GPU-accelerated lung CT segmentation based on level sets and texture analysis.

Area of Science:

• Medical Imaging
• Computer Vision
• Image Processing

Background:

• Accurate lung segmentation is crucial for diagnosing and monitoring thoracic diseases using CT scans.
• Existing methods often struggle with robustness and efficiency in complex 3D datasets.

Purpose of the Study:

• To develop a novel semi-automatic method for robust and accurate lung segmentation in thoracic CT datasets.
• To leverage texture features and GPU acceleration for improved performance.

Main Methods:

• A 3D active surface model based on level sets is employed.
• Texture features, including grey-level co-occurrence matrices and Gabor filters, are integrated.
• Manual initialization on 2D slices guides the semi-automatic segmentation process.
• Graphics Processing Unit (GPU) acceleration is utilized to enhance computational speed.

Main Results:

• The proposed method achieved high segmentation accuracy on the LCTCS 2017 challenge dataset.
• Performance was found to be competitive when compared to other existing segmentation techniques.
• The integration of texture features improved the robustness of the lung segmentation.

Conclusions:

• The developed semi-automatic method offers an accurate and efficient solution for lung segmentation in thoracic CT.
• GPU acceleration significantly boosts the performance of the 3D active surface algorithm.
• This approach shows promise for clinical applications requiring precise lung volume analysis.