bowang-lab/dsh-medomni

README 摘要

DeepSeek Harness × MedOmni: A Composable Agentic Framework for Biomedical Image Analysis Figure preparation assisted by ChatGPT. =22" / Demo https://github.com/user-attachments/assets/8a098d95-8d24-44bd-8435-67e52450c524 Contents - What this plugin does - Demo - Requirements - Quick start - Enable image input for pasted images - Usage examples - Tools - Adding a new tool - Disable / re-enable - Configure - Troubleshooting - Scope - License What this plugin does dsh-medomni is a preview implementation of the MedOmni strategy for medical image analysis. It gives a language-model agent a modality-specific set of tools so it can select an appropriate workflow from a natural-language request: - Report generation: Generate candidate reports for chest X-ray, CT, MRI, and retinal images; compare current and prior chest X-rays for interval change. - Localization: Ground suspected chest X-ray findings with evidence boxes, locate named chest X-ray findings, and label normal chest X-ray anatomy. - Segmentation: Generate masks and overlays for prompted findings or structures in X-ray, ultrasound, retinal images, CT NIfTI volumes, and MRI NIfTI volumes; perform fixed-label anatomical segmentatio…

分类