
MD.ai
Radiology annotation and AI reporting on one platform

Tool details
MD.ai is a New York company founded by Harvard, Duke and Columbia trained physicians, doing two things: helping imaging teams build annotated datasets and validate models, and helping radiologists write reports with LLMs. Several RSNA Kaggle imaging datasets were built on it.
Core capabilities
- DICOM-native annotation: supports most modalities with direct upload, cloud buckets or C-STORE, plus multiplanar reconstruction and hanging protocols.
- AI-assisted labeling: deploy your own models for pre-annotation and validate them across sites in a federated way without sharing patient data.
- De-identification: built-in PHI detection with both pixel and metadata de-identification to stop leaks.
- Report generation: automatic template selection, key-findings mapping, generated impressions and proofreading, with auto-comparison against prior reports.
- Guidelines and coding: applies BI-RADS, LI-RADS and TI-RADS automatically and generates billing codes at signing.
- Developer access: a CLI and Python client, with HL7 and DICOM integration into EHR, HIS and RIS.
Where it fits Suits hospitals and research groups building their own imaging datasets and running model validation, and radiology departments that want faster reports. It serves institutions with IT and compliance capacity; individuals cannot self-serve.
Why it stands out Annotation and reporting are two legs of the same platform, so models trained on datasets can flow straight back into the reporting workflow. It hosted annotation for public datasets such as the RSNA pulmonary embolism and intracranial hemorrhage challenges, and the engineering foundation is sturdier than report-only products.
Editor's take Technical credibility is high for this category: the viewer received FDA clearance in 2023 (K223425), and the team published work concluding GPT-4V is not yet fit for real diagnostic use, which is a more restrained stance than most peers. The gap is pricing opacity, and the clearance covers the viewer, not the report AI, so do not let a single sentence blur that line in procurement.
Pricing
Custom pricingCustom (quote)Priced by data volume and concurrency
Priced by institution size and integration
Additional notes
No price list and no self-serve signup; everything goes through sales. Research collaborations can be discussed separately.
FAQ
User reviews
User Reviews
No reviews yet
No one has reviewed this tool yet