Related Experiment Video
Updated: Aug 8, 2026
06:45
Automated Joint Space Detection Improves Bone Segmentation Accuracy
Published on: November 28, 2025
Explainable Deep Learning for Automated Classification of Skeletal and Tumor Pathologies in 3D Volumetric Medical
Ahmad Almadhor1, Stephen Ojo2, Thomas I Nathaniel3
1Department of Computer Engineering and Networks, College of Computer and Information Sciences, Jouf University, Sakaka, 72388, Saudi Arabia.
Journal of Imaging Informatics in Medicine
|August 7, 2026
Summary
Deep learning models effectively classify 3D medical mesh data, with VGG-16 achieving 99.15% accuracy. This framework enhances tumorous and skeletal structure identification for improved medical applications.
Area of Science:
- Medical Imaging Analysis
- Artificial Intelligence in Healthcare
- 3D Data Processing
Background:
- Accurate classification of 3D medical mesh data is vital for clinical applications like surgical planning and disease diagnosis.
- Deep learning (DL) is increasingly applied to medical imaging, especially for analyzing complex 3D mesh structures.
Purpose of the Study:
- To develop and evaluate an explainable AI-based deep learning framework for classifying tumorous and skeletal structures in 3D medical data.
- To assess the performance of various DL models, including CNN, ResNet-18, VGG-16, and EfficientNet-B0, on a large-scale medical dataset.
Main Methods:
- A robust preprocessing pipeline was created to normalize Stereolithography (STL) neuro-anatomical structures into feature vectors.
- Multiple DL models were trained and evaluated on the MedShapeNet dataset, comprising over 100,000 3D medical shapes.
- Model performance was assessed using quantitative metrics and advanced visualization techniques for explainability.
Main Results:
- The VGG-16 model demonstrated the highest test accuracy (99.15%) and F1 score (99.16%).
- ResNet-18 achieved a high accuracy of 98.87%, closely following VGG-16.
- The framework successfully classified tumorous and skeletal structures with high precision.
Conclusions:
- Deep learning, particularly VGG-16, significantly improves the performance of medical mesh classification tasks.
- The explainable AI framework provides valuable insights into model decision-making processes.
- This approach holds promise for enhancing diagnostic accuracy and clinical decision support in medical imaging.