Mini-CDSS

The project aims to design a Clinical Decision Support System (CDSS) that assists medical practitioners in diagnosing patients by generating an initial patient report, providing a preliminary diagnosis, and offering best practices along with relevant recommendations. Currently, the project is being rebuilt on the LangGraph framework, transitioning from Crew-AI to enhance flow control and leverage the capabilities of Generative AI. A working prototype is expected by February 19th, integrating both text and image analysis to enable multimodal capabilities.

The Project Contains of 3 Main Pillars

1. Main-Graph: Patient may or may not update their clinical/lab reports as pdf here, but we need at least an extract of initial patient encounter with the doctor to give a detailed patient initial report, prelim report (can be looped based on feedback mechanism or exited using satisfied keyword. Then the best practices report is formed and a final report of comprehensive structure is generated.

2. Rag-Graph: Allows the doctor to query with the patient report, uses a reactive layout to fetch most relevant details before answering any query. Fetches data from the vector-DB thus creating a Vector-DB after uploading files is of utmost importance to use this feature.

3. Vision-Capabilities: Using Llama and Gemini Multi-Modal models, we can upload any image and do a over-all diagnosis of what went wrong. We can also analyze a patient's heartbeat as well as body temperature via our Campanion App.