Optical Flow Methods for Lung Nodule Segmentation on LIDC-IDRI Images

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Image Segmentation

Background:

• Lung nodule segmentation is crucial for lung cancer detection in Computer-Aided Diagnosis (CAD) systems.
• Traditional segmentation methods rely on morphology or intensity, often lacking temporal analysis.
• CT scans, composed of sequential 2D slices, resemble video frames, suggesting potential for motion-based techniques.

Purpose of the Study:

• To investigate the efficacy of optical flow methods for lung nodule segmentation in CT scans.
• To introduce a novel application of motion-based segmentation techniques to medical imaging.
• To provide a comparative analysis of different optical flow algorithms for this task.

Main Methods:

• Implementation and application of Farneback, Horn-Schunck, and Lucas-Kanade optical flow algorithms.
• Processing of DICOM 2D image slices from CT scans as sequential frames.
• Comparative analysis of the performance of optical flow methods in segmenting lung nodules.

Main Results:

• Optical flow methods demonstrate effectiveness in segmenting lung nodules from CT scans.
• Thin-sliced CT scans are suitable for motion-based analysis due to their sequential nature.
• The study validates the potential of optical flow for improving nodule segmentation.

Conclusions:

• Optical flow methods offer a promising novel approach for lung nodule segmentation in CT imaging.
• This technique can enhance the capabilities of CAD systems for lung cancer diagnosis.
• Further research can optimize optical flow methods for improved segmentation efficiency and accuracy.