CLASSIFICAÇÃO DE NÓDULOS PULMONARES EM MALIGNO E BENIGNO UTILIZANDO OS ÍNDICES DE DIVERSIDADE DE SHANNON E DE SIMPSON

Abstract

Lung cancer is still the leading cause of cancer mortality worldwide, with one of the lowest survival rates after diagnosis. Therefore, early detection is important to increase the chances of curing the patient. The diagnosis is more accurate if the specialist has more information. In view of the above, this work presents a methodology for characterization about the malignancy or benignity of pulmonary nodules, acting as a second opinion for the expert. The methodology was applied in two different databases, one with 73 nodes, 26 malignant and 47 benign, and other with 1034 nodes and 517 malignant and 517 benign. The Diversity Indices of Shannon and Simpson were used as texture descriptors. The features generated were then subjected to the step of feature selection using the stepwise Discriminant Analysis. After this stage, they were classified by the Support Vector Machine (SVM) where we obtained sensitivity of 85.64%, specificity of 97.89% and accuracy of 92.78%.