OBJECTIVE: The objective was to prevent loss of some implicit structural and local contextual information of lung nodules by current one- (1D) or two-dimensional (2D) schemes. MATERIALS AND METHODS: The testing data set used in this study consisted of computed tomographic scans from 196 different patients in Jilin Tumor Hospital, which consisted of 8428 sections including 108 nodules. By the proposed support vector machine based on three dimensional matrix patterns (SVM(3Dmatrix)) which improves the classifier of SVM, 3D volume of interest of suspected lung nodules can be used directly as the training samples. The 3D scheme may effectively reduce the large numbers of false positives (FPs) by current 1D and 2D schemes. RESULT: Five computer-aided diagnosis (CAD) schemes were investigated for the same 196-case database. SVM(3Dmatrix) achieved a 98.2% overall sensitivity with 9.1 FPs per section, which was in general superior compared to the other four CAD schemes for our application.
OBJECTIVE: The objective was to prevent loss of some implicit structural and local contextual information of lung nodules by current one- (1D) or two-dimensional (2D) schemes. MATERIALS AND METHODS: The testing data set used in this study consisted of computed tomographic scans from 196 different patients in Jilin Tumor Hospital, which consisted of 8428 sections including 108 nodules. By the proposed support vector machine based on three dimensional matrix patterns (SVM(3Dmatrix)) which improves the classifier of SVM, 3D volume of interest of suspected lung nodules can be used directly as the training samples. The 3D scheme may effectively reduce the large numbers of false positives (FPs) by current 1D and 2D schemes. RESULT: Five computer-aided diagnosis (CAD) schemes were investigated for the same 196-case database. SVM(3Dmatrix) achieved a 98.2% overall sensitivity with 9.1 FPs per section, which was in general superior compared to the other four CAD schemes for our application.