MedRLM: Recursive Multimodal Health Intelligence for Long-Context Clinical Reasoning, Sensor-Guided Screening, Evidence-Grounded Decision Support, and Community-to-Tertiary Referral Optimization

MedRLM is introduced as a Recursive Multimodal Health Intelligence framework designed for long-context clinical reasoning and decision support, addressing the limitations of existing medical large language models that rely on single-step prompts. The framework utilizes a Clinical Evidence Graph Memory to integrate diverse patient data, including electronic health records, medical images, and sensor signals, enabling recursive inspection and synthesis of information. This approach enhances clinical decision-making by facilitating deeper reasoning in response to abnormal patterns and supporting clinician review through uncertainty-gated refinement, thereby advancing the capabilities of AI in real-world clinical settings.