Automatic Detection of Periapical Osteolytic Lesions on Cone-beam Computed Tomography Using Deep Convolutional...

Area of Science:

• Oral Radiology
• Artificial Intelligence in Dentistry
• Medical Imaging Analysis

Background:

• Cone-beam computed tomography (CBCT) is crucial for oral radiology.
• Radiolucent periapical lesions (PALs) are common jaw pathologies.
• Manual detection and interpretation of PALs are time-consuming and require expertise.

Purpose of the Study:

• To develop and validate a deep convolutional neural network (CNN) for automated detection of osteolytic PALs in CBCT data.
• To improve the quality and efficiency of diagnosing PALs using AI.

Main Methods:

• A two-step deep learning approach was employed for automated PAL detection.
• Tooth localization and identification utilized SpatialConfiguration-Net.
• Lesion segmentation was performed using a modified U-Net architecture.
• The method was trained and tested on 144 CBCT datasets using 4-fold cross-validation.

Main Results:

• The tooth localization network achieved detection rates between 72.6% and 97.3%.
• The lesion detection sensitivity was 97.1% and specificity was 88.0%.
• The automated method demonstrated high performance despite variations in PAL appearance and data imbalance.

Conclusions:

• The proposed fully automated AI method for osteolytic PAL detection in CBCT data yields excellent results.
• This AI tool has the potential to enhance diagnostic accuracy and efficiency in oral radiology.
• The study highlights the utility of deep learning in addressing challenges in radiological interpretation.