Multi-Agent AI Extracts Oncology Data with High Accuracy

Clinically relevant oncology information is often scattered across various heterogeneous and longitudinal documents, making manual abstraction a time-consuming and burdensome task. This manual process can take nearly half an hour per case, highlighting a critical need for scalable methods that can accurately extract structured data while preserving clinical context. Researchers evaluated the Nimblemind Multi-Agent System (nMAS), a configurable workflow designed for oncology information extraction. nMAS is built to extract 328 specific attributes, defined by clinicians, covering report metadata, diagnosis, staging, and cancer-type-specific details from fragmented oncology documentation. The system's design separates clinician-defined field specifications from the model's execution, combining complexity-aware extraction, report-level consolidation, and source-grounded validation. In a retrospective evaluation involving 230 de-identified oncology documents from 40 patients, nMAS achieved a rank-weighted value-level precision of 82.6%, recall of 87.5%, and an F1 score of 85.0%. This performance significantly surpassed an independently implemented comparator model, which achieved an F1 of 66.4%. These findings strongly support the viability of using a configurable, source-grounded extraction workflow to transform complex, fragmented oncology documentation into reusable structured data, which can then be used for analytics or presented in tumor boards.