Computer-Aided Detection Algorithm Of Lung Nodules Based On Hessian Matrix

ObjectiveTo study the enhancement effect of multi-scale enhancement algorithm based on Hession matrix for circle or like-circle suspected region of lung nodules, realize the computer-aided detection of lung nodules, alleviate the doctor's burden and improve the detection rate of lung nodules.Materials and MethodsIn our study, 44 images were selected and there were 90 nodules. All images were produced by Siemens Somatom 16 CT workstation with 512×512 pixel matrix size, 16-bit allocation, 12-bit depth and 4096 gray-level. The pixel size is 0.74㎜×0.74㎜ with 5㎜ or 7㎜ slice thickness. All images were stored of DICOM.By using Windows XP and VC++6.0 to develop the software, we realized to directly read the 2-Dimensional CT images of DICOM. Gray-level and region-grow algorithm were applied to segment the lung fields, achievdied to delete redundant information and reduce noise by preprocessing CT images. Then we use the gaussian function to simulate the lung nodules, construct Hessian matrix, realize the enhancement of the different nodules by using different scales, segment the ROI of spicious nodules regions, and then extract the feature of the spicious nodules. An experienced radiologist with 10 years experience was asked to detect all the nodules without CAD. And one month later the same radiologist was asked to review all the images to detect nodules with CAD, then the ROC curve was used to evaluate the capability, the we could get the area of ROC.ResultsThe selective multi-scale enhancement algorithm can enhancement the circle or like-circle regions and restrain the line-like regions well. ROC analysis was used to assess the performance of our CAD sytem, the area of ROC(Az) is 0.95 with CAD, but without CAD, the area is 0.88.ConclusionThe multi-scale enhancement algorithm based on Hessian matrix can improve the detection rate of cirle or like-circle region of lung nodules effectively, but the FPF(false positive fraction) is higher. This method is useful because it can improve early detection of nodules in lung cancer.