Proceedings of the Statistical Atlases and Computational Modelling of the Heart Workshop (In Press) - Left ventricle quantification through spatio-temporal CNNs

Congreso

Resumen *

Cardiovascular diseases are among the leading causes of deathglobally. Cardiac left ventricle (LV) quantification is known to be one ofthe most important tasks for the identification and diagnosis of suchpathologies. In this paper, we propose a deep learning method that in-corporates 3D spatio-temporal convolutions to perform direct left ven-tricle quantification from cardiac MR sequences. Instead of analysingslices independently, we process stacks of temporally adjacent slices bymeans of 3D convolutional kernels which fuse the spatio-temporal infor-mation, incorporating the temporal dynamics of the heart to the learnedmodel. We show that incorporating such information by means of spatio-temporal convolutions into standard LV quantification architectures im-proves the accuracy of the predictions when compared with single-slicemodels, achieving competitive results for the LVQuan 2018 Challenge. Información suministrada por el agente en SIGEVA