Related Experiment Video
Updated: Jun 7, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.7K
Deep learning segmentation-based bone removal from computed tomography of the brain improves subdural hematoma
Masis Isikbay1, M Travis Caton2, Jared Narvid3
1Department of Radiology and Biomedical Imaging, University of California San Francisco, 505 Parnassus Ave, M-396, San Francisco, CA 94143, USA.
Journal of Neuroradiology = Journal De Neuroradiologie
|November 9, 2024
Summary
A novel deep learning algorithm accurately removes bone from non-contrast CT head scans, significantly improving subdural hematoma detection for trainee radiologists.
Area of Science:
- Radiology
- Medical Imaging
- Artificial Intelligence
Background:
- Subdural hematoma (SDH) detection on non-contrast CT head (NCCTH) is crucial but challenging due to bone interference.
- Accurate and timely identification of intracranial blood is clinically significant.
Purpose of the Study:
- To evaluate the utility of a NCCTH bone removal algorithm for enhancing SDH detection.
- To assess the impact of bone removal on the diagnostic performance of junior radiology trainees.
Main Methods:
- A deep learning segmentation algorithm was developed and trained for bone removal on 100 NCCTH.
- Segmentation accuracy was validated using quantitative overlap analysis on internal and external datasets.
- A reader study compared SDH detection rates between standard NCCTH and NCCTH with bone removal applied.
Main Results:
- The bone removal algorithm achieved high segmentation accuracy (Dice overlap 0.9999 internal, 0.9957 external).
- SDH detection was statistically improved (P < 0.001) for trainees using NCCTH with bone removal.
- A high percentage of trainees reported improved detection and desire for on-call access to the tool.
Conclusions:
- Deep learning-based NCCTH bone removal is a rapid and accurate method.
- Bone removal enhances SDH detection accuracy and confidence among trainee radiologists.
- This technology shows potential for improving clinical diagnosis of intracranial hemorrhages.
