Automatic Organ Segmentation for CT Scans Based on Super-Pixel and Convolutional Neural Networks

Area of Science:

• Medical Imaging
• Computer Vision
• Artificial Intelligence

Background:

• Accurate organ segmentation in computed tomography (CT) scans is vital for diagnosis and treatment.
• Manual segmentation is time-consuming and prone to variability.
• Automated methods are needed to improve efficiency and consistency.

Purpose of the Study:

• To develop and evaluate an automated organ segmentation framework for CT scans.
• To enhance the accuracy and efficiency of liver and lung segmentation.
• To assist physicians in disease diagnosis by providing precise organ boundaries.

Main Methods:

• A hybrid framework combining Simple Linear Iterative Clustering (SLIC) super-pixels, Support Vector Machine (SVM) classification, and Convolutional Neural Networks (CNNs).
• SLIC segments images into super-pixels, SVM classifies signatures for rough edges, and CNN refines boundaries at the pixel level.
• The framework was tested on abdominal CT scans for liver segmentation and high-resolution CT (HRCT) scans for lung segmentation.

Main Results:

• The automated framework achieved high segmentation accuracy for both liver (Dice coefficient: 97.43%) and lungs (Dice coefficient: 97.93%).
• Liver segmentation was performed efficiently, with a processing time of 38 seconds per slice.
• The method demonstrated precise and efficient organ detection in the tested datasets.

Conclusions:

• The proposed framework offers a precise and efficient solution for organ segmentation in CT and HRCT scans.
• This automated approach can significantly aid radiologists in distinguishing diseases and improving diagnostic workflows.
• The high Dice coefficients indicate the framework's potential as a favorable method for clinical applications.