An improved Scale-Invariant Feature Transform-based method for detecting tumor tissue motion induced by respiration

Area of Science:

• Medical Imaging
• Computational Anatomy
• Biomedical Engineering

Background:

• Scale-Invariant Feature Transform (SIFT) is a key algorithm for target recognition and tracking.
• Accurate measurement of tumor tissue displacement during respiration is crucial for effective treatment planning.

Purpose of the Study:

• To evaluate the efficacy of the SIFT algorithm in quantifying tumor tissue displacement caused by thoracic motion during respiration.
• To compare the performance of SIFT with cross-correlation algorithms in analyzing phantom motion.

Main Methods:

• A thoracic phantom with eight synthetic nodules was used.
• Nodules were subjected to translational (10 pixels) and rotational (5°–45°) movements.
• SIFT and cross-correlation algorithms were applied for motion analysis, with t-tests assessing significance.

Main Results:

• No significant difference was found between SIFT and cross-correlation for translational motion detection.
• SIFT demonstrated superior performance in detecting rotational distortions.
• The t-test showed no significant differences in motion detection across various directions (p > 0.05).

Conclusions:

• The SIFT algorithm is a reliable tool for calculating tissue distortion during respiratory motion.
• SIFT's ability to detect rotational changes makes it valuable for analyzing complex thoracic movements.