Related Experiment Video
Updated: Jan 17, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.3K
Automated Field of View Prescription for Whole-body Magnetic Resonance Imaging Using Deep Learning Based Body Region
Investigative Radiology
|September 16, 2025
Summary
Automated deep learning models accurately prescribe field-of-view (FoV) stations for whole-body MRI (WB-MRI), matching expert performance. This AI-driven approach enhances efficiency and standardization in WB-MRI acquisition.
Area of Science:
- Radiology and Medical Imaging
- Artificial Intelligence in Healthcare
- Deep Learning Applications
Background:
- Manual field-of-view (FoV) prescription in whole-body magnetic resonance imaging (WB-MRI) is critical for image quality but is time-consuming and prone to variability.
- Current manual methods impact patient comfort and workflow efficiency, necessitating innovative solutions.
Purpose of the Study:
- To develop and evaluate an automated system for multi-station FoV prescription in WB-MRI using deep learning (DL)-based 3D anatomic segmentations.
- To assess the accuracy, reliability, and clinical utility of the automated FoV prescription system.
Main Methods:
- A deep learning model (nnUNet-v2) was fine-tuned to segment anatomical structures on fast whole-body localizer (FWBL) images from 374 patients.
- Five consecutive FoVs (head/neck, thorax, liver, pelvis, spine) were automatically generated based on the segmentations.
- Segmentation accuracy was measured using Sørensen-Dice coefficients (DSC), Precision, Recall, and Specificity. Clinical utility was evaluated by expert radiologists and radiographers.
Main Results:
- The DL system achieved high mean DSCs for various anatomical regions, including torso (0.98) and head/neck (0.96).
- Clinical utility assessments showed high acceptability rates (98.3% internal, 87.5% external datasets).
- Automated FoVs were ranked highest in 60% of cases, outperforming less experienced radiographers and matching expert performance.
Conclusions:
- Deep learning-based 3D anatomic segmentations provide accurate and reliable multi-station FoV prescription for WB-MRI.
- The automated system achieves expert-level performance, significantly reducing manual workload and interoperator variability.
- Automated FoV planning promises to standardize WB-MRI acquisition and improve workflow efficiency, potentially increasing clinical adoption.
