A Deep Learning Approach for Tissue and Calcium Classification

G. MATO; ARIEL CURIALE; FLAVIO COLAVECCHIA; DIEGO BUSTOS

San Luis

Conferencia; Conferencia Py Data; 2018

Tissue classification is of paramount importance in many biomedical fields. For example, cardiac function is determined by structural and functional features. In both cases, the analysis of medical imaging studies requires to detect and segment the myocardial tissue. Regarding to the calcium classification, another interesting example of classification, is to analyze calcium deposit to determine the efficiency and uniformity in the formation of bone tissue from human mesenchymal stem cells. In particular, this analysis is carried out by using different features such as the density and area of the calcium deposits which are derived from the calcium classification in the images.In this work we propose to use a deep learning technique to assist the automatization of tissue and calcium classification in Microscopy images and Cardiac Magnetic Resonance image (CMR) with a spatial resolution of 0.961 x 0.961 µm and 1.36 x 1.36 mm respectively. The deep learning approach used presents several improvements to previous works, for instance, the use of the Jaccard distance as optimization objective function. Also, it is incorporated a residual learning strategy and a batch normalization layer to train the fully convolutional neural network. The method was entirely implemented in python by using the Keras and Tensorflow libraries for GPU. Our results demonstrate that this architecture outperforms previous approaches based on a similar network architecture, and that provides a suitable?