A bio-inspired, adaptive task outsourcing system for an energy-efficient IoT edge-cloud continuum in healthcare.

Abstract

A bio-inspired, adaptive task outsourcing system for an energy-efficient IoT edge-cloud continuum in healthcare, including: an IoMT sensing layer comprising one or more wearable or implantable medical sensors configured to collect physiological patient data and generate one or more processing tasks; an intelligent edge gateway that is communicatively coupled to the IoMT perception layer and configured to receive one or more processing tasks; and a cloud analytics layer that is communicatively coupled with the intelligent edge gateway, including the intelligent edge gateway: a criticality index module (CI module) configured to calculate a criticality index for each processing task, representing a clinical urgency level associated with the patient's physiological data; a cost estimation module configured to estimate, for each of a variety of eligible execution nodes, including at least one of: execution on a portable device, execution on an edge gateway, execution on a fog node, and execution on a cloud server, a predicted power consumption, a predicted end-to-end latency, and a predicted privacy risk; a bio-inspired metaheuristic scheduler configured to select an execution node from the multitude of eligible execution nodes based on a multi-criteria cost function that combines the predicted energy costs, the predicted end-to-end latency, and the privacy risk using weighting parameters that are dynamically adjusted depending on the calculated criticality index; and a privacy guardian configured to apply a privacy mechanism to the processing task when the selected execution node is outside a defined local trust boundary, where the selected execution node performs the processing task and generates an output or a warning message for transmission to a clinician interface, so that the system dynamically balances energy consumption, response latency, and data privacy based on the patient's clinical urgency and network conditions.

Description

[0001] The present invention relates generally to the fields of health informatics, the Internet of Medical Things (IoMT), and distributed computing architectures. In particular, the invention relates to a bio-inspired, adaptive task offloading system for energy-efficient and privacy-conscious processing in IoT edge cloud networks in healthcare.

[0002] The rapid growth of the Internet of Medical Things (IoMT) has led to the widespread use of wearable sensors, smart medical devices, and networked health monitoring systems for continuous patient observation. Devices such as ECG monitors, pulse oximeters (SpO2 sensors), temperature sensors, blood pressure monitors, blood glucose meters, and activity trackers generate continuous streams of physiological data. These systems are increasingly being used in hospitals, home care settings, elderly care monitoring, and telemedicine platforms. In traditional cloud-centric architectures, patient data collected by wearable devices is transferred to centralized cloud servers for processing, analysis, and storage. While cloud platforms offer scalable computing resources and advanced analytics capabilities, transferring raw data to the cloud presents several challenges.These include increased network bandwidth consumption, higher latency, dependence on a stable connection, and potential risks to patient data privacy. In time-critical medical events such as cardiac arrhythmias or sudden oxygen desaturation, delays due to network congestion or remote processing can jeopardize patient safety.

[0003] On the other hand, performing all data processing locally on portable devices is not a viable solution due to their inherent resource limitations. Portable and implantable medical devices are typically subject to strict limitations regarding battery capacity, computing power, and storage availability. Performing computationally intensive tasks such as real-time anomaly detection or predictive analytics can significantly reduce battery life and decrease device uptime. This creates a trade-off between responsiveness and energy efficiency. Edge and fog computing paradigms have emerged as intermediary layers between IoMT devices and the cloud infrastructure. Edge gateways, such as smartphones or bedside hub devices, offer higher computing power and are closer to the patient, enabling lower-latency processing.However, existing mechanisms for task allocation in such multi-layered architectures are often based on static rules or fixed scheduling guidelines, such as periodic data transfer or threshold-based forwarding. These approaches cannot dynamically adapt to changing patient conditions, network fluctuations, device power levels, and data privacy requirements. Furthermore, health data is highly sensitive and subject to strict regulatory and ethical constraints. Many existing systems lack adaptive data privacy control mechanisms that consider the sensitivity of raw medical data before it is transmitted across local trust boundaries.

[0004] To solve this problem, the present invention offers a bio-inspired, adaptive task outsourcing system for an energy-efficient IoT edge-cloud continuum in healthcare.

[0005] The system dynamically determines an optimal execution location for tasks involving the processing of health data.

[0006] The system offers a swarm intelligence-based metaheuristic scheduler configured to evaluate multiple potential execution nodes—including wearables, edge gateways, fog nodes, and cloud servers—and select a near-optimal execution path based on a multi-parametric cost function.

[0007] The system reduces end-to-end latency in time-critical clinical events by dynamically prioritizing low-latency nodes as a patient's clinical urgency increases.

[0008] The system extends the battery life of portable medical devices by offloading computationally intensive tasks, provided this is clinically safe.

[0009] The system has a criticality index (CI) module that is configured to assess a patient's urgency based on vital parameter thresholds, trend fluctuations, and anomaly values, and adjusts appointment scheduling priorities accordingly.

[0010] The system implements a cost function that takes into account the predicted energy consumption, the predicted execution latency, and the data privacy risk, with the weighting being dynamically adjusted based on the calculated criticality index.

[0011] The system offers an adaptive data protection mechanism configured to apply data protection mechanisms, including encryption, feature-only transmission, and privacy-compliant transformations, before transferring health data outside a local trust boundary.

[0012] The system provides a backup solution in case a node is unavailable, thus ensuring the reliability and continuity of task execution in healthcare.

[0013] The system offers a scalable framework that can be used in hospitals, home care, telemedicine platforms, and emergency care systems that require adaptable, energy-efficient, and privacy-compliant remote patient monitoring.

[0014] The present invention discloses a bio-inspired, adaptive task offloading system for energy-efficient, latency-aware, and privacy-compliant processing in IoT edge cloud networks in healthcare. The system is designed to intelligently distribute healthcare data processing tasks based on real-time contextual conditions across wearable medical devices, edge gateways, fog nodes, and cloud servers. The system comprises an IoMT sensing layer configured to acquire physiological patient data from wearable sensors, an intelligent edge gateway configured to receive and manage tasks generated from the acquired data, and a cloud analytics layer configured to perform large-scale analytics and long-term storage.

[0015] The intelligent edge gateway acts as a decision and control layer and houses the central adaptive scheduling mechanism. A criticality index (CI) module calculates a clinical urgency score for each task based on predefined thresholds for vital signs, trend fluctuations, and / or anomaly detection models. The CI reflects the medical urgency associated with the task and directly influences the task outsourcing strategy. A cost estimation module evaluates multiple potential execution nodes, including execution on wearables, edge gateways, fog nodes, and in the cloud. For each potential node, the system estimates the anticipated energy consumption, end-to-end latency, and data privacy risk associated with data transmission or processing.A multi-criteria cost function integrates the predicted energy, latency, and data privacy components using dynamically adjustable weights. The weight values ​​are determined as a function of the calculated CI.

[0016] If the CI indicates high clinical urgency, the system increases the weighting of latency to prioritize a rapid response. If the CI indicates lower urgency, the system increases the weighting of energy efficiency and data privacy. A bio-inspired metaheuristic scheduler, implemented using a swarm-based optimization approach such as ant colony optimization (ACO), particle swarm optimization (PSO), or whale optimization (WOA), scans the execution space of candidates and selects a near-optimal execution node based on the evaluated cost function. The scheduler adaptively updates its selection probabilities according to current network conditions, device power status, and the clinical context.Before transferring data outside a defined local trust boundary, an adaptive data protection mechanism applies a safeguard. Depending on policy restrictions and the sensitivity of the data, this safeguard may include encryption, transmitting only features, or adding noise to protect privacy.

[0017] The system further includes a fallback mechanism configured to select the next available execution node if a selected node becomes unavailable, thus ensuring reliability. For highly critical tasks, low-latency nodes, such as the edge gateway, are prioritized during the fallback selection process. By dynamically balancing energy consumption, latency, and data privacy risk based on patient condition and network context, the invention improves the battery life of wearables, reduces response times in medical emergencies, enhances data privacy compliance, and enables scalable deployment in hospitals, home healthcare systems, telemedicine platforms, and emergency care infrastructures.

[0018] The present invention provides a bio-inspired, adaptive task offloading system that enables energy-efficient, latency-aware, and privacy-compliant processing in IoT edge cloud networks in healthcare. The invention is described with reference to exemplary embodiments, which serve only for illustration and are not intended to limit the scope of the invention. The system operates within a multi-layered healthcare architecture designed to dynamically assign computational tasks generated by patient monitoring devices. The system comprises an IoMT sensing layer, an intelligent edge gateway layer, an optional fog computing layer, and a cloud analytics layer.The IoMT sensing layer comprises wearable and implantable medical sensors configured to continuously capture physiological parameters such as electrocardiogram (ECG), oxygen saturation (SpO2), body temperature, blood pressure, glucose levels, and physical activity.

[0019] These devices typically have limited resources and operate on limited battery capacity. The intelligent edge gateway is logically and physically located near the patient and can be implemented as a smartphone, bed hub, or embedded controller. The gateway receives sensor data streams, performs preprocessing, makes scheduling decisions, and manages communication with the fog or cloud infrastructure. The cloud analytics layer comprises centralized servers configured for large-scale analytics, long-term storage, and clinician-oriented systems. In certain implementations, a fog node can be deployed within a hospital or regional network to provide intermediate computing capacity with reduced latency compared to remote cloud servers. IoMT devices generate data packets or computational tasks based on sampled physiological signals.

[0020] A task can include signal filtering, feature extraction, anomaly detection, predictive modeling, or alert generation. Upon receipt at the edge gateway, each task is encapsulated with metadata, including a timestamp, device identifier, task type, data size, and contextual information. The gateway maintains a task queue and initiates an adaptive evaluation process for each incoming task. The system includes a criticality index (CI) module configured to quantify the clinical urgency associated with a given task. The CI can be calculated using a threshold-based assessment of vital signs relative to predefined safety margins, trend analysis to detect sudden or progressive deviations from baseline, or anomaly values ​​generated by lightweight machine learning models running at the gateway.For example, the detection of arrhythmic ECG patterns or sudden oxygen desaturation leads to a high CI value, while routine monitoring data with stable trends leads to a low CI value.

[0021] The calculated CI directly influences planning priorities. For each potential execution node—including execution on wearables, edge gateways, fog nodes, and cloud servers—the system estimates the expected cost parameters. These include the anticipated energy consumption based on the required compute cycles and communication overhead; the anticipated end-to-end latency, including transmission delay, network propagation delay, queue delay, and processing time; and a data privacy risk assessment, which depends on whether sensitive data leaves a defined local trust boundary and the type of protection applied.

[0022] These predicted values ​​are dynamically updated based on runtime statistics and current network conditions. A multi-criteria cost function is defined for each potential execution decision. This function combines predicted energy consumption, predicted latency, and data privacy risk using weighted coefficients. The values ​​of these coefficients are dynamically assigned as a function of the criticality index. If the AI ​​indicates high clinical urgency, the weight of latency is increased to prioritize a faster response. If the AI ​​indicates lower urgency, energy efficiency and data privacy are given greater weight, conserving the wearable's battery life and minimizing the unnecessary disclosure of sensitive data.

[0023] This adaptive weighting mechanism enables context-aware trade-offs between response speed, energy consumption, and data privacy. The scheduler uses a bio-inspired metaheuristic optimization approach to scour the decision space of possible execution paths. In one embodiment, an ant colony optimization technique is employed, where each ant represents a possible path for task execution, such as "wearable-to-edge," "wearable-to-fog," or "wearable-to-cloud." The paths are evaluated based on their cost function, and lower-cost paths leave higher pheromone levels, thus reinforcing favorable decisions across successive iterations.

[0024] In alternative embodiments, particle swarm optimization can be employed, where each particle represents a possible relocation configuration and updates its position in the search space based on individual and collective best solutions. In another embodiment, a whale optimization algorithm can be used to iteratively refine proposed solutions. The scheduler operates within a limited decision window to ensure real-time responsiveness and outputs the selected execution node for each task. Before payload data is transferred beyond the defined local trust boundary, the system activates an adaptive data protection module. Depending on the sensitivity classification and policy rules, the data protection module applies one or more protection mechanisms.In one embodiment, raw sensor streams are converted into feature vectors at the edge gateway, and only derived features are transmitted to fog or cloud layers. In another embodiment, payload data is encrypted before transmission. In yet another embodiment, privacy-protecting noise is added to the aggregated statistics, where permitted. Privacy protection can enforce predefined restrictions, such as prohibiting the transmission of raw ECG signals beyond the gateway, while allowing the storage of protected feature-level data in the cloud.

Claims

[1] A bio-inspired, adaptive task outsourcing system for an energy-efficient IoT edge cloud continuum in healthcare, comprising: an IoMT sensing layer comprising one or more wearable or implantable medical sensors configured to collect physiological patient data and generate one or more processing tasks; an intelligent edge gateway that is communicatively coupled to the IoMT perception layer and configured to receive one or more processing tasks; and a cloud analytics layer that is communicatively coupled with the intelligent edge gateway, including the intelligent edge gateway: a criticality index module (CI module) configured to calculate a criticality index for each processing task, representing a clinical urgency level associated with the patient's physiological data; a cost estimation module configured to estimate, for each of a variety of eligible execution nodes, including at least one of: execution on a portable device, execution on an edge gateway, execution on a fog node, and execution on a cloud server, a predicted power consumption, a predicted end-to-end latency, and a predicted privacy risk; a bio-inspired metaheuristic scheduler configured to select an execution node from the multitude of eligible execution nodes based on a multi-criteria cost function that combines the predicted energy costs, the predicted end-to-end latency, and the privacy risk using weighting parameters that are dynamically adjusted depending on the calculated criticality index; and a privacy guardian configured to apply a privacy mechanism to the processing task when the selected execution node is outside a defined local trust boundary, where the selected execution node performs the processing task and generates an output or a warning message for transmission to a clinician interface, so that the system dynamically balances energy consumption, response latency, and data privacy based on the patient's clinical urgency and network conditions. [2] System according to claim 1, wherein the IoMT sensing layer comprises one or more wearable sensors selected from electrocardiogram (ECG) sensors, oxygen saturation (SpO2) sensors, temperature sensors, blood pressure sensors, blood glucose meters and activity trackers. [3] System according to claim 1, wherein the criticality index is calculated on the basis of at least one of the following components: predefined threshold rules for vital parameters, trend deviation analysis or an anomaly value generated by a lean machine learning model running on the intelligent edge gateway. [4] System according to claim 1, wherein the multi-criteria cost function comprises weighted components corresponding to the predicted energy costs, the predicted end-to-end latency and the data privacy risk, and wherein the weight assigned to latency is increased when the criticality index exceeds a predefined urgency threshold. [5] System according to claim 4, wherein, if the criticality index is below a predefined urgency threshold, the weights associated with energy costs and data privacy risk are increased relative to the latency weighting. [6] System according to claim 1, wherein the bio-inspired metaheuristic scheduler uses an ant colony optimization algorithm in which possible execution paths are evaluated based on pheromone updates that are proportional to the inverse of the value of the multi-criteria cost function. [7] System according to claim 1, wherein the bio-inspired metaheuristic scheduler uses a swarm-based optimization technique selected from particle swarm optimization (PSO) or whale optimization algorithm (WOA) to determine the execution node. [8] System according to claim 1, wherein the data protection mechanism applies at least one data protection mechanism selected from the encryption of user data, the transmission of feature vectors instead of raw physiological data or the addition of privacy-preserving noise to aggregated statistics. [9] System according to claim 1, further comprising a fallback mechanism configured to select the next feasible execution node if a previously selected execution node is no longer available.