MedScribe AI
An open-source AI medical agent system that chains computer vision and generative AI into a clinical report generation pipeline, producing a complete radiology report within seconds from image upload.
Timeline (last 90 days)
MedScribe AI is an open-source AI agent project built by a six-person Pakistani team in a 48-hour hackathon (Pak Angels Generative & Agentic AI Hackathon, Cohort 11), used to generate complete clinical reports from chest X-ray images.
MedScribe AI adopts a three-stage pipeline architecture: a vision model identifies lesions, four agents collaborate to generate the report, and a bilingual interactive interface.
MedScribe AI uses a ResNet18 model trained on the Kaggle pediatric chest X-ray dataset at its foundation, and with the help of the TorchXRayVision open-source medical imaging foundation model, achieves 90.5% accuracy, 98.4% precision, and 93.2% F1 score according to data published by the project team.
MedScribe AI implements serial collaboration of four AI agents (Vision, Reasoning, Reporting, Escalation) based on the LangGraph framework.
The Reasoning agent in MedScribe AI uses a ReAct loop and can dynamically decide subsequent tool calls based on visual findings, rather than following fixed rules.
The reasoning layer of MedScribe AI uses openai/gpt-oss-120b running through Groq, achieving fast and free inference, and patient X-ray data is not sent to any proprietary server.
MedScribe AI uses Streamlit to build the application layer, allowing users to upload chest X-rays and return within seconds two versions comprising a professional clinical report for doctors and a plain-language summary for patients.
MedScribe AI supports bilingual output in English and Urdu, and reports can be exported as PDF, including the original X-ray and a Grad-CAM heatmap.
MedScribe AI's pricing logic is that clinics pay 50 USD per month for AI preliminary triage, far below the cost of hiring a part-time radiologist.
The pediatric chest X-ray dataset used by MedScribe AI is limited in scale, the model's generalization ability has not yet been validated on adult imaging or multi-center data, and the accuracy and compliance of the clinical reports have not been systematically evaluated in a real medical environment.
1 more timeline events
All Facts (10)
MedScribe AI's target paid scenarios include: private clinics in Pakistan and South Asia without an on-site radiologist, telemedicine platforms that need automated preliminary reports, and hospital chains handling large volumes of routine chest X-rays.
50%UnverifiedMedScribe AI adopts a three-stage pipeline architecture: a vision model identifies lesions, four agents collaborate to generate the report, and a bilingual interactive interface.
50%UnverifiedMedScribe AI uses a ResNet18 model trained on the Kaggle pediatric chest X-ray dataset at its foundation, and with the help of the TorchXRayVision open-source medical imaging foundation model, achieves 90.5% accuracy, 98.4% precision, and 93.2% F1 score according to data published by the project team.
50%UnverifiedMedScribe AI implements serial collaboration of four AI agents (Vision, Reasoning, Reporting, Escalation) based on the LangGraph framework.
50%UnverifiedThe Reasoning agent in MedScribe AI uses a ReAct loop and can dynamically decide subsequent tool calls based on visual findings, rather than following fixed rules.
50%UnverifiedThe reasoning layer of MedScribe AI uses openai/gpt-oss-120b running through Groq, achieving fast and free inference, and patient X-ray data is not sent to any proprietary server.
50%UnverifiedMedScribe AI uses Streamlit to build the application layer, allowing users to upload chest X-rays and return within seconds two versions comprising a professional clinical report for doctors and a plain-language summary for patients.
50%UnverifiedMedScribe AI supports bilingual output in English and Urdu, and reports can be exported as PDF, including the original X-ray and a Grad-CAM heatmap.
50%UnverifiedThe pediatric chest X-ray dataset used by MedScribe AI is limited in scale, the model's generalization ability has not yet been validated on adult imaging or multi-center data, and the accuracy and compliance of the clinical reports have not been systematically evaluated in a real medical environment.
50%UnverifiedMedScribe AI is an open-source AI agent project built by a six-person Pakistani team in a 48-hour hackathon (Pak Angels Generative & Agentic AI Hackathon, Cohort 11), used to generate complete clinical reports from chest X-ray images.
50%