5G technology, with its low latency and high capacity, has transformed healthcare delivery, especially in distant areas. 5G wireless networks in healthcare to demonstrate their potential benefits has been extensively studied ( Agiwal et al., 2016). How 5G might assist rural healthcare with connectivity problems, opening the door to telemedicine and remote operations ( Gupta et al., 2021). 5G network slicing enables low-latency telemedicine, crucial for distant healthcare ( Qin et al., 2020).

The impact of 5G technology on surgical and IoT healthcare applications have been investigated ( Park & Choi, 2021, Ahmad et al., 2022). A real-time 5G infrastructure designed for smart healthcare services was introduced ( Hossain et al., 2022). 5G technology enhances telehealth services in rural areas ( Patel et al., 2022).

New approaches of considering latency improvement have emerged. Fog computing resource allocation with predictive analytics ( Zhang et al., 2018). Load balancing and task scheduling aiming at latency reduction in healthcare applications using certain methodologies have been studied ( Wang & Wang, 2019, Hong et al., 2020). A dynamic resource allocation technique was suggested that considers data urgency and hence helps to lower latency in Internet of Things enabled healthcare systems ( Sharma & Singh, 2022). A fog computing multi-objective optimization framework including cost, energy consumption, and delay was presented ( Liu et al., 2021).

Despite being understudied, 5G and fog computing are increasingly gaining traction. 5G enhances healthcare fog computing by diminishing latency ( Jiang et al., 2021). Empirical studies executed to illustrate advancements in rural telehealth and remote monitoring ( Patel et al., 2022, Álvarez et al., 2023). Case study for demonstrating the functionality of integration in rural healthcare has been studied ( Xu & Zhang, 2023). The potential of 5G to enhance data processing in fog nodes for healthcare IoT applications was investigated ( Cheng & Zhang, 2022).

A balance of security, energy economy, and latency is clearly necessary. There is provisions to lower energy consumption and latency in fog-5G systems, enabling their efficient running in rural locations with restricted power supplies ( Liu et al., 2023, Tang & Wang, 2022).

The papers demonstrate number of gaps in the current literature: •

There has been a lack of practical implementation and testing in rural locations.

More adaptive and foresight algorithms specific to the specific requirements of the healthcare domain are needed. Further research in this area should be conducted on actual applications of such integrated systems in the rural healthcare environment. It is found that there is an upsurge towards highly sophisticated forecasting models that can deal with emergencies and also deal with medical data. The role of bringing such high level technologies in the undeveloped regions in terms of the effects they have on society. Lastly, with the solution of the latency issues, integration of 5G and fog computing will immensely change the rural healthcare. The reviewed literature provides a good foundation on which additional improvements can be achieved to ensure that healthcare services provided in rural communities are not only as fast and efficient, but even more so. The formulation of the problem in mathematical terms is aimed at approaching the latency reduction in fog-5G-based rural healthcare.

This research project will primarily aim to minimize the latency in the overall case of the rural health system by streamlining the workload of data processing between the fog nodes and the cloud service, using 5G as a connectivity medium. The formulated objective function is: L _total=w ₁.L _fog+w ₂.L _cloud

Where:

L total represents the total system latency.

L fog is the average latency for data processed by fog nodes.

L cloud is the average latency for data processed by the cloud.

w 1 and w 2 are the weights representing the proportion of data processed by fog nodes and cloud, respectively, with the constraint: w 1 + w 2 = 1

Let n be the number of fog nodes (assumed to be 15), d i be the distance from the IoT device to the nearest fog node, t process _ fog be the processing time at a fog node, assumed to be low due to localized processing, t 5 g be the latency introduced by 5G communication (estimated to be between 1-10 ms).

The fog node latency can be expressed as: L fog = 1 N ∑ i = 1 N ( t process _ fog + t 5 g + d i C 5 g )

Where:

C 5 g is the speed of 5G signal.

N is the total number of IoT devices (assumed to be 1000).

Let D be the distance from the rural area to the cloud server (significantly larger than d _i ), t process _ fog be the processing time in the cloud, which is higher due to centralized computing and t network be the latency due to traditional network connectivity to the cloud (assumed to be around 100 ms).

The cloud latency can be expressed as : L cloud = t process _ cloud + t network + 2 D C internet

Where:  C internet represents the average internet speed.

2.8.1 Data Urgency Constraint

Each data packet is assigned an urgency level u i (ranging from 1 to 10). Critical data (where u i > 5 ) must have latency below L threshold (20 ms).

The constraint is: ∀ i , where u i > 5 _, L i ≤ L threshold

2.8.2 Resource Utilization Constraints

Let R fog be the total processing resource available at fog nodes, R cloud be the total processing resource available in the cloud and R i be the resource requirement for processing data packet i.

The constraints are: w 1 . ∑ i = 0 N r i ≤ R fog

2.8.3 Bandwidth Constraints

Let B 5 G be the total bandwidth available for 5G channels (assumed to be 12 channels).

Let B i be the bandwidth requirement for transmitting data packet i.

The constraints are: w 1 . ∑ i = 0 N b i ≤ B 5 G

For cloud communication, traditional internet bandwidth ( B internet ) should also be considered, but with 5G connectivity, it is not a significant constraint. w 1 + w 2 = 1 , 0 ≤ W 1 ≤ 1 , 0 ≤ W 2 ≤ 1

By considering the important character of healthcare data, the available computational resources, and the real-time conditions of the network, this problem formulation aims to optimize latency so enabling the system to respond to the changing needs of healthcare in rural areas.

The details of the proposed algorithms are summarized in Table 1.

2.9.1 PRA Algorithm Outline: The proposed Predictive Resource Allocation (PRA) method is presented in Algorithm 1, which dynamically allocates network resources based on predicted demand and current consumption patterns. -

Input: The present status of the network, previous data consumption, priority of patient information, Historical (H): 10000 runs, Existing Need: D _c, Time Frame (W): last 1000 runs

2.9.2 LADR Algorithm Outline: The detailed procedure of the Latency-Aware Dynamic Routing (LADR) mechanism used to select the optimal routing path based on latency constraints is demonstrated in Algorithm 2. -

Input (Λ: LatencyMap, 15 fog nodes, 12 5G channels): Data urgency, current network latency, fog node status.

2.9.3 ANS Algorithm Outline: The procedure for the proposed Adaptive Network Slicing (ANS) mechanism, which dynamically assigns network slices based on system load and priority requirements, is shown in Algorithm 3. -

Input: ϵ: (DataUrgency, 1 to 10), ϕ: (CurrentNetworkLoad, 0 to 1, Data packet characteristics, network load.

The flowchart shows the data handling technique of the suggested system model. Healthcare IoT devices begin the process with data collecting; classification and forwarding to a fog node for urgency evaluation follows. Depending on urgency, we either send data to the cloud over a 5G network for additional processing or handle data locally. Choosing where to handle the data considers dynamically both local and cloud processing options. Once processing is finished, the data is restored into the system for dynamic changes and system flexibility to maximize it once more. Seeking to lower delay in rural health systems, this flowchart effectively shows data flow and decision-making process. The detailed data handling mechanism and processing architecture of the proposed system are illustrated in Figure 1.

The F5GLO is used in this section to compare the results of the simulation of Local Server Model and Traditional Cloud Model. F5GLO proves its ability to significantly decrease the response time, with the maximum decrease of latency up to 87%. F5GLO is also characterized by a better energy efficiency and high throughput in addition to relying on more powerful technology, such as fog computing and 5G technology, to improve the performance of the rural health system.

3.2.1 Average Latency: The F5GLO technology demonstrates the 87 percent enhancement of the conventional cloud computing and dropping of the latency by 13 milliseconds. The mean latency of Local Server Model is 40 milliseconds and this is not as low in terms of latency reduction as F5GLO can provide when used in combination with 5G to process data locally. Conversely, the Traditional Cloud Model that is based on remote data processing has the highest average of 100 milliseconds of latency out of the three models. The comparison shows that F5GLO has significantly enhanced the responsiveness and latency of the rural health system. As shown in Figure 3, the F5GLO model achieves significantly lower latency compared to the Local Server and Traditional Cloud models.

3.2.2 Peak Load Handling: Local Server Model is better than the cloud-based technique, but latencies of 80 ms to the peaks suggests local processing limitations without integration to 5G. This demonstration indicates the difficulty in maintaining minimal delays in the situation when the demand on the service is growing. Traditional Cloud Model has the highest delay of 150 milliseconds due to use of centralized processing. By using the peak load management tool provided by F5GLO, rural healthcare centres will become more efficient and can be better equipped to deal with emergencies. Figure 4 presents the comparison of peak load latency across the F5GLO, Local Server, and Traditional Cloud models.

3.2.3 Latency Distribution: The results of the assessment have shown that the F5GLO model now whirs along with significantly faster speed of about 19.5 ms lag on average, compared to 120 ms, which was about 87% reduced. It retains its low-latency appeal even with the heavy load, and works in a variety of situations. The latency distribution is skewed to the left with approximately 90% of data packets received in less than 20 ms. The Local Server Model is faster than the Traditional Cloud Model but lagged behind F5GLO; nevertheless, the former is significantly faster than cloud-only solutions. In this configuration 90 percent of the packets arrive within 40 ms, and only a handful of them go after 80 ms. The delays in the Traditional Cloud Model skyrocket because of transmission and processing requests: approximately 90 percent of packets end up having an average delay of approximately 100 ms (maximum delay is approximately 150 ms). The latency distribution across the evaluated models is summarized in Table 2 and illustrated in Figure 5.

3.2.4 Average Throughput: F5GLO framework is more effective by having the average of approximately 0.077 packets/ms or 77 packets/s. This is the enhanced performance of the fog processing and high-speed 5G connections that reduce delays and accelerate the processing of data. F5GLO performs even better when we consider the latency and throughput jointly. F5GLO has lower latency and higher throughput, increasing the efficiency of data processing. The Local Server Model reduces the distance between source and processor, averaging approximately 0.050 packets/milliseconds and it is faster than pure cloud setups, yet, still, not faster than F5GLO. The Traditional Cloud Model reduces to a rate of approximately 0.010 packets per millisecond due to centralized processing and transmission latencies and creates higher latency as well. Distributed architecture of F5GLO, which has localized decision-making, maintains throughput approximately eight times compared to the conventional cloud solutions. This near combination of low latency and high throughput indicates F5GLO will provide a good solution to data-intensive healthcare applications where quick delivery is important and latency is a significant limitation. The throughput characteristics of the proposed and existing models are depicted in Figure 6.

3.2.5 Resource Utilization: The F5GLO model has a peak throughput of 0.100 packets/ms, which is more than the Local Server Model and Traditional Cloud Model at 0.075 packets/ms and 0.015 packets/ms respectively. Such throughput capacity maximum highlights the scalability of F5GLO and its ability to handle great data volumes in order to provide a timely response in the healthcare process. Implementation of mist nodes to process in F5GLO lessens the workload on the cloud and therefore enhances the ability to perform more activities or functions in this system. ANS technique enhances the use of bandwidth, thus, the throughput is increased and resources are used more efficiently in every model. Figure 7 and Table 3 present the comparison of resource utilization across the F5GLO, Local Server, and Traditional Cloud models.

3.3.1 Average Energy Use: F5GLO system consumes approximately 3.5 energy units per packet, hence it is more energy efficient in comparison to the Local Server Model and the Traditional Cloud Model. Local Server Model has a speed of approximately 4.0 units per packet. It is facilitating local processing yet consumes more power because there is poor data management in the 5G network. The Traditional Cloud Model consumes approximately 5.0 units per packet, the most energy consuming since processing occurs in remote data centers and data transfers require more power. F5GLO involves processing over fog nodes, and this implies that fewer data are transmitted to the cloud. This reduces the volume of information transmission, hence, improving energy efficiency and reducing latency. The simulation outcomes clearly show such benefits in efficiency and prove that F5GLO is more efficient in terms of managing energy consumption than the rest.

3.3.2 Energy vs. Latency Trade-off: The F5GLO model is also able to keep low power consumption in intense workload or when dealing with data streams of large volume. Using 7.0 energy units per packet, the F5GLO model will be used at peak throughput. Even though the incorporation of local processing, indeed makes the energy consumption higher, particularly, as the existing 5G networks are not optimized regarding energy efficiency, its consumption remains significantly lower than the one of the Local Server Model, which consumes approximately 10.0 units per packet under the same conditions. F5GLO is considered to be compatible with adaptive control. It also tries to balance the speed and energy usage with a better distribution of resources. The energy saving will be dependent on the location of the fog nodes and the distribution of the workload, and this can vary in various configurations. However, local data processing with the help of the fog nodes will always reduce additional data transmissions. This provides F5GLO with an improved trade-off between latency and energy consumption. Such a trade-off is essential in rural healthcare since timely and accurate information has to be deployed with careful power management, although the two tend to be conflicting ultimately. The energy–latency trade-off among the evaluated models is summarized in Table 4.

The outcomes of the simulation have shown that the reduction of latencies through the use of 5G of the network and the implementation of fog computing can enhance the healthcare services delivery to rural locations. In this paper, the discussion, observation, and implication in the three models are discussed. Since F5GLO paradigm cuts down the latency by 87 percent on the Local Server paradigm and the Traditional Cloud Model.

The F5GLO design has a process called Predictive Resource Allocation (PRA) that enables resource allocation on the basis of predictions. It implies that even a scaling up does not require the introduction of any latency, which is essential when more devices or patients are introduced into the system. Local Server Model can be scaled because it can process data locally but lacks dynamic resource management in comparison with F5GLO. Whereas the Traditional Cloud Model is capable of scaling up its data processing capabilities, it tends to increase its latency as it scales and demonstrates that this model cannot be used in dynamic and latency-sensitive settings such as healthcare. The scalability characteristics of the proposed and existing models are depicted in Figure 8.

The increased throughput also suggests that the system will be able to handle additional data or additional patients at the same time at the expense of the timeliness of service, therefore, ideal in scenarios such as mass casualty situations in rural environments where resources are typically scarce.

Although we aimed not only to reduce latency, but also to make our method energy efficient, it is another benefit level, especially in those regions where power costs may be sporadic or expensive. Latency and energy optimization Twin optimization is the full design of the F5GLO system.

ANS algorithm is capable of reducing emergency data delay and maintaining network performance to other services by creating network slices according to the urgency of data. This makes sure that it transmits important healthcare information with utmost urgency. The influence of network slicing on latency and service prioritization is depicted in Figure 9.