Effective Categorization of Lung Abnormalities using Squeezenet Features and Machine Learning Models

Lung cancer is the primary cause of cancer-related fatalities in the developed countries. This is due to the changes in lifestyle of the peple and also it is clinically evident only after progressing to advanced stages. Therefore, early detection of lung cancer is essential for patient survival. X-ray, computer tomography and magnetic resonance imaging approaches are used to detect the lung cancer. In this study, it is aimed to develop CAD system to classify the different types of lung cancers using histopathological images. Deep learning techniques are widely used techniques in the medical image classification. Results show that logistic regression is found to achieve an accuracy of 96% in differentiating the tumor cells. High value of 94% for MCC clearly indicates that the features extracted using squeezenet features has the ability to categorize abnormality. SVM with RBF kernel displays better peformance in identifying the categories of tumor with an accuracy of 89.8%. k-NN model with by distance weights is found to show comparatively high peformance in detecting the abnormalities in the images with accuracy and AUC values of 98.4% and 99.8% respectively. This study is among the few studies which has achieved significant performance for multiclass classification.