Related Experiment Video
Updated: Oct 7, 2025
06:59
A Pipeline using Bilateral In Utero Electroporation to Interrogate Genetic Influences on Rodent Behavior
Published on: May 21, 2020
4.2K
Federated Deep Learning to More Reliably Detect Body Part for Hanging Protocols, Relevant Priors, and Workflow
Ross W Filice1, Anouk Stein2, Ian Pan3
1Department of Radiology, MedStar Georgetown University Hospital, 3800 Reservoir Road, NW CG201, Washington DC, 20007, USA. ross.w.filice@gunet.georgetown.edu.
Journal of Digital Imaging
|January 12, 2022
Summary
Artificial intelligence (AI) accurately identifies body parts in radiology exams, improving prior exam retrieval and display protocols. This enhances workflow efficiency and diagnostic accuracy by overcoming limitations in current data standards.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence
Background:
- Radiology exam preparation relies on accurate body part information for prefetching prior studies and optimizing hanging protocols.
- Current Digital Imaging and Communications in Medicine (DICOM) standards for body part information are often variable, inaccurate, lack granularity, or are missing.
- These data deficiencies lead to suboptimal prefetching and hanging protocols, impacting radiologist workflow and potentially diagnostic quality.
Purpose of the Study:
- To evaluate the efficacy of modern artificial intelligence (AI) techniques, specifically federated deep learning, for automatic body part detection in radiological examinations.
- To demonstrate how AI-driven body part categorization can enhance the reliability of prefetching and hanging protocols.
- To explore the potential of AI in optimizing radiological image display and dynamic hanging protocols.
Main Methods:
- Utilized federated deep learning techniques to analyze image data within radiological examinations for automatic body part identification.
- Developed and implemented AI models trained on diverse radiological image datasets.
- Integrated AI-based body part detection into the radiology workflow for prefetching and hanging protocol management.
Main Results:
- AI techniques, particularly federated deep learning, achieved highly accurate automatic detection of body parts from radiological image data.
- The AI-driven approach significantly improved the reliability of body part categorization compared to traditional DICOM encoding.
- Demonstrated potential for enhanced prefetching of relevant prior examinations and more effective hanging protocol implementation.
Conclusions:
- AI-powered body part detection offers a robust solution to the limitations of current DICOM standards.
- This technology enables more efficient and accurate radiology workflows through improved exam prefetching and display.
- Future applications include dynamic hanging protocols and advanced image display optimization using AI.
