Hybrid Task Scheduling for Optimized Neural Network Inference on Skin Lesions in Resource-Constrained Systems

This study proposes a hybrid task scheduling framework to optimize neural network inference for skin lesion classification in resource-constrained environments. The framework integrates Earliest Deadline First scheduling with Dynamic Voltage and Frequency Scaling to balance real-time performance and energy efficiency. By dynamically adjusting task execution priorities and processor frequency, the proposed method reduces missed deadlines, optimizes resource utilization, and minimizes power consumption. The approach effectively adapts to varying workloads, ensuring that AI-driven skin lesions medical imaging tasks meet real-time constraints without excessive computational overhead. This hybrid scheduling method can be extended to other healthcare applications, including realtime anomaly detection in medical environments.