Project description
Motivation
The rapid evolution and complexity of oncology guidelines, coupled with their abstract nature and limited accessibility, create significant knowledge gaps for both patients and physicians. While patients frequently turn to online resources, the reliability of these sources is often questionable, highlighting the need for accurate and personalized information. The increasing number of cancer cases, coupled with the limited number of specialists, underscores the urgency for tools that can effectively disseminate and personalize evidence-based medical knowledge.
Solution
The Aidvice AI-driven medical knowledge assistant improves access to complex oncology guidelines by providing personalized, up-to-date, explainable, and understandable information for patients and doctors. It seeks to bridge knowledge gaps, support symptom management, and support continuity of care. The assistant uses retrieval-augmented generation (RAG) powered by large language models (LLMs), where an automated fact-checking mechanism is applied to detect and correct any inaccuracies, ensuring trustworthy and reliable answers to medical questions.
Impact
Within 12 months, Aidvice will be evaluated in a closed testing environment by selected physicians and patients focusing on the accuracy, safety, and practical utility of its responses, supporting evidence, and source attribution.
Partners and Sponsors
Aidvice is done in collaboration with the Department of Radiation oncology and radiotherapy at the TUM University Hospital (Klinikum rechts der Isar).
The project is sponsored by Google.org, the charitable arm of Google, as part of the Google.org Accelerator for Generative AI. It was selected as one of the 20 projects among 3000 applications from 120 countries.
