Automatic reorientation to generate short-axis myocardial PET images

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Cardiology

Background:

• Accurate reorientation of Positron Emission Tomography (PET) images to short-axis (SA) views is crucial for myocardial imaging diagnosis.
• Existing methods may lack efficiency or accuracy in image redirection and quantitative analysis.

Purpose of the Study:

• To develop and validate an automated system for reorienting and quantitatively analyzing myocardial PET images.
• To improve the diagnostic accuracy and efficiency of myocardial PET image interpretation.

Main Methods:

• Developed an automated reorientation system with modules for regional division (LVG-CED), myocardial segmentation (PSO-FCM), and ellipsoid fitting (LLS-SVD).
• Evaluated 36 combinations of segmentation and fitting methods using correlation coefficients (PCC, KCC, SCC).
• Quantitatively analyzed reoriented short-axis images using polar maps and assessed performance across three image classifications.

Main Results:

• The developed LVG-CED method achieved a 100% success rate in regional division.
• The combination of PSO-FCM and LLS-SVD demonstrated superior performance in correlation coefficients.
• The automated system showed >96% similarity to manual methods in quantitative analysis and excellent performance across all image classifications.

Conclusions:

• The developed algorithm system enables accurate automatic reorientation and quantitative analysis of PET myocardial metabolic images.
• The system is effective even for images affected by interference, offering a robust solution for clinical applications.