Exploiting Temporal Correlation Mechanism for Designing Temperature-Aware Energy-Efficient Routing Protocol for Intrabody Nanonetworks

An Intrabody Nanonetwork (IBNN) is composed of integrated nanoscale devices, implanted inside the human body to collect diagnostic information and tuning medical treatments. The non-invasive continuous monitoring and precision of these nanoscale devices in the diagnostic of diverse diseases is improving advanced monitoring, therapeutic, and telemedicine services. The unique feature constraints of these nanoscale devices (such as inadequate energy, storage, and computational resources) along with the molecular absorption thermal challenges due to Electromagnetic (EM) communication in the Terahertz band (THz) are the primary limitations in enabling efficient routing in IBNNs. Our proposed protocol explicitly addresses these challenges and proposes a routing scheme, which enables temperature-aware energy-efficient data transmission to avoid hotspot formation and controlled energy consumption. Furthermore, our proposed temporal correlation based data decision approach allows only those Nano Bio Sensors (NBSs) to transmit the periodic data packets that have updated information for avoiding unnecessary energy consumption and antenna radiation exposure on biological cells. The presented work also considers the instant data retrieval requirement of the healthcare system and introduces an on-demand data retrieval approach that ensures instant transmission of updated information to the healthcare system. The effectiveness of our proposed scheme is evaluated by comparing it with the flooding scheme and thermal-aware routing algorithm (TARA) using the Nano-SIM tool. The results obtained from extensive simulations validate that our proposed protocol achieves 75% - 85 % low temperature rise and improved network lifetime.


I. INTRODUCTION
Advances in nanotechnology are boosting the development of nanoscale devices [1], which have tremendous potential for enabling cellular monitoring and precision [2], [3]. These nanoscale devices, which consist of data sensing, processing, and computation components, leverage the idea of Intrabody Nanonetworks (IBNNs) [1], [4]. The idea of IBNNs is based on the collaborative effort of a large number of implanted nanoscale devices to empower the emergence of a more sophisticated healthcare monitoring and diagnosis The associate editor coordinating the review of this manuscript and approving it for publication was Oussama Habachi . system. Nanoscale devices with their unique features, such as enabling non-invasive cellular level monitoring, early detection of cancerous cells and precision and accuracy in drug delivery and diagnosis are increasing IBNNs potential for a wide range of nanomedicine applications. These nanomedicine applications include in-vivo glucose monitoring system [5], monitoring and detection of cancer biomarkers and human chorionic gonadotrophin [6], [7], which are envisioned to reshape the future of traditional healthcare system.
The emerging scope of IBNNs is leading to a growing interest in addressing the challenges of IBNNs, such as enabling nanoscale communication and the development of novel communication protocols. Among the various methods investigated for enabling nanoscale communication such as molecular communication, nanomechanical communication and acoustic communication [1], Electromagnetic (EM) communication in the Terahertz (THz) band has gained significant attention [8]. The reason behind this rising interest is the development of graphene-based plasmonic nanoantennas that support EM waves propagation in the THz band with the extremely higher bit data rates [4]. Accordingly, the feasibility and scope of EM waves communication in IBNNs is leading to its adaptation for designing EM-based routing schemes [9]- [12]. However, channel modeling in the THz band is regarded as a significant challenge in enabling EM communication in the THz band [13], [14]. EM waves propagation in THz band triggers molecular absorption phenomena, which affects the properties of the propagating medium in terms of attenuation and thermal noise. The issue of thermal noise temperature due to molecular absorption was first highlighted by Jornet and Akyildiz in [14]. Later, other studies [15]- [17] also focused on the molecular absorption noise temperature. In [18], Elayan et al. investigated the photothermal effects of molecular absorption in IBNNs. In our previous work [19], we analyzed the thermal effects of nanonodes density on heat generation for the first time. Based on the existing studies [14]- [18], the thermal effects of molecular absorption can cause a rise in the temperature of the body, which can lead to the severe damages to the human body cells, including causing a heat stroke and multiple organ failures.
Thermal un-awareness during routing decisions can also significantly increase the temperature in the biological medium as frequent emission of EM waves for transmitting data increases the biological cells heat exposure time. Furthermore, frequent selection of optimum path (such as shortest route) and unavailability of the next-hop during data transmission is also a major cause of hotspot formation in the biological medium [20]. Thermal-awareness at the network layer has been extensively considered in Wireless Body Area Networks (WBANs) [21]. However, due to different characteristics of IBNNs and WBANs such as energy, transmission range, data rates, storage and computational resources, they are not well suited for unique constraints of IBNNs [22].
The effectiveness of IBNNs routing schemes require a comprehensive understanding of IBNNs thermal challenges along with the in-depth consideration of Nano Bio Sensors (NBSs) limitations in energy, storage, computation and transmission range. The unique feature constraints of these NBSs are illustrated as; NBSs are equipped with severely inadequate energy, with the maximum energy an NBS may harvest, it can transmit only 8 packets of 200 bits during its entire lifetime [9], which significantly reduces the number of messages that can be transmitted by an NBS. In addition to limited energy, the extremely short transmission range of these NBS (i.e., 10 mm) also limits the direct data transmission opportunities. In the context of storage and computational resources, the storage capacity of these NBS does not support maintaining a routing table at the NBS level. Therefore, topology-awareness and selection of an optimum forwarding next-hop are nearly impossible at NBS level [4], [8], [23].
The existing routing schemes in IBNNs [9]- [11], [24] considered only the inadequate energy and storage challenges in designing a routing protocol. To the best of our knowledge a routing protocol that also considers the thermal challenges of a routing protocol has not been proposed. This is the first routing scheme that explicitly considers the energy and thermal challenges of data routing in IBNNs. Our proposed scheme realizes the ultimate objective of saving maximum energy and preventing temperature rise by introducing a temporal correlation based routing scheme. Our proposed temporal correlation based routing scheme increased the transmission of updated information and avoid redundant data transmission for avoiding network resources consumption and heat generation. In addition, we have proposed a novel energy-efficient and thermal-aware forwarded data aggregation approach that facilitates the selection of forwarding hop without storing the topology information at the NBS level. It further aims at the aggregation of multiple forwarded packet transmission into one single copy for transmission to the destination for avoiding hotspot formation and unnecessary energy consumption. The primary contributions of our proposed work are underlined as follows: 1) Our proposed routing protocol is based on temporal correlation to ensure the temperature-aware energyefficient periodic and on-demand data transmission to the external healthcare system. 2) In our proposed work, we aim to estimate the temperature rise in the biological medium using the Finite-Difference Time-Domain (FDTD) method. Our proposed scheme considers the two primary causes of heat generation in the biological medium, which are described as antenna radiation and molecular absorption noise temperature. 3) We proposed a novel method for data forwarding and aggregation, the intermediate NBSs aggregate the forwarded data for transmitting one single copy to the destination. In addition, our proposed forwarded aggregation methods enable the transmission of critical data with high priority for reducing delay and improving the accuracy of the critical information. We conducted extensive simulation using the Nano-SIM tool to evaluate the performance of our proposed scheme. The obtained results validate that our proposed scheme enables energy-efficient and controlled temperature rise routing in IBNNs as compared to the state-of-theart schemes. The rest of the paper is organized as follows: In section II, we discussed the existing scheme in detail. Section III comprehensively demonstrate our proposed temperature-aware energy-efficient routing scheme. Section IV presents performance evaluation, simulation results and discussion of the proposed routing scheme. In Section V, we draw up the conclusion. VOLUME 8, 2020

II. RELATED WORK
The last decade has witnessed an emerging interest in the development of communication protocols for IBNNs. Research communities have focused their attention on investigating the primary challenges of IBNN (such as channel modeling and designing novel communication protocols) for broadening the scope of IBNNs applications. This section deals with the brief discussion of the thermal challenges associated with EM-based communication in the THz band and the limitations of existing EM-based routing schemes to explore the opportunities and challenges of enabling EM-based routing in IBNNs.

A. THERMAL EFFECTS OF EM-BASED COMMUNICATION IN THE THz BAND
Nanoscale communication using EM waves allows NBSs to communicate in a distributed manner to accomplish diverse responsibilities. However, EM wave propagation in the THz band is severely challenged by the molecular absorption phenomena. Molecular absorption is caused by the movement of the molecules present in the biological medium that affects the medium in terms of attenuation and thermal noise. Molecular absopption noise phenomena can be explained as; the emission of EM waves at specific frequencies in the THz band excites the molecules present in the medium. The excited molecule has an internal vibration, which means the atoms have periodic motion while the movement of the molecule is translational and rotational. Due to this vibration, a part of the energy of the transmitted EM wave is converted into kinetic energy or, from a communication perspective, is simply lost. Subsequently, the internal vibration of the molecules cause the emission of EM radiation at the same frequency as of the propagating incident wave. The amount of radiation that is capable of propagating through the absorbing medium is defined as transmittance, which can be calculated using BeerLamberts law, given in (1): where f represents the frequency of the EM waves propagating at a distance d, and β b denotes the molecular absorption coefficients. Jornet et al. first determined the value of β b for gas molecules in [14], which depends on the composition of the medium. Later, Elayan et al. [25] followed the same approach for calculating β b for IBNNs communication using (2): where λ w represents the effective wavelength defined as λ n , and n , n are real and imaginary parts of the tissue's refractive index n. Using the calculated value of the molecular absorption coefficient, the transmittance of the medium is determined as given in [25].

B. EM-BASED ROUTING SCHEME FOR IBNNs
This section briefly discusses the existing EM-based routing schemes to shed light on the limitations of the existing schemes and to emphasize the need for new routing protocols for IBNNs.
The flooding scheme proposed in [26] aimed at conserving energy resources of the IBNN by controlling the direction of the forwarded packets. The selective flooded scheme achieved the objective of energy conservation by avoiding duplicate packets transmission. In [9], a protocol stack was proposed for minimizing energy consumption in IBNNs, the presented greedy scheme selected only a single node for transmitting answer message based on the remaining energy. Selection of a single node for transmitting answer messages supported in reduced energy consumption of the network. Another research effort that attempted to save energy is presented in [10], which employed nanocluster controllers for aggregating data with minimum energy consumption from NBSs. Data transmission to the nanocluster located at the minimum distance reduced data transmission distance and save energy consumption. In [27], the authors examined the energy harvesting options available for NBSs for improved network lifetime. The authors of [11], introduced a fuzzy logic and bio-inspired firefly algorithm based routing scheme for reducing energy consumption during the data reporting process. The optimized selection of NBSs using a bio-inspired firefly algorithm led to an improved network lifetime.
The brief discussion of existing schemes confirm that the primary focus of the state-of-the-art-schemes is on the energy consumption challenges of IBNNs. A routing scheme that also considers the thermal effects of data routing in IBNNs has not been proposed. To the best of our knowledge, this is the first scheme that considers the thermal and energy challenges of IBNNs.

C. EXISTING THERMAL-AWARE ROUTING SCHEME FOR WBANs
The objective of this section is to discuss the existing thermal-aware routing scheme for WBANs to demonstrate the importance of thermal-aware routing scheme. Furthermore, this section also stresses the reasons behind the unadaptability of the existing thermal-aware WBANs schemes for IBNNs. Tang.et.al proposed the Thermal-Aware Routing Algorithm (TARA) in [28]. In TARA, nodes estimate the surrounding temperature by monitoring the activities of their neighboring nodes. Based on the surrounding temperature, the node avoids data transmission in the hotspot area. In another work, Least Temperature Routing (LTR) is proposed [29]. In LTR, the node forwards the packet to the coolest neighbor to avoid hotspot formation while reducing the delay. In another effort to prevent temperature rise, Least Total-Route Temperature (LTRT) routing is performed [29]. LTRT is a hybrid routing scheme that considers the characteristic of the LTR and shortest hop routing algorithm for selecting the next-hop. In [30], another Hotspot Prevention Scheme (HPR) is introduced that is based on the characteristics of LTR and ALTR. The introduced scheme minimizes the delay while preventing hotspots. In [31], Bag et al. proposed a routing algorithm for network of homogeneous and Id-less biomedical sensor nodes (RAIN). RAIN avoids hotspots formation by using a probabilistic approach for next-hop selection. In [32], the authors proposed weighted Based Energy and Temperature aware Routing Protocol (WETRP). WETRP assigns equal weights to residual energy, temperature and link status to distribute data traffic evenly for conserving maximum energy resources, reducing delay and controlling temperature rise.
According to the brief discussion of existing thermal-aware WBAN schemes, we conclude that in WBAN schemes, nodes have enough energy, storage, and computational resources for monitoring the neighboring nodes activities, maintaining a routing tables and accordingly selecting an optimum nexthop. While in IBNNs, NBSs do not have enough resources for storing routing tables and computing next-hop. Therefore, monitoring activities of neighboring NBSs and subsequently selecting the next hop is an over exhausting task for energy constraint NBSs. Moreover, other characteristics of IBNNs (such as transmission range and data rate) also require extensive modifications of existing WBAN schemes before their adaptation for IBNNs. In Table 1, we further highlight the differences between our proposed work and existing schemes to shed more light on the contribution of our work.

III. TEMPORAL CORRELATION BASED TEMPERATURE-AWARE ENERGY-EFFICIENT ROUTING PROTOCOL FOR IBNNs
In this section, we aim to propose temporal correlation based temperature-aware energy-efficient routing scheme, which VOLUME 8, 2020 explicitly considers the peculiar requirement of IBNNs. The primary objective of our proposed scheme is to prevent temperature rise and reduce energy consumption during transmission of data from NBSs to nanorouter. To address the routing challenges of IBNNs, our proposed routing protocol consists of two different data retrieval mechanisms; named as temperature-aware energy-efficient periodic data retrieval scheme and temperature-aware energy-efficient on-demand data retrieval scheme. Our proposed temperature-aware energy-efficient periodic data retrieval approach aims at minimum consumption of energy resources and controlled heat generation during collection of periodic data from the NBSs. Whereas, the goal of our temperature-aware energy-efficient on-demand data retrieval scheme is to take into account the urgent need for reporting physical parameters in an emergency. Therefore, in the on-demand data retrieval approach, when a nanorouter receives a request message, it selects an optimum number of NBSs for generating response messages. The complete demonstration of proposed NBSs selection mechanism is further discussed in the section III-E. In addition, the Fig. 1 also highlights the working of our proposed scheme.

A. SYSTEM ARCHITECTURE
In this work, we consider a healthcare monitoring system that consists of an IBNN and an external healthcare monitoring system. IBNN encompasses two different kinds of nanoscale devices (i.e., NBSs and nanorouter), which are implanted inside the human body. Implanted NBSs are severely constraint in terms of energy, transmission range, storage and computational resources, whereas nanorouter has more resources than NBSs. Therefore, nanorouters handle more complicated operations, including temperature estimation, data aggregation, and transmission of aggregated information to the external healthcare system using nanointerface. Nanointerface establishes the connection between IBNN and the external healthcare system for remote data transmission via the Internet.

B. NETWORK MODEL, ASSUMPTION AND DEFINITIONS
The considered IBNN is modeled as undirected graph G = (V , E). V is the set of NBSs (NS) and nanorouter(NR) such that (NS, NR) ∈ V . E is the set of edges connecting vertices (i.e. NS and NR). In this work, we assume that NBSs are homogenous, which are deployed using a constant mobility model. Whereas, nanorouter maintains a fixed position.
Definition 1: The temporal correlation proportion is based on the reading reported by the NBS (NS i ) during last periodic data interval denoted as PR (t(i−1)) tran and the value observed by the NBS (NS i ) for the current periodic interval expressed as CR (t(i)) obs . Temporal correlation proportion value determine the correlation between the current observed and previous transmitted reading using (3).
where CR (t(i)) obs and PR (t(i−1)) tran denotes the value observed in the current periodic interval and the value transmitted in the previous periodic interval, respectively. Definition 2: A location µ in the event area A is designated as a hotspot, if location µ has a temperature rise greater than the specified threshold (i.e, the normal temperature of the body).
Nanorouter estimates the change in temperature by monitoring the activities of the NBSs in location µ. Let us consider that location µ is covered by n number of NBSs. The NBSs located in the location µ can be same and different at different time instances (i.e. due to movement of NBSs). NBSs positioned at location µ at various time intervals can be expressed as After each request rate interval, nanorouter calculate the temperature rise in location µ. The rise in temperature is a function of power radiated by n number of NBSs and molecular absorption noise due to heat absorption in the biological medium.

C. TOTAL TEMPERATURE RISE ESTIMATION
In this section, we aim to provide a detailed explanation of the temperature rise estimation process. In our proposed protocol, nanorouter calculates the increase in the temperature by considering the two main contributors of heat generation in the biological medium (i.e., antenna radiation and molecular absorption noise temperature). NBSs radiate EM waves into the biological medium for transmitting data, the radiated energy is absorbed by the surrounding biological cells that emit this absorbed energy as heat into their surroundings. Consequently, emitted energy leads to increased molecular absorption noise temperature, which increases the observed temperature of the biological medium. Temperature rise at location µ is calculated using diffusive heat equation [32], expressed in (4). We have assumed the biological tissues to be homogeneous without any sharp edges.
where the mass density and specific heat capacity of the heated particle are denoted using ρ and C p , respectively. The rate of change of temperature is expressed as ∂T ∂t , K denotes the thermal conductivity and T mol shows the change in temperature caused by molecular absorption noise temperature.
T rad represents the temperature increased due to antenna radiations and Q heat is the rate of generated heat energy per unit volume. Q heat can also be described as the energy source formed due to EM absorption in the particle [33], [34].

1) ANTENNA RADIATIONS
NBSs equipped with point dipole antenna are the primary source of heat generation in the biological medium. The length of point dipole antenna is denoted using l, and the dipole movement I o l is set to 1mA. The electric field intensity of a point dipole antenna is determined using the geometry of the radiating antenna and its current distribution. Illustration of proposed protocol model: Before periodic data retrieval nanorouter estimates increase in the temperature and avoid data collection from the hotspot areas. Those NBSs that are located in the normal temperature rise area transmit the periodic data based on temporal correlation decision. Whereas for on-demand data retrieval, nanorouter nominates those NBSs that are located in the normal temperature rise area and have maximum selection score. Subsequently, nanorouter transmits the aggregated data to the external healthcare system.
A half-wavelength dipole antenna experiences maximum intensity at (θ = π 2 ) [18], which can be expressed using (5): where the angle between the antenna axis and the vector to the observation point is denoted as θ, the distance to the antenna, and the average power radiated by the antenna is expressed using d and P o . The average radiated power P o can be calculated using (6): where R o denotes the radiation resistance of the antenna, determined using (7): where l is the length of antenna, the total heat generated in a location µ is determined by combining the power radiated by VOLUME 8, 2020 n number of radiating antenna that emit EM waves, as given in (8): Power rad n (8) where the coefficient Power rad n represents the power radiated by N antennas.

2) MOLECULAR ABSORPTION NOISE TEMPERATURE
The energy absorbed by the surrounding cells due to exposure of EM waves leads to an increase in molecular absorption noise temperature. The rise in temperature is determined using the heat model developed in [18]. The heat generated from the N par − 1 particles located at the nearest distance d n from the heat emitting source d s (i.e. antenna) can be rewritten as [18], using (9): whereK = k+k o 2 , and Q par is the heat produced by the N par that absorbed the heat and became the heat sources. The overall change in temperature is estimated by determining the heat produced in the medium. According to the law of conservation of energy, the heat generated (Q in ) is equivalent to heat dissipated (Q out ), hence change in temperature can be calculated using (10): (10) where Q in and Q out shows heat generated and heat dissipated by the heat particle, given in (11) and (13): where I o is the light intensity and σ abs is the absorption cross section calculated as (12): where α is the density of the heat particles N o in volume V o and β b is the molecular absorption coefficient defined in (1). Heat dissipated Q out can be calculated as (13): Now we estimate the temperature rise in the location µ can be following the similar approach used in [18], [33]. We considered the surface area of the location µ is a small cuboid of volume V , where identical heat particles N par are extended over the cuboid, the molecular absorption noise temperature can be calculated using (14): dx dy dz (14) where Q hp is the heat produced by identical particles. Further, expression of T p varies according to the typical array geometry and various illumination conditions [18], [33].

3) TEMPERATURE RISE CALCULATION USING FDTD
The total temperature rise is estimated using FDTD approach. FDTD is an EM modeling technique that discretizes the differential form of time and space. Due to the low complexity of FDTD in terms of time and memory, it is suitable for measuring heat gain [28]. Now we modify the bioheat equation given in (4) for calculating temperature rise in location µ using FDTD technique. The location µ is presented using three coordinates (x, y, z). The new modified heat equation for overall temperature rise is expressed in (15): where T t+1 is the temperature of the location µ (x, y, z) at time t+1, δ t and δ are discretized time step and the discretized space step respectively. From this equation, we can find the temperature of the location µ (x, y, z) at time t + 1, which is a function of the temperature at location µ (x, y, z) at time t, as well as a function of the temperature of surrounding , (x, y, z + 1), and (x, y, z − 1) at time i.

D. TEMPERATURE-AWARE ENERGY-EFFICIENT PERIODIC DATA TRANSMISSION
The implanted NBSs are responsible for generating periodic data packets at specified time intervals that consume significant energy resources of NBSs and require frequent emission of EM waves in the medium. To enable energy-efficient periodic data transmission with controlled temperature rise, we have proposed a temporal correlation based routing scheme. The purpose of employing temporal correlation is the nature of the continuous glucose monitoring applications, which exhibits a significant temporal correlation among the consecutive detected reading. In IBNNs, the transmission of temporally correlated information results in the collection of similar information at the expense of energy resources and increased heat generation. Therefore, we have proposed a temporal correlation based periodic data transmission mechanism, which ensures that NBSs do not frequently radiate EM waves for transmission of similar information. Since NBSs consume maximum energy during data communication and high data communication rate affects the networks in terms of increased antenna radiation exposure and energy consumption. Therefore, evading redundant data transmission support in achieving steady temperature rise in the biological medium and reduced energy consumption.

1) HOTSPOT IDENTIFICATION
Due to the limited resources of NBSs, they are unable to efficiently estimate the change in their surrounding 75912 VOLUME 8, 2020 temperature. Therefore, nanorouters are responsible for enabling temperature-aware parodic data transmission. Nanorouter monitors the change in temperature and prevents parodic data retrieval from the hotspot areas. Nanorouter identifies hotspot areas by estimating temperature rise using (15) after every parodic data transmission interval and broadcast information about the hotspot areas. Any location µ in the coverage area that has estimated temperature rise above the allowed temperature threshold is excluded from the generation of periodic data packets. The temperature threshold is set according to the average temperature of the body. The restriction of parodic data transmission avoids any further antenna exposure on biological cells in the hotspot areas and provide enough time to the heated region to stabilize its temperature.

2) PERIODIC DATA RETRIEVAL FROM NORMAL TEMPERATURE RISE AREAS
NBSs that are located in the areas with normal temperature rise are allowed to generate parodic data packets. The complete mechanism of temporal correlation based data transmission decision and data transmission process is elaborated in this section.

a: TEMPORAL CORRELATION BASED DATA TRANSMISSION DECISION
To eliminate similar data transmission, NBSs decide about the data transmission by determining the temporal correlation proportion using (3). Based on temporal correlation proportion, we have constructed the following binary hypothesis given in (16): where φ is the Temporal Coherency Threshold (TCT). If the value of the temporal correlation proportion is higher than the TCT, then the data transmission decision is positive, which means that NBS has updated detected value that needs to be transmitted. Whereas, if R(T ) is 1, then the observed reading is similar, and the detected value is suppressed.

b: DATA TRANSMISSION
Once data transmission decision is made, NBSs with the positive updated information (active NBSs) transmit data to the nanorouter via a single-hop or multi-hop method. Active NBSs broadcast announcement messages to notify neighboring NBSs about their transmission. The announcement message contains the information about the NBS ID and its respective layer based on the closeness degree of the NBS (i.e., the distance between a nanorouter and the NBS).

radiusofNBS). The width of each layer corresponds to the transmission radius of the NBS to enable direct data transmission from the lower layer to the intermediate higher layer. The highest layer is expressed as the α r , which is the nearest layer from the nanorouter.
The Broadcast of announcement messages provides an opportunity to the topology un-aware NBSs about the possibility of data transmission through an optimum forwarding next-hop. NBSs select only those NBSs for forwarding periodic data that have broadcast the announcement messages. The twofold objective of our proposed forwarded data aggregation scheme is elaborated as; first, excluding the involvement of inactive NBSs in data forwarding for saving their energy that will be consumed in data forwarding. Second, avoiding transmission of multiple forwarded packets by combining the packets at the intermediate forwarding NBSs and transmitting one single copy to the destination for conserving energy and preventing heat generation.

3) PERIODIC DATA RECEIVAL FROM HOTSPOT AREAS
Based on the announcement of nanorouter about hotspot areas, NBSs located in the hotspot areas cannot transmit the periodic data even if they have updated information. NBSs only generate periodic messages if they have detected high priority information that cannot be suppressed. The NBSs that transmit the high priority information is overheard by the NBSs, which are located in the close vicinity. Correspondingly, other NBSs with the high priority information stop their transmission with the objective of saving biological tissues from overexposure of EM waves. The data transmission mechanism from the hotpot area is further demonstrated in Algorithm 2.

4) PERIODIC DATA AGGREGATION AT THE INTERMEDIATE NBSs
After receiving data from the lower layer, the intermediate forwarding NBSs perform low complexity operations to VOLUME 8, 2020 combine the data packets for transmitting a single data copy to the destination. Our proposed forwarded data aggregation method combines the data based on the priority of the transmitted packet and forwarding similarity threshold. (17):

Definition 4: The periodic reading received at the intermediate NBS for the interval (t(i)) denoted as (PRR t(i) ) and the periodic reading detected for the interval (t(i)) at the intermediate NBS denoted as (PRD t(i) ) are correlated if forwarding correlation proportion is less than or equal to forwarding similarity threshold, forwarded data correlation can be calculated as
where PRD t(i) and PRR t(i) denotes the periodic reading detected and received at the intermediate NBS, respectively. While σ represents the forwarding similarity threshold. In this work, we consider the value of σ to be 15% that can be changed according to a specific application. The aggregation of the data is performed according to the forwarding similarity threshold and data priority level of the source and intermediate NBS.
Definition 5: The priority of the data is set according to the characteristics of the application. In our proposed work, we have defined the data criticality level according to hypoglycemia condition for the low and high level of glucose given in [35]. For a low glucose level, the critical levels are defined as low glucose level 1 (LGL1) and low glucose level 2 (LGL2). The glucose range of LSL1 is given as (70 mg/dL < Glucose ≥ 54mg/dL), which is categorized as low-level hypoglycemia alert that requires instant monitoring. In LSL2, the glucose reading of < 54 mg/dL signifies very low-level hypoglycemia and requires immediate action. While the priority of high glucose level is classified as high glucose level 1 (HGL1) and high glucose level 2 (HGL2). HGL1 has reading (250 mg/dL ≤ Glucose > 180 mg/dL), considered as the elevated high glucose level and require immediate monitoring. Whereas glucose value of greater than 250 mg/dL is indicated as the HGL2 necessitating immediate action.
The mechanism of data aggregation based on data priority level (i.e. LGL1, LGL2, LGL3 and LGL4) and forwarding similarity threshold is demonstrated below in detail: 1) Intermediate NBS and source NBS both have data priority of the same level: The Intermediate NBS calculates forwarding correlation proportion using (17) and compares it with forwarding similarity threshold. In this case, let intermediate NBS A and source NBS B both has same priority level and FDC(proporation) < σ , then it is comprehended that NBS A has reading that is similar to the reading received from NBS B. Thus, average operation is performed to combine the readings of NBS A and NBS B. Aggregation from multiple sources that have transmitted almost the similar readings is performed using (18):  (19): where PRR LPL k denote the low periodic reading received. After intermediate NBS has received all the readings, intermediate NBS assign low weightage (ω 2 ) to low priority readings given in Equation (19) and high weightage to high priority readings calculated in 75914 VOLUME 8, 2020 equation (18), such that the assigned weights ω 1 + ω 2 are equal to 1. Finally the accumulated reading is calculated using (20).

E. TEMPERATURE-AWARE ENERGY-EFFICIENT ON-DEMAND DATA RETRIEVAL
Diagnosis and treatment of physiological parameters also demand instantaneous collection of detected readings. Therefore, we have introduced an on-demand data retrieval mechanism; the objective of our proposed scheme is to enable instant transmission of data to the healthcare system when a request is made. The on-demand data transmission approach outlines our proposed scheme more proactive about the patients health condition. To realize the goal of our proposed approach, nanorouter handles the immediate requirement of diverse healthcare parameters by selecting the optimum NBSs for generating on-demand answer packets from those regions that are not experiencing temperature rise. The proposed selection standard is regulated by nanorouter that nominates NBSs based on their most recent transmission interval, estimated residual energy and distance from nanorouter. The reason behind our unique selection parameter of the most recent transmission interval is to increase the probability of receiving updated information as those NBSs that have recently transmitted updated information are more likely to have meaningful information. While the other selection parameter (i.e., residual energy and distance) strengths selections of those NBSs whose participation will not cause the death of the network and reduces multi-hop transmission. The mathematical expression for the aforementioned selection is given in (21): where MRTI NS , E R , and D NS→NR represents the most recent transmission interval of the NBS, residual energy, and distance of NBS from nanorouter, respectively. The nanorouter nominates those NBSs for transmitting answer messages that have a maximum Selection points .

F. ENERGY CONSUMPTION ANALYSIS
The objective of presented section is to perform a detailed energy consumption analysis of IBNN by considering the number of messages transmitted and received by nanotransceivers using the TS-OOK modulation scheme at the physical layer [36]. Based on the TS-OOK modulation scheme, the energy expended during transmission of a pulse is 1 pJ (i.e., E tx = 1 pJ ), whereas the energy consumed during the pulse reception is 0.1 pJ (i.e., E rx = 1 pJ ). Accordingly, (22) and (23) provide the amount of energy consumed during transmission and reception of x number of bits.
where γ represents the transmission probability of symbol 1 in the x streams of bits, and its weight is generally set to 0.5 as 0 and 1 both have the same probability of occurrence. Whereas, during the reception process, the entire stream of bits is considered. Accordingly, the amount of energy consumed after rth periodic interval is expressed as (24): where NT (M AM ) and NT (M AN ) shows the total number of transmitted answer message and announcement messages. Whereas, NR(M AM ), NR(M AN ) and NR(M RM ) denote total number or received answer messages, announcement messages, and request messages respectively. The amount of energy harvested during the periodic time interval (t) is expressed using E H . In the context of the energy harvesting mechanism, the technical limitations of nanonetworks do not support conventional energy harvesting mechanisms such as solar energy, wind power, and underwater turbulence. To address these energy harvesting challenge, novel schemes are proposed for recharging nanoscale devices. One of the prominent techniques of energy harvesting is the introduction of a piezoelectric nanogenerator [37]. The piezoelectric nanogenerator is composed of an array of ZnO nanowires, a reflecting circuit, and an ultra nanocapacitor. The nanowires compression produces an electric current between the ends of nanowires that recharge the ultra nanocapacitor. The mechanical vibration methods such as air conditioning and heartbeat provide the compressed release cycle. The frequency of the compressed release cycles f c provides the time for recharging the ultra nanocapacitor. Based on the energy harvesting model, the voltage of the charging nanocapacitor after β compressed cycles can be rewritten as [37], using (25): where C cp denotes the total capacitance of the ultra nanocapacitor, V g shows the generator voltage and Q represents the harvested charge per cycle. The typical values of these quantities C cp , Q and V g are 9 cF, 6 pc and 0.42 V respectively. According to [37], after β cycles the energy E cp stored in ultra-nanocapacitor is given using (26): where C cp represents the capacitance of ultra-nanocapacitor, V g corresponds to the voltage of the generator and β cycles demonstrates the number of cycles necessary to reach the energy level given by (27): where V g is the generator voltage, C cp is the capacitance of ultra-nanocapacitor and Q represents the harvested charge per cycle. Subsequently, the amount of energy harvested during periodic time interval (t) is obtained by determining the number of cycles β required for charging the ultrananocapacitor, as expressed in (28): where cycles β correspond to number of cycles given in (27) and cycles t represents the number of cycles during the periodic time interval (t).
In the view of the amount of energy harvested and consumed in packet transmission and reception process, the remaining energy of the network is obtained using (29): where E o denote the initial energy of the network and E t con.
represents the amount of energy consumed till the periodic time interval (t).

IV. PERFORMANCE EVALUATION
This section evaluates the proposed routing scheme in IBNNs to highlight the practical utility of the proposed protocol for healthcare applications. To shed light on the efficacy of the proposed scheme, we have performed an extensive comparison with the flooding scheme and TARA. The flooding scheme is considered as the benchmark scheme for comparing new routing schemes for IBNNs. In the flooding scheme, the direction of the forwarded packets is controlled to save energy and bandwidth [26]. Whereas, In TARA each node monitors the activities of its neighboring nodes for preventing forwarding data in a hotspot area [28]. The reason behind using TARA is the unavailability of a thermal-aware scheme for IBNNs. Furthermore, the detailed comparison with TARA also underlined the significant energy consumption and delay issues of TARA regarding its adaption in IBNNs.

A. NANO-SIM SIMULATIONS
We have performed detailed simulations using the Nano-Sim tool. The Nano-Sim tool is an emerging tool integrated into NS-3 for evaluating the performance of EM-based communication schemes in the THz band [26]. We consider an IBNN, which is expended on a radius of 10 mm with a varying number of NBSs (i.e., 100 and 150) and one nanorouter. The implanted NBSs are moving in the network area using a constant mobility model, whereas, nanoroute maintains a fixed position at the center of the network during the entire simulation time. The moving NBSs are equipped with a sensing unit for monitoring the physiological parameters. NBSs transmit detected physiological readings to the nanorouter at the periodic intervals and on the demand of the external healthcare system. At the physical layer, we have adapted the same settings as given in [36]. The size of answer messages is set to 176 bits. Whereas, the size of the request message and announcement message flowing in the network is set to 36 bits. For providing a fair comparison, we have used the same simulation parameters for all the schemes. Furthermore, table 2 summarizes the tissues' properties for IBNNs along with the simulation parameters. We consider the same properties of the tissues as adapted in [18] for photothermal modeling of IBNNs.

B. PERFORMANCE METRICS
We use the following metrics to evaluate the performance of our proposed scheme. VOLUME 8, 2020  • Average temperature rise: Average temperature rise represents the average increase in the temperature observed in the network.
• Average remaining energy: This represents the average energy left in the network.
• Alive NBSs: It shows the number of alive NBSs in the network.
• Number of notifications: It corresponds to the total number of readings reported by the NBSs.
• Average end-to-end delay: It expresses the average endto-end delay experienced by a packet.
• Packet delivery ratio: It represents the total number of packets transmitted to the total number of packets received.

C. SIMULATION RESULTS
This section provides a comprehensive discussion of the obtained results to highlight the potential of the proposed scheme for enabling continuous healthcare monitoring. The presented results have 95% confidence interval computed after 10 iterations to reduce statistical fluctuations impact.  Fig. 4 demonstrates that increased traffic rate also increases the heat generation. Therefore, from Fig. (4a,4b), we can observe that in WFCS, the increase in the temperature is 40% more than our proposed scheme. Our proposed scheme and WFCS, both experienced the maximum temperature rise at 10% TCT due to high packet generation rate. However, our proposed scheme still observed a low temperature rise due to our efficient forwarded data aggregation scheme, which prevents the biological medium from overheating. Whereas, the flooding scheme experienced the maximum temperature rise for all the TCT values (i.e., 10%, 15%, 20%, 25% and 30%), due to thermal-unawareness during routing decisions. While TARA experiences more temperature rises due to higher packet generation and packet forwarding rate, which ultimately increases the temperature. Fig. 4a depicts that once the network is dead the flooding scheme and TARA experience no further change in the temperature. To shed more light on the contribution of our proposed scheme on low average temperature rise, we have also compared the average temperature rise observed at the end of simulations (see Fig. 4c). Fig. 4c, demonstrate that our proposed scheme realizes the objective of controlled temperature rise by enabling temperature-aware routing and controlled traffic flow.
2) AVERAGE TEMPERATURE RISE COMPARISON OF OUR PROPOSED SCHEME WITH THE DENSITY OF 150 NBSs AND INCREASING TCT OVER TIME Fig. (5a,5b) shows that for increasing densities of NBSs, our proposed temperature-aware routing scheme maintains a steady temperature rise whereas WFCS results in a high temperature rise due to forwarding of all packets without any aggregation at the intermediate NBS. The flooding scheme and TARA experience more rise in temperature than our proposed scheme due to the high rate of traffic generation and data forwarding. In Fig. 5c, the average temperature rise at the end of simulations is presented to highlight the impact of our proposed scheme further. From Fig. 5, we can observe that our proposed scheme achieves the objective of hotspot prevention and maintains the lowest average temperature rise.
3) AVERAGE REMAINING ENERGY COMPARISON OF OUR PROPOSED SCHEME WITH THE DENSITY OF 100 NBSs AND INCREASING TCT OVER TIME Fig. 6a depicts that the increasing value of TCT leads to lower consumption of energy resources. In WFCS, the maximum energy is consumed for 10% TCT. Whereas, for our proposed scheme even at 10% TCT energy consumption is lower than WFCS for all the TCT values (i.e, 10%, 15%, 20%, 25% and 30%). When TCT is set to 30%, our proposed scheme experienced the lowest amount of energy consumption due to suppression of the maximum number of notifications. 75918 VOLUME 8, 2020   Fig. 6b, it is evident that our proposed scheme improves 70% to 80% better residual energy as compared to both TARA and the flooding scheme. The reason behind improved remaining energy is our proposed temporal correlation based data transmission decision that suppresses the redundant data transmission and enables transmission of updated information, which ultimately results in better remaining energy.

4) AVERAGE REMAINING ENERGY COMPARISON OF OUR
PROPOSED SCHEME WITH THE DENSITY OF 150 NBSs AND INCREASING TCT OVER TIME From Fig. 7, it is evident that the increased NBSs density improves the residual energy of the network. Our proposed scheme achieves the maximum remaining energy for all the TCT values (i.e, 10%, 15%, 20%, 25% and 30%). Whereas in WFCS, energy delation is more rapid than our proposed scheme. The comparison of our proposed scheme with WFCS illustrates that our proposed forwarded data aggregation scheme has a great impact on energy conservation. Furthermore, the comparison of our proposed scheme with the flooding scheme and TARA presented in Fig. 7b depicts that even for the increased densities of NBSs, the remaining energy is not substantially improved. TARA experiences the lowest remaining energy due to high data forwarding and packet retransmission rate that leads to the maximum consumption of energy resources. Fig. 8 illustrate that our proposed scheme results in the maximum number of alive NBSs for the increasing densities of NBSs (i.e., 100 and 150) at all the time instances. For both densities of NBSs, at time 0.5 seconds, for 30% TCT, our proposed scheme has almost 85% alive NBSs. While VOLUME 8, 2020  for 10% TCT, almost 60% of NBSs are alive. WFDAS also results in a better alive number of NBSs as compared to the state-of-the-art-schemes. From Fig. 8b, we can see that the flooding scheme and TARA have the lowest number of alive NBSs. In TARA, NBSs died more rapidly due to monitoring of the neighboring NBSs and increased rate of packet  retransmission for avoiding hotspots. From Fig. (8a,8b), we can observe that our controlled packet generation and efficient data forwarding strategy significantly increase the lifetime of NBSs. Fig. 9 shows that when the TCT increases, the number of notifications reported by our proposed protocol decreases for the increasing densities of NBSs (i.e., 100 and 150). From Fig. 9a, we can see that when the NBSs density is 100 and TCT is 10%, maximum notifications are reported, whereas, for 30% the minimum number of notifications are reported. While the number of notifications in the flooding scheme and TARA remains fixed since all the reading detected by the NBSs are reported without any redundant data transmission control mechanism. Similarly, in Fig. 9b, we can see that the number of readings reported by our proposed scheme for the density of 150 is almost similar to the ones presented in Fig. 9a. This is because whenever the temporal correlation proportion is above the TCT, the data is notified. Therefore, for the lower value of TCT, our proposed scheme reports the maximum readings.  Fig. 10, we can observe that our proposed scheme experiences minimum average-end to-end delay for increasing densities of NBSs (i.e., 100 and 150) even at 10% TCT. The reason behind the negligible difference in the average end-to-end delay is the optimized selection of next-hop. Our proposed scheme experiences the lowest average endto-end delay for 150 NBSs since when the NBSs density is increased, the number of potential next-hop NBSs also increases, which leads to reduced delay. Whereas, WFCS experiences more average end-to-end delay as compared to   our proposed scheme. The reason for the increased delay in WFCS is due to the forwarding of all packets without any aggregation at the intermediate NBS. Accordingly, the high packet forwarding rate ultimately leads to an increased average end-to-end delay. In the same manner, the flooding and TARA both show the maximum average end-to-end delay for increasing densities of NBSs due to increased packet generation and forwarding rate. Fig. 11 visually compare the packet delivery ratio of our proposed scheme with WFCS, flooding and TARA schemes. From Fig. (11a,11b), it is evident that our proposed scheme achieve maximum packet delivery ratio for increasing density of NBSs. The high packet delivery ratio is due to minimum data congestion. Since WFCS transmits the forwarded packet without any suppression, therefore WFCS acheive almost 80% to 90% packet delivery ratio, which is still better than flooding and TARA scheme that have packet delivery ratio of less than 70% due to high data traffic.

8) PACKET DELIVERY RATIO COMPARISON OF OUR PROPOSED SCHEME WITH THE INCREASING DENSITIES OF 100 AND 150 NBSs AND INCREASING TCT OVER TIME
In view of the detailed discussion of simulation results, it is evident that suppression of correlated data during data generation and data forwarding avoid transmission of redundant information and regulate data traffic under control. Therefore, our proposed strategy outperforms exiting schemes and results in controlled temperature rise, reduced end-to-end delay, improved network lifetime and packet delivery ratio.

V. CONCLUSION
The proposed routing protocol explicitly considers the feature and thermal challenges of IBNNs for designing a temperature-aware energy-efficient routing protocol for IBNNs. Furthermore, our proposed novel forwarded data aggregation approach combines forwarded packets to transmit only single to the destination for saving energy resources and avoiding antenna radiation exposure time. The obtained results confirm that our proposed approach reduces energy consumption and prevent temperature rise as compared to the state-of-the-art-schemes. In addition, our proposed strategy also avoids congestion and achieve low average end-to-enddelay. According to the comprehensive results and discussions, we conclude that our proposed protocol enables safer routing in terms of heat generation in a biological medium and improves the lifetime of an IBNN.