Largest Versus Smallest Nodules Marked by Different Radiologists in Chest CT Scans for Lung Cancer Detection

In this paper, we present a novel approach to find and select texture features of solitary pulmonary nodules (SPNs) detected by computed tomography (CT) and evaluate the performance of grafted decision tree based classifier in differentiating benign from malignant as well as from metastasis SPNs. We compared the results of smallest as well as largest nodule of a patient visible in different slices of CT scan and conclude that by taking the slice of a patient with largest area nodule is better in classifying the SPNs in 3 classes as compared to considering the nodules with smallest area of the same patient. It also reflects that specificity as well as sensitivity is much better which can further assist the physician in yielding the right decision at right time in the detection and diagnosis of lung cancer. This study reveals that there could be a significant improvement in the field of lung nodule detection at an early stage of lung cancer and also ensures that unnecessary biopsies can be avoided using the proposed methodology and feature set.