Towards Building AI Doctor and Digital Assistant by Capitalizing on A Mixture of AI Medical Systems

Global healthcare faces mounting challenges, including rising demand, limited specialist availability, and unequal access to diagnostic resources. To address these issues, this work proposes a unified framework for constructing an AI doctor and digital assistant by capitalizing on a mixture of interoperable medical AI systems. Our approach integrates diverse modalities such as medical images, X-rays, CT scans, MRIs, PET scans, videos, and laboratory data to support end-to-end healthcare workflows. We demonstrate foundational building blocks through two exemplar systems: (1) automatic rosacea detection, enhanced with interpretable and privacy-preserving techniques, and (2) laparoscopic image desmoking, using novel loss functions and differentiable Wiener filtering for real-time surgical support. These contributions highlight two pillars of the AI doctor vision: accurate and trustworthy disease detection, and enhancement of clinical imaging workflows. Looking forward, we extend the framework toward additional disease-specific systems, including breast cancer, tumors, Parkinson’s, and Alzheimer’s detection, thereby advancing multimodal and modular healthcare solutions. Ultimately, this work lays the foundation for scalable, reliable, and accessible AI-driven healthcare, enabling equitable clinical support for patients and providers worldwide.