Related Experiment Video
Updated: Jul 31, 2025
13:35
Segmentation and Linear Measurement for Body Composition Analysis using Slice-O-Matic and Horos
Published on: March 21, 2021
10.6K
A vendor-agnostic, PACS integrated, and DICOMcompatible software-server pipeline for testing segmentation algorithms
Lei Zhang1, Wayne LaBelle1, Mathias Unberath2
1University of Maryland, Baltimore.
Research Square
|May 10, 2023
Summary
This study introduces a deployable pipeline for real-time AI/ML medical image segmentation testing. The system integrates PACS with on-premises processing, enabling visualization of segmentation results and volumetric data at the workstation.
Area of Science:
- Medical Imaging Analysis
- Artificial Intelligence in Medicine
- Radiology Workflow Integration
Background:
- Need for reproducible AI/ML in medical imaging at the clinical bedside.
- Enabling real-time shadow testing of segmentation algorithms on new studies.
- Integrating Picture Archiving and Communication System (PACS) with on-premises image processing.
Approach:
- Developed a containerized, deployable pipeline for clinical workflow shadow testing.
- Pipeline components: router/listener, anonymizer, OHIF web viewer, DCM4CHEE DICOM archive.
- On-premises workstation for DICOM/NIfTI conversion and image processing.
Key Points:
- Visualizes DICOM images with segmentation masks and volumetry (mL) via DICOM SEG and SR.
- Demonstrated feasibility with a traumatic pelvic hematoma segmentation model.
- Mean total clock time from PACS send to archive completion: 5 minutes 32 seconds.
Conclusions:
- Software seamlessly integrates with existing PACS for prospective DL model testing.
- Pipeline executed via a single command shell script.
- Open-source code available with configuration instructions.
