Description

Lung diseases identification based on analysis and processing of medical images is important to assist medical doctors during the diagnosis process. In this context, this paper proposes a new feature extraction method based on human tissue density patterns, namely Analysis of Human Tissue Densities in Lung Diseases. The proposed method uses human tissues radiological densities, in Hounsfield Units, to perform the features extraction on thorax computerized tomography images. We compared the proposed method against the Gray Level Co-occurrence Matrix and Statistical Moments to accomplish the performance evaluation alongside four machine learning classifiers. Overall, the results revealed that the proposal achieved higher accuracy ratios while it took the lowest runtime in all performed experiments. Thus, we consider our proposal as a valid alternative to be used in real-time applications.