@

초록

Medical imaging analysis in resource-limited settings face critical challenges, including inadequate computational infrastructure, unreliable connectivity, and a shortage of medical expertise. While cloud-based solutions have dominated medical artificial intelligence (AI), their reliance on stable internet and centralized data processing raises concerns about latency, privacy, and accessibility. On-device AI processes data locally on edge device and emerge as a transformative approach to address the given limitations. This review systematically explores the technical foundations, applications, and challenges of lightweight on-device AI for medical imaging in low-resource environments. Case studies across infectious disease detection using portable ultrasound, and retinal diagnostics demonstrate the feasibility of deploying compact models on devices such as Raspberry Pi-based platforms and smartphone. However, significant bottlenecks persist including hardware constraints, data scarcity, and ethical risks such as algorithmic bias. This study purpose the roadmap for future research to advocate for edge-cloud collaboration, bio-inspired ultra-low-power AI, and to bridge clinical and engineering domains. This review synthesizes cutting-edge advancements and provides practical insights for achieving equitable healthcare through decentralized and privacy-preserving AI solutions.

키워드