Fully Automated Detection of the Appendix Using U-Net Deep Learning Architecture in CT Scans.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Accurate appendix segmentation is crucial for diagnosing appendicitis.
• Manual appendix identification is time-consuming and relies on radiologist expertise.
• Automated methods are needed to improve efficiency and accuracy.

Purpose of the Study:

• To develop a fully automated deep learning approach for appendix detection in CT scans.
• To utilize a U-Net architecture for efficient and high-performance appendix segmentation.
• To evaluate the diagnostic reliability of the proposed model.

Main Methods:

• A U-Net deep learning architecture was employed for appendix segmentation.
• The model was trained on an annotated dataset of abdominal CT scans.
• Data augmentation techniques were applied to extend the training dataset.
• Hyperparameter optimization was used to refine the model's performance.

Main Results:

• The U-Net model achieved high segmentation performance with a Dice Similarity Coefficient (DSC) of 85.94%.
• Key metrics included Volumetric Overlap Error (VOE) of 23.29% and Average Symmetric Surface Distance (ASSD) of 1.24 mm.
• The model demonstrated superior performance compared to other methods, leveraging U-Net's contextual understanding.

Conclusions:

• The proposed U-Net model offers reliable appendix segmentation in CT scans.
• Deep learning shows significant potential for improving clinical outcomes in appendix detection.
• Limitations include segmentation challenges when the appendix is near other structures.