RECIST 1.1 lesion labeling for a CT reading workflow
AI-assisted lesion segmentation, classification and anatomical labeling on CT following the client's RECIST-based protocol.
A clinical AI provider running a CT lesion segmentation pipeline needed radiologists to annotate baseline oncology studies following a detailed RECIST 1.1 protocol, including per-lesion classification, anatomical site correction and specific handling of lymph nodes and bone lesions.
Segment measurable lesions with the client's AI-assisted tool, classify each lesion (malign / benign / undetermined / treated), correct auto-assigned anatomical sites and annotate non-measurable findings as ROIs.
Dedicated radiologists were trained on the client's classification tables, lymph node and bone lesion sub-protocols, and specific rules for context-dependent classification. Each series was tracked in a shared spreadsheet and reviewed for adherence to the protocol before hand-off.
The client received a consistent, protocol-compliant reading of each study. The rules around lymph node sub-locations and bone lesion morphology were applied uniformly across annotators, which reduced downstream cleanup on the client's side.