Cloud-Edge AI System Boosts Rural Clinical Screening.

Delivering advanced medical AI capabilities to rural areas with limited resources presents significant challenges, including scarce bandwidth and compute power, alongside the need to integrate diverse medical data. This research proposes a novel cloud-edge collaborative architecture designed to overcome these constraints for multimodal clinical screening. The system employs lightweight, domain-specific AI models deployed at the edge (e.g., local clinics) to process raw medical data and transform it into compact, structured outputs. These outputs are then sent to a cloud-based Large Language Model (LLM) which synthesizes them into comprehensive clinical summaries. An LLM-based orchestrator dynamically selects the most relevant diagnostic tools based on patient context, ensuring comprehensive coverage without processing unnecessary inputs. Evaluated on 20 multimodal clinical cases across various scenarios (cardiac, obstetric, trauma) and simulated network profiles, the hybrid system demonstrated 98-99% diagnostic tool recall and 92-96% precision. It matched or exceeded cloud-only baselines in clinical accuracy, while maintaining consistent latency (25-35 seconds) regardless of bandwidth and achieving 4-15x lower token costs. This architecture significantly improves factual grounding compared to purely cloud-based approaches under deployment constraints.