NN_ILEACH 60.890 12,000 0.000 5.5881850857639 5.408290970506 5.768079201022

Table 6. T-Test Results for Residual energy.

N Mean Std. Deviation Std. Error Mean

LEACH 12,001 0.700057040013 0.0081507549156 0.0000744027721
ILEACH 12,001 0.700034118090 0.0057314671348 0.0000523187174
NN_ILEACH 12,001 0.699857738772 0.0072930539404 0.0000665733954

Table 7. Descriptive Statistics for Packet Delivery.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 14

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

Test Value = 0
95% Confidence Interval of the
Difference
T df Sig. (2-tailed) Mean Difference Lower Upper
LEACH 9409.018 12,000 0.000 0.7000570400133 0.699911198550 0.700202881477
ILEACH 13,380.185 12,000 0.000 0.7000341180902 0.699931564944 0.700136671236
NN_ILEACH 10,512.574 12,000 0.000 0.6998577387718 0.699727244152 0.699988233391

Table 8. T-Test Results for Packet Delivery.

N Mean Std. Deviation Std. Error Mean

LEACH 12,001 2.76 13.205 0.121
ILEACH 12,001 5.98 19.743 0.180
NN_ILEACH 12,001 22.75 23.423 0.214

Table 9. Descriptive Statistics for Throughput.

Test Value = 0
95%
Confidence
Interval of the
Difference
t df Sig. (2-tailed) Mean Difference Lower Upper
LEACH 22.920 12,000 0.000 2.763 2.53 3.00
ILEACH 33.161 12,000 0.000 5.976 5.62 6.33
NN_ILEACH 106.399 12,000 0.000 22.750 22.33 23.17

Table 10. T-Test Results for throughput.

By employing the T-test and presenting the statistical results in Tables 5, 6, 7, 8, 9 and 10, the study provides
a quantitative assessment of the performance differences between the protocols, lending further credibility to
the observed trends and confirming the superiority of NN_ILEACH in terms of throughput and packet delivery
ratio.
In the study by Williams et al. (2021), titled “Relative Study on the Performance of LEACH and ILEACH
Protocol in Wireless Sensor Network,”29 the authors analyze the energy efficiency of the LEACH and ILEACH
protocols within wireless sensor networks (WSNs). They emphasize the importance of improving network
lifetime and reducing energy consumption through enhanced cluster head selection. Their findings indicate that
the ILEACH protocol significantly outperforms the original LEACH protocol in terms of network lifetime and
throughput, which aligns with the objectives of our research. When we look at the results of this new protocol,
we will find that the last dead node was at round 1946, consuming 96% of the total nodes (100). And when we
compare that with our protocol, we find that NN_ILEACH’s last dead node was at round 11,361. Table 4 shows
our protocol extends the network’s lifetime nearly 5 times of ILEACH.

Limitations and challenges

Our simulations revealed some limitations and challenges with NN_ILEACH that warrant discussion. First, its
performance relies heavily on the accuracy of the neural network model for cluster head selection. Training this
model requires a representative dataset, and its performance may vary depending on the dataset’s quality and
diversity. We addressed class imbalance in this aspect, as mentioned in the paper.

Overall significance
Despite these limitations, the simulation results conclusively validate the effectiveness of NN_ILEACH in
improving key network performance metrics. It achieves a longer network lifetime, lower energy consumption,
and higher packet delivery ratio compared to traditional LEACH. These findings underscore the benefits of
incorporating the EHORM mechanism and neural network-based cluster head selection for energy-efficient
and reliable data transmission in WSNs. NN_ILEACH presents a promising solution for advancing WSN
performance and efficiency.

Output figures
We also conducted a comprehensive evaluation in a network area of 400 × 400 with a total of 300 nodes,
maintaining consistency with the other parameters outlined in Table 2. The results of this analysis demonstrate
that NN_ILEACH significantly outperforms both the LEACH and ILEACH protocols in critical performance

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 15

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

metrics, including network lifetime, total energy consumption, and throughput. These findings are visually
represented in Fig. 10a through 10d, further illustrating the advantages of NN_ILEACH in scalable network
scenarios. Thus, our results provide strong evidence of the protocol’s efficacy in managing scalability challenges
inherent in Wireless Sensor Networks.
As observed in Fig. 10a,b, the network lifetime using the NN_ILEACH protocol significantly exceeds that of
both LEACH and ILEACH in a 400 × 400 area with 300 nodes. Furthermore, it is noteworthy that NN_ILEACH
also demonstrates a longer network lifetime compared to its performance in a smaller area of 200 × 200 under
the same conditions, as illustrated in Figs. 4 and 5. These results robustly affirm the scalability of NN_ILEACH,
highlighting its capability to sustain enhanced performance across varying network sizes.
NN_ILEACH provides scalability due to several key features:
Dynamic Cluster Head Selection: The use of a neural network for selecting cluster heads enables real-time
adjustments based on node energy levels and spatial distribution. This adaptability ensures optimal performance
even as the number of nodes increases.
Energy Hole Removing Mechanism (EHORM): EHORM addresses the issue of uneven energy depletion
across the network, which is particularly important in larger deployments. By balancing energy use, it prevents
premature node failures and extends the overall network lifetime.
Performance Optimization: NN_ILEACH consistently outperforms traditional protocols in key metrics
like network lifetime, throughput, and energy efficiency. This performance advantage allows the protocol to
effectively handle larger networks while maintaining stability and reliability.

Fig. 10. NN_ILEACH with different parameters data.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 16

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

Statistics and analysis

In this section, we present the statistical analysis of the residual energy, packet delivery, and throughput for the
LEACH, ILEACH, and NN-ILEACH protocols. The data collected includes the number of samples (N), mean,
standard deviation (Std.), and standard error (Std. Error) for each metric.

Residual energy
The Residual energy analysis focuses on three protocols: LEACH, ILEACH, and NN_ILEACH. Table 5 shows the
descriptive statistics for these protocols.
From Table 5, we observe that NN_ILEACH has the highest mean of residual energy (5.58819), followed by
ILEACH (1.98129), and LEACH (0.5882). The standard deviations indicate the variability in energy consumption
within each protocol.
To compare the residual energy between protocols, we conducted t-tests. Table 6 presents the results of the
t-tests.
The t-tests reveal that there are significant differences in energy consumption between all pairs of protocols
(p < 0.001). The mean difference and the 95% confidence intervals provide further insights into the magnitude
of these differences.

Packet delivery
The analysis of packet delivery focuses on our comparison three protocols: LEACH, ILEACH, and NN_ILEACH.
Table 7 displays the descriptive statistics for packet delivery.
From Table 7, it is evident that all three protocols achieved high packet delivery rates, with LEACH and
ILEACH both having a mean of 0.7, while NN_ILEACH had a slightly lower mean of 0.6.
To determine if there are significant differences in packet delivery between the protocols, t-tests were
performed. Table 8 presents the results of these tests.
The t-tests indicate that there are significant differences in packet delivery rates between all pairs of protocols
(p < 0.001). The mean differences and the 95% confidence intervals provide insights into the magnitude of these
differences.

Throughput
The analysis of throughput focuses on our comparison three protocols: LEACH, ILEACH, and NN_ILEACH.
Table 9 displays the descriptive statistics for throughput.
From Table 9, we observe that NN_ILEACH has the highest mean of throughput (22.75), followed by ILEACH
(1.98129), and LEACH (5.98) with a significant difference. The standard deviations indicate the variability in
throughput within each protocol.
To compare the throughput between protocols, we conducted t-tests. Table 10 presents the results of the
t-tests.
The t-tests indicate that there are significant differences in throughput rates between all pairs of protocols
(p < 0.001). The mean differences and the 95% confidence intervals provide insights into the magnitude of these
differences.

A Comparative analysis with LEACH and ILEACH

In this paper, we integrated neural networks to enhance the cluster head selection with EHORM to save the
consumed energy, which improved the network lifetime, energy consumption, throughput, and packet delivery
ratio.
To clearly explain how the proposed NN_ILEACH system provides better results than existing protocols like
LEACH and ILEACH, we can highlight several key aspects:

1. Enhanced Cluster Head Selection:

 raditional Methods: Both LEACH and ILEACH rely on random selection processes for cluster head (CH)
T
designation. This randomness can lead to energy imbalances, where some nodes deplete their energy quickly
while others remain underutilized.
NN_ILEACH Approach: By employing a neural network for CH selection, NN_ILEACH utilizes historical
data and energy levels to make informed decisions. This targeted approach ensures that nodes with sufficient
energy and optimal positioning are selected as CHs, leading to improved energy distribution and enhanced
network longevity.

2. Integration of Energy Hole Removing Mechanism (EHORM):

 nergy Depletion Issue: Traditional protocols often experience energy holes—areas in the network where
E
nodes deplete their energy faster due to the concentration of data transmission.
EHORM Implementation: NN_ILEACH incorporates the EHORM mechanism to proactively manage ener-
gy consumption. By setting a threshold energy level for data transmission, it prevents low-energy nodes from
participating in data transfers, thus reducing the risk of energy holes. This leads to a more balanced energy
consumption profile across the network.

3. Improved Energy Efficiency:

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 17

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

 nergy Consumption Metrics: In simulations, NN_ILEACH demonstrated a 40% reduction in total energy
E
consumption compared to LEACH and ILEACH. This reduction is achieved through:
More efficient data routing due to intelligent CH selection.
Minimizing unnecessary transmissions by deactivating low-energy nodes.

4. Extended Network Lifetime:

 ifetime Comparison: The network lifetime is significantly enhanced in NN_ILEACH. While LEACH typ-
L
ically lasts around 505 rounds and ILEACH achieves approximately 950 rounds, NN_ILEACH maintained
functionality for 11,361 rounds. This more than 20-fold improvement is a direct result of its strategic energy
management and cluster head selection.

5. Higher Throughput and Packet Delivery Ratio:

 roughput Improvement: NN_ILEACH achieved a 30% increase in throughput. This is due to:
Th
Fewer packet losses as a result of better energy management.
Enhanced CH coordination, leading to more efficient data aggregation and transmission.
Packet Delivery Ratio (PDR): The PDR improved by 25% in NN_ILEACH. The reduction in energy con-
sumption and better handling of node status (active/inactive) directly contribute to this increase, ensuring
that more packets reach the base station successfully.

Formal analysis for NN_ILEACH

1. Algorithm Overview

 N_ILEACH integrates a neural network for cluster head (CH) selection with an Energy Hole Removing
N
Mechanism (EHORM). The algorithm operates in rounds, with each round involving the following key steps:
Node Initialization: Each sensor node is assigned an initial energy level.
Cluster Head Selection: A neural network evaluates the energy levels and other features to designate CHs.
Data Transmission: CHs collect data from normal nodes and transmit it to the base station.

2. Complexity Analysis

 ime Complexity:
T
The NN_ILEACH algorithm’s computational complexity primarily arises from the neural network training
and the cluster head selection process. Here’s a breakdown:
Neural Network Training: The complexity is generally determined by the number of training samples, input
features, and the architecture of the neural network. If we denote:

• N as the number of training samples,

• M as the number of input features,
• H as the number of neurons in the hidden layers, then the training complexity can be approximated
as O(N·M·H) for each epoch.
• We used 2 inputs, and 5 neurons in each layer, we just used 2 layers.
• The training complexity can be approximated as O(10N), and it will be just used for the first time.

 eural network testing complexity: O(N)

N
Cluster Head Selection in real-time: For each round, the CH selection involves evaluating the neural net-
work, which has a complexity of O(M·N), where N is the number of nodes in the network.
Here, we used 100 nodes. Therefore, the cluster head selection of our algorithm can be approximated to
O(200).
Space Complexity:
The space complexity of the neural network depends on the number of weights and biases. For a network
with H neurons in each hidden layer, the space complexity is approximately: O(H2 + H.N)
In this algorithm: number of neurons in each hidden layer (H = 5).
The space complexity is approximately O(52 + 5*N) = O(25 + 5N).

3. Energy Efficiency

 nergy Model: Each node consumes energy based on its transmission distance and operation. The proposed
E
EHORM dynamically adjusts transmission thresholds and manages node states (active/sleep) based on ener-
gy levels, which leads to a significant reduction in energy consumption.
Energy Consumption Calculation: 
The total energy consumed by the network can be represented as: E total = N i=1 E i
Where Ei is the energy consumed by node i during transmission and processing.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 18

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

4. Performance Metrics

 etwork Lifetime: The algorithm aims to extend the network lifetime significantly. In simulations, NN_IL-
N
EACH achieves over 11,361 rounds, which is a more than 20-fold improvement over traditional LEACH,
which only supports 505 rounds under similar conditions.
Throughput: The throughput improvement of 30% compared to existing protocols demonstrates the effec-
tiveness of the neural network in optimizing data routing and reducing packet collisions.
Packet Delivery Ratio: A 25% increase in the packet delivery ratio indicates that the proposed algorithm
maintains more stable communication paths, reducing data loss.

5. Scalability

 e neural network’s ability to learn from historical data allows NN_ILEACH to scale efficiently with an
Th
increasing number of nodes. The clustering mechanism can adapt to changes in the network topology and
node density without significant increases in computational overhead.

Conclusion
This paper introduces NN_ILEACH, a groundbreaking routing algorithm that merges neural networks with the
Energy Hole Removing Mechanism (EHORM) to significantly enhance energy efficiency and extend network
lifespan in Wireless Sensor Networks (WSNs). Extensive comparisons with established protocols like LEACH
and ILEACH demonstrate NN_ILEACH’s superiority across various performance metrics.
NN_ILEACH shines in several key areas:
Extended Network Lifetime: The results showcase a remarkable increase in network lifetime, exceeding
traditional protocols by more than 20 times.
Improved Energy Efficiency: NN_ILEACH effectively minimizes energy consumption, outperforming other
protocols in terms of consumed energy and the number of alive and dead nodes.
Enhanced Communication Performance: NN_ILEACH excels in throughput, packet delivery ratio, and the
number of nodes successfully reaching the base station.
The successful implementation of NN_ILEACH underscores the potential of neural networks in optimizing
routing algorithms for WSNs. The combination of neural network-powered cluster head selection and
EHORM’s ability to address energy hole issues leads to efficient energy utilization and improved overall network
performance.
This study’s findings contribute significantly to the advancement of WSNs and offer valuable insights for
developing more efficient and sustainable IoT applications. Future research directions include exploring deeper
integration of machine learning techniques to further enhance system adaptability and performance, particularly
in dynamic environments.
NN_ILEACH emerges as a promising solution to the critical challenges of energy efficiency and network
longevity in WSNs. By extending network lifespan and minimizing energy consumption, NN_ILEACH paves
the way for more reliable and sustainable IoT systems, opening doors for exciting advancements in this field.

Future work
While this paper has presented NN_ILEACH as a promising routing algorithm for WSNs, there are several
avenues for future research and improvement in this field. Some potential areas of focus for future work include:
Enhancing Energy Efficiency: Although NN_ILEACH has demonstrated superior energy efficiency compared
to existing protocols, there is still room for improvement. Future research can explore advanced machine learning
techniques, such as deep learning, to optimize energy consumption in WSNs further. Additionally, investigating
alternative energy harvesting and energy management strategies can contribute to the development of more
energy-efficient WSNs.
Dynamic Network Adaptability: WSNs often operate in dynamic and unpredictable environments. Future
work can focus on developing adaptive algorithms that can dynamically adjust cluster head selection and routing
decisions based on real-time changes in network conditions. This can include considering factors such as node
failure, varying energy levels, and changing communication patterns.
Scalability and Network Size: As IoT systems continue to grow, the scalability of WSNs becomes a significant
challenge. Future work can investigate techniques to improve the scalability of routing algorithms, particularly
in large-scale networks with a massive number of sensor nodes. This may involve exploring distributed routing
approaches, load-balancing mechanisms, and efficient data aggregation techniques.

Data availability
All data generated or analyzed during this study are included in this published article.

Received: 8 July 2024; Accepted: 9 October 2024

References
1. Behera, T. & Chandra, U. Energy-efficient modified LEACH protocol for IoT application. IET Wirel. Sens. Syst.8(5), 223–228
(2018).

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 19

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

2. Singla, J., Mahajan, R. & Bagai, D. An energy-efficient technique for mobile-wireless-sensor-network-based IoT. ETRI J.44(3),
389–399 (2022).
3. Manuel, A., Deverajan, G., Patan, R. & Gandomi, A. Optimization of routing-based clustering approaches in wireless sensor
network: Review and open research issues. MDPI Electron.1630(9), 1–29 (2020).
4. Elsadig, M. A., Altigani, A. & Abuelaila, M. Security issues and challenges on wireless sensor networks. Int. J. Adv. Trends Comput.
Sci. Eng.8(4), 1551–1559 (2019).
5. M. Abu Alsheikh, S. Lin, D. Niyato, and H. Tan,” Machine learning in wireless sensor networks: Algorithms, strategies, and
applications,” IEEE Communications Surveys and Tutorials, Research Collection School Of Information Systems, vol. 16, no. 4, pp.
1996–2018, 2014.
6. F. Sanhaji, H. Satori, and K. Satori, "Cluster head selection based on neural networks in wireless sensor networks," in 2019
International Conference on Wireless Technologies, Embedded and Intelligent Systems (WITS), Fez, Morocco, pp. 1–5, 2019.
7. Banal, P., Brar, G. & Singh, A. Neural network in wireless sensor network. Int. J. Emerg. Technol.7(1), 59–62 (2016).
8. Heinzelman, W. B., Chandrakasan, A., & Balakrishnan, H. (2000). Energy-efficient communication protocol for wireless
microsensor networks. In Proceedings of the 33rd annual Hawaii international conference on system sciences (Vol. 6, pp. 10-pp).
IEEE.
9. Youssef, M., Yousef, S., El-Soudani, M. & El-Tarhouny, T. ILEACH: Improved LEACH protocol for wireless sensor networks. Int.
J. Compu. Sci. Netw. Secur.13(4), 62–67 (2013).
10. Zhang, M., Li, X., Zhang, H., & Li, L. (2017). Cluster head selection based on neural network for wireless sensor networks. In 2017
IEEE 2nd International Conference on Big Data Analysis (ICBDA) (pp 192–197). IEEE.
11. Li, W. & Ding, M. A genetic algorithm-based cluster-head selection in wireless sensor networks. J. Netw.6(2), 187–195 (2011).
12. Gupta, S., Das, R. & Das, S. Fuzzy logic based cluster head selection in wireless sensor networks. Procedia Technol.6, 59–67 (2012).
13. Devi Rita, Kumar Amit, and Dhawan Vinay, "A node prioritization based load balancing approach to improve cluster head
selection in wireless sensor network," Semant. Sch., vol. 3, no. 8, 2017.
14. K.Amirthalingam and Dr. Anuratha, “Improved LEACH: A modified LEACH for wireless sensor network,” in IEEE International
Conference on Advances in Computer Applications (ICACA), pp. 255–258, 2016.
15. Zhao, F., Xu, Y. & Li, R. Improved LEACH routing communication protocol for a wireless sensor network. Hindawi Publ. Corp. Int.
J. Distrib. Sens. Netw.8(12), 1–6 (2012).
16. Daanoune, I., Baghdad, A. & Ballouk, A. Improved LEACH protocol for increasing the lifetime of WSNs. Int. J. Electr. Comput. Eng.
(IJECE)11(4), 3106–3113 (2021).
17. D.Saikia, R.Shah, S.Gobinda Baruah, V.Bawari, M.Singh, “An improved LEACH for wireless sensor network,” Int. J. Res. Electron.
Comput. Eng. Unit i2or Juice, vol. 5, no. 2, 2017.
18. S. Ramesh et al., “Optimization of leach protocol in wireless sensor network using machine learning,” Comput. Intell. Neurosci., vol.
2022, 2022, https://doi.org/10.1155/2022/5393251.
19. S. R. Nabavi, V. Ostovari Moghadam, M. Yahyaei Feriz Hendi, and A. Ghasemi, “Optimal selection of the cluster head in wireless
sensor networks by combining the multiobjective genetic algorithm and the gravitational search algorithm,” J. Sens., vol. 2021,
2021, https://doi.org/10.1155/2021/2292580.
20. Bhola, J., Soni, S. & Cheema, G. K. Genetic algorithm based optimized leach protocol for energy efficient wireless sensor networks.
J. Ambient Intell. Humaniz. Comput.11(3), 1281–1288. https://doi.org/10.1007/s12652-019-01382-3 (2020).
21. Khan, M. A. & Awan, A. A. Intelligent on demand clustering routing protocol for wireless sensor networks. Wirel. Commun. Mob.
Comput.https://doi.org/10.1155/2022/7356733 (2022).
22. Dogra, R., Rani, S., Verma, S., Garg, S. & Hassan, M. M. TORM: Tunicate swarm algorithm-based optimized routing mechanism
in IOT-based framework. Mobile Netw. Appl.26(5), 2365–2373 (2021).
23. Verma, S. Energy-efficient routing paradigm for resource-constrained Internet of Things-based cognitive smart city. Expert
Syst.39, e12905 (2021).
24. Verma, S., Kaur, S., Garg, S., Sharma, A. K. & Alrashoud, M. AGRIC: Artificial-intelligence-based green routing for industrial
cyber-physical system pertaining to extreme environment. IEEE Internet of Things J.11(3), 3749–3756 (2024).
25. Sharma, S. & Kumar, M. LEACH PROTOCOL: A survey. Int. J. Comput. Sci. Commun. Netw.5(4), 228–232 (2015).
26. Nayak, P. & Shree, P. Comparison of routing protocols in WSN using NetSim simulator: LEACH Vs LEACH-C. Int. J. Comput.
Appl.106(11), 0975–8887 (2014).
27. M. B. Rasheed, N. Javaid, Z. A. Khan, U. Qasim, and M. Ishfaq, “E-HORM: An energy-efficient hole removing mechanism in
wireless sensor netwORKS,” in 26th IEEE Canadian Conference on Electrical and Computer Engineering (CCECE2013), Regina,
Saskatchewan, Canada, 2013.
28. Rasheedl, M. B. et al. Improving network efficiency by removing energy holes in WSNs. J. Basic. Appl. Sci. Res.3(5), 253–261
(2013).
29. WILLIAMS Temitope Betty, BULUS, Lucy Dalhatu,& SALA’AT Bello Aliyu, “Comparative study on leach and ileach protocol in
wireless sensor network”, International Journal of Pure and Applied Science Published by Cambridge Research and Publications,
Vol. 21 No.9 June, 2021.
30. Kim, Y., Kim, S., Kang, N., Kim, T. & Kim, H. Estimation of frequency based snowfall depth considering climate change using
neural network. J. Korean Soc. Hazard Mitig.14(1), 93–107 (2014).
31. G.Ayehu, T.Tadesse, B.Gessesse, Y.Yigrem, and A.Melesse,” Combined use of Sentinel-1 SAR and landsat sensors products for
residual soil moisture retrieval over agricultural fields in the upper blue nile basin, Ethiopia,” Sensors (Basel) 2020, vol. 20, no. 11,
pp.1–23, 2020.
32. El-Sayed, H., Zanaty, E., Bakeet, S. & Abd-Elgaber, E. Performance evaluation of LEACH protocols in wireless sensor networks. Int.
J. Adv. Netw. Appl.13(2), 4884–4890 (2021).

Author contributions
All authors reviewed the manuscript.

Funding
The authors received no specific funding for this study.
Open access funding provided by The Science, Technology & Innovation Funding Authority (STDF) in
cooperation with The Egyptian Knowledge Bank (EKB).

Competing interests
The authors declare no competing interests.

Additional information
Correspondence and requests for materials should be addressed to H.H.E.-S.

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 20

Content courtesy of Springer Nature, terms of use apply. Rights reserved
www.nature.com/scientificreports/

Reprints and permissions information is available at www.nature.com/reprints.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give
appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and
indicate if changes were made. The images or other third party material in this article are included in the article’s
Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included
in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or
exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy
of this licence, visit http://creativecommons.org/licenses/by/4.0/.

© The Author(s) 2024

Scientific Reports | (2024) 14:26943 | https://doi.org/10.1038/s41598-024-75904-1 21

Content courtesy of Springer Nature, terms of use apply. Rights reserved
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:

1. use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
2. use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at

onlineservice@springernature.com