Abdominal muscle segmentation from CT using a convolutional neural network.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Computed Tomography (CT) is crucial for diagnosing diseases and monitoring treatment response, often by assessing muscle mass.
• Manual segmentation of abdominal muscles on CT slices is time-consuming and labor-intensive for radiologists.
• Changes in abdominal muscle mass are key indicators of patient treatment efficacy.

Purpose of the Study:

• To develop and evaluate a fully convolutional neural network (CNN) for automated abdominal muscle segmentation on CT images.
• To provide an efficient and accurate tool for quantifying muscle mass changes during patient treatment.

Main Methods:

• A fully convolutional neural network (CNN) architecture was implemented for semantic segmentation of abdominal muscles.
• The CNN model was trained and validated on a dataset of CT scans.
• Performance was assessed using metrics including Dice similarity coefficient, precision, and recall on an independent test set.

Main Results:

• The CNN achieved a high mean Dice similarity coefficient of 0.92.
• The method demonstrated excellent performance with a mean precision of 0.93 and a mean recall of 0.91.
• The automated segmentation significantly reduced the time required compared to manual methods.

Conclusions:

• The developed CNN provides an effective and automated solution for abdominal muscle segmentation in CT.
• This tool can streamline clinical workflows, enabling faster and more efficient monitoring of patient treatment progress.
• The CNN-based approach offers a valuable advancement for clinical applications in medical imaging analysis.