Convolutional Neural Network Performance for Sella Turcica Segmentation and Classification Using CBCT Images

Introduction

The study focuses on the application of advanced technology, specifically Convolutional Neural Networks (CNNs), in the segmentation and classification of Sella Turcica using Cone Beam Computed Tomography (CBCT) images. The aim is to improve the accuracy and efficiency of radiographic interpretation in dentistry.

Methodology

• Image Acquisition: CBCT images were obtained using the NewTom 5G CBCT device with specific parameters.
• Ground Truth Labeling: Two dentomaxillofacial radiologists labeled the Sella Turcica shapes in the CBCT images.
• Models: A Sella Turcica Segmentation Model was developed using deep learning techniques, specifically a U-Net architecture.
• Pre-Processing Steps: Image enhancement techniques such as intensity normalization and contrast limited adaptive histogram equalization were applied.
• Model Training: The deep CNN segmentation model was trained with 500 epochs and a specific learning rate.

Results

• The study successfully segmented and classified Sella Turcica shapes (flattened, round, oval) using the developed AI models.
• The use of CNNs and deep learning techniques improved the accuracy of segmentation and classification compared to traditional methods.

Discussion

• The integration of AI algorithms in radiographic interpretation offers advantages in terms of accuracy, efficiency, and prevention of misdiagnoses.
• The study contributes to the growing body of research on AI applications in dentistry and medical imaging.

Conclusion

The research demonstrates the potential of CNNs and deep learning models in automating the segmentation and classification of anatomical structures like Sella Turcica in CBCT images. This technology has the potential to enhance diagnostic accuracy and streamline radiographic interpretation in dentistry.

Funding and Ethical Considerations

• The study was supported by the Eskisehir Osmangazi University Scientific Research Projects Coordination Unit.
• Ethical standards were followed, and informed consent was obtained from all participants.
• Data availability and conflicts of interest were addressed in accordance with established guidelines.