Radiology Copilot

Research applicability

Supports imaging researchers studying segmentation failure modes and medical educators building reviewed MRI exercises.

Getting started

Untested editorial guidance: read Installation and Limitations, prepare a local environment and public research cases, then compare masks and report drafts with reference data before considering any research use.

Maintenance

README retrieved on 2026-10-17; future maintenance not verified.

License

Consult repository and each dependency/data licence; no independent legal audit.

Workflow & human oversight

Workflow level left unspecified for this resource-level listing. Researchers choose scope, inspect inputs and methods, and verify generated outputs before reuse.

Workflow level is not an evidence score; higher is not better evidence.

Approach

The README describes MONAI SegResNet segmentation and deterministic measurements, with an LLM selecting registered tools and producing reports pending user review. Research feedback compares predictions, reference masks and user corrections and supports CSV export; it does not automatically retrain models.

Limitations

README-only review; no code or clinical validation performed. The authors explicitly prohibit clinical decision use. Their small-sample Dice and model-comparison claims are not independently reproduced and are not adopted as effectiveness evidence here. User corrections are not necessarily clinician-reviewed; model generalization and privacy require separate assessment.

Reproducibility

Primary README source check only; no installation, execution, security test or performance reproduction.

Adoption requirements

Python environment, project dependencies, separately downloaded MONAI weights and appropriately licensed MRI data. LLM tool calling additionally requires a supported provider or local model; without it the documented fallback is deterministic. Confirm rights and consent before any third-party data transfer.