Aprendizado Profundo para a classificação de blocos de tecidos pulmonares em imagens de Tomografia Computadorizada de Alta Resolução

• Aline Dartora
• Lucas Ferrari de Oliveira

DOI: https://doi.org/10.5753/sbcas.2017.3707

Resumo

Interstitial lung disease (ILD) involves several imaging patterns that are observed in high resolution computed tomography (HRCT). The automated tissue characterization is an essential component of a computer aided diagnostic (CAD) system for ILD research. Deep convolutional neural networks learns features directly from training data rather than extracting them manually, avoiding the need for feature extrator optimization. In this context, the present work investigate about a classification method using a deep learning program, in order to improve the performance of CAD systems for the diagnosis of ILDs. Preliminary results achieved a general recognition rate of 74.9 % in the classification of the five classes of ILD refered in this study.