Lung nodule detection by KNN classifier and active contour modelling and 3D visualization

In this paper, an automatic computer-aided detection (CAD) scheme for lung nodules detection and 3D visualization of them is proposed. There are four steps in this method. In the first step, we use median filter for removing noise from slices and morphological operation for image enhancement then we segment lung regions from the CT data by using adaptive threshold algorithm and active contour modeling. The second step is nodule detection that contains two parts: feature extraction and classification. For extracting feature, we use a 2D stochastic features for accurate nodule detection and a 3D anatomical features for reducing the value of false positive(FP). Furthermore, KNN classifier is used in this paper. In the third step, the nodule contours is extracted by active contour modeling. In the final step, 3D visualization technique is applied on the segmented nodule to represent better visual results. At the end, experimental results of our method are indicated with good performance comparing with some other efficient methods. We achieved 90% detection rate and 5.63FP/Scan.