Abdomen tissues segmentation from computed tomography images using deep learning and level set methods.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiotherapy

Background:

• Accurate segmentation of abdomen tissues (liver, kidney) is vital for radiation therapy planning.
• Low contrast between organs complicates automated segmentation.
• Existing methods struggle with precise delineation of abdomen anatomy.

Purpose of the Study:

• To develop an automated abdomen tissue segmentation method using deep learning.
• To improve the accuracy of liver and kidney segmentation in medical images.
• To enhance the efficacy of radiation therapy planning through precise anatomical segmentation.

Main Methods:

• An attention-based deep learning model (U-net with attention) was employed for initial segmentation.
• Image cropping was utilized to focus on relevant abdominal regions.
• Level set evolution with three energy terms was applied for segmentation optimization.

Main Results:

• The method achieved high mean Dice scores for liver segmentation: 96.2% (FLARE21) and 95.1% (LiTS).
• Excellent mean Dice scores were obtained for kidney segmentation: 96.6% (FLARE21) and 95.7% (LiTS).
• The proposed approach demonstrated superior performance compared to other segmentation techniques.

Conclusions:

• The attention-based deep learning method provides accurate abdomen tissue segmentation.
• This technique offers significant improvements for radiation therapy planning.
• The model shows strong potential for clinical application in medical imaging analysis.