Optimizing Medical Image Classification Models for Edge Devices

Machine learning algorithms for medical diagnostics often require resource-intensive environments to run, such as expensive cloud servers or high-end GPUs, making these models impractical use in the field. We investigate of model quantization and GPU-acceleration chest X-ray classification on edge devices. employ 3 types (dynamic range, float-16, full int8) which we tested trained Chest-XRay14 Dataset. achieved a 2–4x reduction size, offset by small decreases mean AUC-ROC score 0.0%–0.9%. On ARM architectures, integer was shown improve inference latency up 57%. However, also observe significant increases x86 processors. GPU acceleration improved latency, but this outweighed kernel launch overhead. show that optimization diagnostic has potential expand their utility day-to-day devices used patients healthcare workers; however, improvements are context- architecture-dependent should be relevant before deployment low-resource environments.