Artificial intelligence-assisted detection of soft tissue calcifications and ossifications in CBCT

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Soft tissue calcifications and ossifications (STCO) are common findings on cone beam computed tomography (CBCT).
• Accurate detection and classification of STCOs are crucial for diagnosis and treatment planning.
• Integrating artificial intelligence (AI) can potentially enhance the interpretation of complex imaging data.

Purpose of the Study:

• To develop and evaluate an AI system for detecting various types of STCOs on CBCT images.
• To assess the diagnostic accuracy of the AI system in both single-class and multi-class classification scenarios.
• To determine the potential of AI-assisted CBCT interpretation in clinical practice.

Main Methods:

• Retrospective review of CBCT images from 287 patients.
• Segmentation of STCOs in the axial plane and identification across all planes.
• Training an AI model to detect 10 specific STCO types and a combined class, with data split for training, testing, and validation.

Main Results:

• The single-class AI model achieved high performance with sensitivity, precision, and F1-score of 0.98, 0.91, and 0.94, respectively.
• The multi-class AI model demonstrated strong results with sensitivity, precision, and F1-score of 0.88, 0.80, and 0.84, respectively.
• The AI system exhibited superior accuracy in single-class classification tasks.

Conclusions:

• The developed AI system shows high diagnostic accuracy for detecting STCOs on CBCT.
• Single-class classification by the AI yielded superior performance compared to multi-class classification.
• AI-assisted CBCT analysis holds promise for improving diagnostic efficiency, interdisciplinary collaboration, and clinical decision-making.