Technical Developments in Radiology: Enhancing Research Workflows with PACScrawler

Area of Science:

• Medical Informatics
• Radiology Research

Background:

• Medical imaging research is often slowed by data silos.
• Accessing imaging studies and reports from Picture Archiving and Communication Systems (PACS) and Radiology Information Systems (RIS) is challenging for cohort creation.

Purpose of the Study:

• To develop PACScrawler, an open-source tool to streamline cohort assembly.
• To enable efficient search and retrieval of imaging studies and reports across PACS and RIS.

Main Methods:

• Developed PACScrawler with a backend crawler to index DICOM/RIS metadata and reports into an Apache Solr/Lucene index.
• Implemented a web interface for multi-parameter search and an on-demand retrieval module for imaging data.
• Evaluated indexing speed, retrieval efficiency compared to manual methods, and utility for LLM-assisted query generation.

Main Results:

• Indexed over 2.3 million examinations from 316,310 patients between 2012 and 2024.
• Routine indexing of 1079 examinations took 26.5 minutes (1.4 seconds/study).
• PACScrawler retrieved 10 head MRI exams 2x faster than manual PACS export; 17/20 LLM-generated queries were usable.

Conclusions:

• PACScrawler reduces technical barriers for clinical imaging research.
• The tool facilitates retrospective cohort identification, data retrieval, and monitoring of imaging trends.
• PACScrawler is adaptable to other clinical settings with PACS/RIS integration.