From Scanner to Science: Reusing Clinically Acquired Medical Images for Research

Area of Science:

• Medical Imaging
• Data Science
• Clinical Research Informatics

Background:

• Medical imaging research is expanding, increasing demand for diverse and large datasets.
• Utilizing clinically acquired data is challenging due to complex, time-consuming transfer and processing.
• Existing methods struggle with the unique complexities of medical imaging data reuse.

Purpose of the Study:

• To present Locutus, a manifest-driven Extract, Transform, Load (ETL) process.
• To address the challenges of reusing clinically acquired medical imaging data for research.
• To facilitate efficient data downloading, deidentification, and delivery.

Main Methods:

• Locutus employs a five-phase workflow: initialization, data preparation, extraction, transformation (deidentification), and loading.
• The process is modular and manifest-driven, ensuring adaptability to specific data requests.
• It is designed to be robust against errors and uncertainties in clinical data, incorporating human oversight.

Main Results:

• The Locutus workflow has successfully processed 32,962 imaging accessions for research.
• The system demonstrates a scalable approach to handling large volumes of medical imaging data.
• The process is transitioning towards regular monitoring with reduced human intervention.

Conclusions:

• Locutus provides an effective solution for reusing clinically acquired medical imaging data.
• The ETL process enhances data accessibility for medical research.
• Future work will focus on addressing technical challenges and optimizing human roles.