Multiple Instance Learning for Benign vs. Malignant Classification of Lung Nodules in CT Scans

Multiple Instance Learning (MIL) is used to classify nodules as malignant or benign on 275 benign cases and 96 malignant samples from the publicly available Lung Image Database Consortium Image Collection (LIDC-IDRI) for Computer Aided Diagnosis without predefined Regions of Interest (ROIs) from lung cancer screening thoracic CT scans. By fusing the results of two texture features based on the Gray Level Co-occurrence Matrix (GLCM) and Local Binary Patterns (LBP), a classification was performed over 5 fold cross validation using Support Vector Machines for Multiple-Instance Learning (MI-SVM) classifier. The resulting average Area Under Curve, Specificity, Sensitivity, and Accuracy were: 0.9696, 0.9855, 0.6979 and 0.9111 respectively.