Novel 3D heart left ventricle muscle segmentation method for PET-gated protocol and its verification

Area of Science:

• Medical Imaging
• Cardiovascular Imaging
• Image Segmentation

Background:

• Accurate segmentation of the left ventricle is crucial for assessing cardiac function.
• Existing methods for cardiac segmentation may have limitations with diverse PET data characteristics.

Purpose of the Study:

• To propose and validate a universal method for left ventricular myocardium segmentation using gated PET data.
• To enhance the B-spline Explicit Active Surface (BEAS) algorithm with mechanical properties for improved segmentation accuracy.

Main Methods:

• Extended the BEAS algorithm by incorporating tensile strength and bending resistance into the active model.
• Applied the segmentation model to gated PET data from 60 patients.
• Validated the method against QGS (Cedars-Sinai) and ECTb (Emory University) software, and using left ventricular phantom data.

Main Results:

• The proposed method demonstrated strong correlations with established software: R² of 0.867 for QGS and 0.845 for ECTb.
• Calculated ventricular volumes closely matched the true cavity volumes in phantom experiments.
• The segmentation method showed high agreement with reference software for volume measurements.

Conclusions:

• The developed method provides reliable left ventricular volume measurements from gated PET data.
• Results are comparable to current established segmentation techniques.
• Phantom data confirmed the physical accuracy of the volume measurements.