eduzhai > Applied Sciences > Engineering >

AtrialJSQnet A New Framework for Joint Segmentation and Quantification of Left Atrium and Scars Incorporating Spatial and Shape Information

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 12 pages

Abstract: Left atrial (LA) and atrial scar segmentation from late gadolinium enhancedmagnetic resonance imaging (LGE MRI) is an important task in clinical practice. , to guide ablation therapy and predict treatment results for atrialfibrillation (AF) patients. The automatic segmentation is however stillchallenging, due to the poor image quality, the various LA shapes, the thinwall, and the surrounding enhanced regions. Previous methods normally solvedthe two tasks independently and ignored the intrinsic spatial relationshipbetween LA and scars. In this work, we develop a new framework, namelyAtrialJSQnet, where LA segmentation, scar projection onto the LA surface, andscar quantification are performed simultaneously in an end-to-end style. Wepropose a mechanism of shape attention (SA) via an explicit surface projection,to utilize the inherent correlation between LA and LA scars. In specific, theSA scheme is embedded into a multi-task architecture to perform joint LAsegmentation and scar quantification. Besides, a spatial encoding (SE) loss isintroduced to incorporate continuous spatial information of the target, inorder to reduce noisy patches in the predicted segmentation. We evaluated theproposed framework on 60 LGE MRIs from the MICCAI2018 LA challenge. Extensiveexperiments on a public dataset demonstrated the effect of the proposedAtrialJSQnet, which achieved competitive performance over the state-of-the-art.The relatedness between LA segmentation and scar quantification was explicitlyexplored and has shown significant performance improvements for both tasks. Thecode and results will be released publicly once the manuscript is accepted forpublication via this https URL.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...