Algorithms for Assisted Diagnosis of Solitary Lung Nodules in Computerized Tomography Images

The present work seeks to develop a computational tool to suggest the malignancy or benignity of Solitary Lung Nodules by means of analyzing texture and geometry measures obtained from computarized tomography images. Three groups of methods are proposed with the purpose of suggesting the diagnosis for such nodules. The methods are divided according to their common characteristics. Group I uses four geostatistical functions denominated semivariogram, semimadogram, covariogram and correlogram to analyze nodules’ texture. Group II describes measures based only on the nodules’ geometry, such as convexity, sphericity, and measures based on the curvature. Finally, Group III analyzes the Gini coefficient and nodules’ skeleton, which take into account both the nodules’ geometry and texture. A sample with 36 nodules, 29 benign and 7 malignant, was analyzed and the preliminary results of these methods are very promising in characterizing lung nodules. All groups of proposed methods have the area under the ROC curve value above 0.800, using Fisher’s Linear Discriminant Analysis and Multilayer Perceptron Neural Networks. This means that the proposed methods have great potential in the discrimination and classification of Solitary Lung Nodules.