GreenQMedNet: Reducing Computational Footprint in Medical Data Analysis Through Explainable Quantum Hybridization

Deep learning has revolutionized medical data classification, significantly improving diagnostic accuracy. However, conventional deep learning models often exhibit high computational intensity, energy inefficiency, and limited scalability, posing considerable challenges for sustainable multimedia computing. Furthermore, elevated false-positive rates in these models place an additional burden on already strained clinical resources. To address these issues, we propose a novel energy-efficient hybrid quantum-classical learning framework tailored for high-precision medical data analysis. Our architecture synergistically integrates a compact convolutional encoder, responsible for classical representation learning, with a lightweight quantum neural network dedicated to classification. This jointly optimized system achieves a substantial reduction in model size, using approximately 98.06% fewer parameters and 16.7% fewer FLOPs compared to lightweight baselines such as SqueezeNet. At the same time, it enhances classification precision, most notably by substantially reducing false positives. The efficacy of our approach is rigorously evaluated across five distinct medical datasets. The results demonstrate that the hybrid architecture achieves competitive classification performance while significantly reducing model parameters and computational footprint. These findings highlight the potential of hybrid quantum-classical architectures as an early-stage approach toward energy-efficient medical AI systems.