Lung nodule Detection and Classification using Deep Neural Network

Lung nodule detection is a complicated process. It is an interesting topic in the field of biomedical imaging. To detect lung nodules Computer-Aided Detection (CAD) plays an important role. Most of the CAD system consists of a nodule detector and a feature based classifier. In our method the nodule detection part, the CT images processed through Adaptive Histogram Equalization (AHE), thresholding, and edge detection. Gray Level Co-occurrence Matrix (GLCM) is used to extract features that is the part of the classifier. Feature such as Energy, Entropy, Correlation, Homogeneity, Dissimilarity, Contrast are extracted. A dataset is created tabulating the values of the features. This dataset is used to train the Deep Neural Network (DNN) model. This proposed model is able to label the nodules as benign or malignant with 88.25% of accuracy. LIDC and Kaggle database are used for the experiment purpose. For more: Lung nodule Detection and Classification using Deep Neural Network – IEEE Conference Publication