DICOMScribe

DICOMScribe is a web + desktop AI app that helps radiologists draft consistent, structured reports faster by leveraging the current study plus the patient’s prior imaging and reports. Instead of trying to “diagnose” from pixels alone (a regulatory and liability minefield), it focuses on high-confidence workflow automation: pulling priors, extracting key measurements, surfacing interval changes, and generating a first-pass narrative and impression in the site’s preferred template. It integrates with PACS/RIS via DICOM and HL7/FHIR, and can run in a “no-image-upload” mode where inference happens on-prem or in a hospital-controlled VPC. The goal is to reduce time spent on comparison language, stable findings, and boilerplate while keeping the radiologist fully in control with clear provenance (what came from priors vs. what was newly observed). Expect meaningful time savings, but not magic—edge cases and poor priors will still require manual work.