Related Experiment Video
Updated: Jun 26, 2025
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
Improving Automated Hemorrhage Detection at Sparse-View CT via U-Net-based Artifact Reduction.
Johannes Thalhammer1, Manuel Schultheiß1, Tina Dorosti1
1From the Department of Physics, School of Natural Sciences (J.T., M.S., T.D., F.P., D.P., F.S.), Munich Institute of Biomedical Engineering (J.T., M.S., T.D., T.L., F.P., D.P., F.S.), Department of Diagnostic and Interventional Radiology, School of Medicine, Klinikum rechts der Isar (J.T., M.S., T.D., F.P., D.P.), Institute for Advanced Study (J.T., F.P., D.P.), and Computational Imaging and Inverse Problems, Department of Computer Science, School of Computation, Information, and Technology (T.L.), Technical University of Munich, Boltzmannstrasse 11, 85748 Garching, Germany.
Deep learning artifact reduction significantly improved automated hemorrhage detection in sparse-view head CT scans. This method maintained high diagnostic accuracy even with substantially fewer X-ray views.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Sparse-view CT scans reduce radiation dose but introduce artifacts, potentially hindering diagnostic accuracy.
- Automated detection of intracranial hemorrhage is crucial for timely diagnosis and treatment.
Purpose of the Study:
- To evaluate the efficacy of deep learning-based artifact reduction in sparse-view cranial CT.
- To assess the impact of this artifact reduction on automated hemorrhage detection performance.
Main Methods:
- A U-Net model was trained for artifact reduction on simulated sparse-view cranial CT scans (3000 patients).
- EfficientNet-B2 was trained for automated hemorrhage detection on full-view CT data (17,545 patients).
- Performance was compared against unprocessed and total variation (TV) postprocessed images using AUC and DeLong test.
Main Results:
- U-Net artifact reduction improved image quality and automated hemorrhage detection compared to unprocessed and TV-processed images.
- Hemorrhage detection performance remained high (AUC 0.97) with U-Net processing, even when reducing views from 4096 to 256.
- Structural similarity index measure increased, indicating enhanced image quality with U-Net postprocessing.
Conclusions:
- Deep learning-based artifact reduction using U-Net substantially enhances automated hemorrhage detection in sparse-view cranial CT.
- This approach allows for significant reduction in the number of CT views while preserving diagnostic performance.
