An Efficient Pipeline for Abdomen Segmentation in CT Images

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Biomedical Engineering

Background:

• Computed tomography (CT) scans often present image quality challenges like discontinuous edges and poor contrast, hindering accurate abdomen segmentation.
• Existing segmentation techniques struggle with these handicaps, impacting their utility in real-time diagnostic systems.
• Efficient abdomen segmentation is crucial for subsequent analyses such as feature selection and classification in medical imaging.

Purpose of the Study:

• To develop an efficient and robust statistical pipeline for abdomen segmentation in CT scans.
• To create a method that is unaffected by common image quality disadvantages inherent in CT imaging.
• To provide a reliable foundation for real-time diagnostic systems requiring precise abdominal region analysis.

Main Methods:

• A statistical pipeline integrating intensity-based, morphological, and histogram-based procedures was designed.
• The pipeline was optimized using 16 training CT images with inherent segmentation disadvantages.
• Performance was evaluated on 16 test and 26 validation CT images using six key performance metrics.

Main Results:

• The proposed method demonstrated high segmentation accuracy across training, testing, and validation datasets.
• Achieved average Jaccard index of 98.95/99.36/99.57%, Dice coefficient of 99.47/99.67/99.79%, and classification accuracy of 99.38/99.63/99.87%.
• The pipeline proved effective in overcoming common CT image handicaps, ensuring reliable abdomen segmentation.

Conclusions:

• A novel statistical pipeline effectively performs abdomen segmentation in CT scans, robust to common image quality issues.
• The developed method offers a significant advancement for applications requiring precise abdomen segmentation, including organ and tumor analysis.
• This study provides a detailed and reliable approach for abdomen segmentation, supporting the development of advanced real-time diagnostic tools.