Project Overview
A radiology network needed to reduce reporting time while improving consistency of findings across modalities. We designed an AI pipeline for detection and triage with seamless PACS/RIS integration and clinician‑friendly overlays.
Challenges
- DICOM variability and strict privacy requirements
- Class imbalance and calibration for rare pathologies
- Real‑time GPU inference and viewer integration
Neuradigi Solution
We built a secure DICOM pipeline, trained UNet/ResNet‑based models with MONAI, and deployed GPU‑backed inference services with monitoring and human‑in‑the‑loop review.
1
Data & Pipeline: DICOM ingestion, anonymization, QA, and augmentation.
2
Modeling: UNet/ResNet ensemble, cross‑validation, calibration, drift checks.
3
Deployment: TorchScript/Triton serving, PACS/RIS integration, dashboards.
Business Outcomes
- 96% detection accuracy (AUC‑aligned)
- 70% faster case triage
- 3x cases processed per day
- 30% fewer false positives
