Related Experiment Video
Updated: Jul 9, 2026
06:45
Automated Joint Space Detection Improves Bone Segmentation Accuracy
Published on: November 28, 2025
Deep Learning-Assisted Three-Dimensional Segmentation of Vertebrobasilar Artery Calcification in Cone Beam Computed
Kardelen Demirezer1, Salih Taha Alperen Özçelik2, Oğuzhan Altun3
1Department of Oral and Maxillofacial Radiology, INONU University, Malatya, Turkey. kardemirezer.1999@gmail.com.
Journal of Imaging Informatics in Medicine
|July 7, 2026
Summary
A new deep learning model, ImprovedVertebroV5, automates vertebral artery (VBAC) segmentation from CBCT scans, improving accuracy and efficiency for stroke risk assessment.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Vascular Neurology
Background:
- Vertebral artery (VBAC) assessment is vital for stroke risk evaluation.
- Manual segmentation of VBACs is time-consuming and prone to inter-observer variability.
Purpose of the Study:
- To develop and validate a novel deep learning model for automated VBAC segmentation.
- To optimize the model for detecting small vascular structures in CBCT images.
Main Methods:
- Developed ImprovedVertebroV5, a 3D U-Net architecture with focal attention, small-object detection, and pyramid pooling.
- Trained and evaluated the model on a clinical dataset of CBCT scans from patients undergoing vertebrobasilar evaluation.
Main Results:
- The V5 model achieved a mean Dice score of 0.7312 ± 0.2080 and IoU of 0.6100 ± 0.2180.
- Demonstrated high precision (0.7785 ± 0.1347), recall (0.7457 ± 0.2526), and specificity (0.9997 ± 0.0003).
- Showed statistically significant improvements over the V4 benchmark in multiple performance metrics.
Conclusions:
- The proposed small object-optimized architecture shows promising performance for automated VBAC segmentation on CBCT.
- This model may facilitate opportunistic screening for VBAC abnormalities in dental imaging settings.