TY  - GEN 
T1  - Left Ventricle Segmentation in Dynamic 82Rb PET/CT Using Deep Convolutional Neural Networks
A1  - Amirian, Mohammadreza
A1  - Chevalley, Arthur
A1  - Andrearczyk, Vincent
A1  - Klein, Ran
A1  - DeKemp, Robert
A1  - Moulton, Eric
A1  - Kamani, Christel H.
A1  - Prior, John O.
A1  - Jreige, Mario
A1  - Depeursinge, Adrien
Y1  - 2025
N2  - Precise delineation of the left ventricle (LV) enables advanced analyses of endo-to-epicardium blood perfusion patterns and their gradients in PET, which can support various clinical applications, including coronary microvascular dysfunction. Approaches like simple thresholding or semi-automatic curve fitting often fall short in accurately capturing the LV boundaries in dynamic $^{82}$Rb PET/CT. They also require expert manual input.
This study presents a reliable manual, multi-modal LV delineation strategy, and a fully automatic segmentation algorithm using deep convolutional neural networks (CNNs).
ER  -