Automated meningioma detection using skull X ray images with deep learning and machine learning classifiers.
Authors
Affiliations (4)
Affiliations (4)
- College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea.
- Department of Neurosurgery, St. Vincent's Hospital, College of Medicine, The Catholic University of Korea, 93 Jungbu-daero, Paldal-gu, Suwon, 16247, Korea.
- Department of Neurosurgery, Incheon St. Mary's Hospital, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea.
- Department of Neurosurgery, St. Vincent's Hospital, College of Medicine, The Catholic University of Korea, 93 Jungbu-daero, Paldal-gu, Suwon, 16247, Korea. [email protected].
Abstract
This study aimed to develop a novel diagnostic tool for detecting meningioma using skull X-ray images, combining deep learning with traditional machine learning classifiers. The goal was to explore the potential of using a cost-effective and widely accessible imaging modality for automated meningioma detection. Skull X-ray images from 158 meningioma patients (632 images) and 201 control subjects (804 images) without brain tumors or vascular diseases were retrospectively collected from St. Vincent's Hospital, South Korea. Anteroposterior, Towne, and lateral views were included in the analysis. EfficientNetB0 served as a backbone of the deep learning model. It was enhanced with transfer learning and attention mechanisms. Extracted features were integrated into traditional classifiers such as Random Forest and XGBoost to improve classification performance. Model performance was evaluated using metrics such as accuracy, sensitivity, specificity, F1-score, and AUROC. External validation was conducted using data (824 images) from Incheon St. Mary's Hospital, South Korea. In the internal validation cohort, the hybrid EfficientNetB0-Random Forest model achieved the highest accuracy (0.97), with an AUROC of 0.999. External validation results showed that Random Forest achieved the best performance among classifiers, with an accuracy of 0.74 and an AUROC of 0.76. Grad-CAM visualizations highlighted the model's focus on key cranial regions for meningioma detection. This study demonstrated the feasibility of using skull X-rays for automated meningioma detection. By integrating deep learning and traditional machine learning techniques, the proposed approach offers a promising diagnostic tool, especially for resource-limited settings where advanced imaging modalities might not be readily available.