A pre-trained 3D Autoencoder is used to reconstruct an identity segmentation result.
Optimum setting parameters are useful for segmentation and computational performance.
Propose volume patch extraction from Delayed Enhancement Magnetic Resonance Imaging.
A fusion of 3D Autoencoder with 3D U-Net proves better myocardial segmentation.
An anatomical network is used to segment the myocardium and left ventricular cavity.
Objectives: In this work, a new deep learning model for relevant myocardial infarction segmentation from Late Gadolinium Enhancement (LGE)-MRI is proposed. Moreover, our novel segmentation method aims to detect microvascular-obstructed regions accurately. Material and methods: We first segment the anatomical structures, i.e., the left ventricular cavity and the myocardium, to achieve a preliminary segmentation. Then, a shape prior based framework that fuses the 3D U-Net architecture with 3D Autoencoder segmentation framework to constrain the segmentation process of pathological tissues is applied. Results: The proposed network reached outstanding myocardial segmentation compared with the human-level performance with the average Dice score of ‘0.9507’ for myocardium, ‘0.7656’ for scar, and ‘0.8377’ for MVO on the validation set consisting of 16 DE-MRI volumes selected from the training EMIDEC dataset. Conclusion: It is concluded that our approach's extensive validation and comprehensive comparison against existing state-of-the-art deep learning models on three annotated datasets, including healthy and diseased exams, make this proposal a reliable tool to enhance MI diagnosis.Le texte complet de cet article est disponible en PDF.
Keywords : Myocardial infarction segmentation, LGE-MRI, Microvascular-obstructed regions