CT-based multi-organ segmentation using a 3D self-attention U-net network for pancreatic radiotherapy

Area of Science:

• Radiotherapy
• Medical Imaging
• Artificial Intelligence

Background:

• Manual segmentation of organs-at-risk (OARs) is time-consuming and labor-intensive in radiotherapy planning.
• Accurate OAR segmentation is crucial for effective treatment and minimizing side effects.

Purpose of the Study:

• To develop a deep learning-based method for rapid and accurate segmentation of multiple OARs in the pancreas.
• To expedite the radiotherapy treatment planning process.

Main Methods:

• A retrospective study of 100 patients with CT scans.
• Development of a 3D deep attention U-Net for segmenting eight OARs (large bowel, small bowel, duodenum, left kidney, right kidney, liver, spinal cord, stomach).
• Performance evaluation using Dice similarity coefficient (DSC), sensitivity, specificity, Hausdorff distance 95% (HD95), mean surface distance (MSD), and residual mean square distance (RMSD).

Main Results:

• The deep learning model achieved high accuracy in segmenting OARs, closely matching manual contours.
• Mean DSC values ranged from 0.86 ± 0.06 (duodenum) to 0.96 ± 0.01 (liver).
• Quantitative metrics demonstrated the method's effectiveness in precise organ delineation.

Conclusions:

• The proposed deep learning method significantly accelerates radiotherapy treatment planning through rapid multi-OAR segmentation.
• This approach holds potential for improving dose delivery accuracy and reducing gastrointestinal toxicity in pancreatic adaptive radiotherapy.