A Comparative study of Lung Cancer detection using supervised neural network

Abstract

Different data analysis and classification methods have been used for diagnosis and detection of lung carcinoma. Since the cause of lung carcinoma stay incomprehensible, circumvention becomes unattainable, hence early revealing of tumor in lungs is the first way for treatment of lung carcinoma. Therefore, biomedical image processing and machine learning is the process used to classify the presence of lung carcinoma. The motive of this study is to develop precision and determine specific value for detection of lung carcinoma at an earlier stage using a mixture of biomedical image processing technique and knowledge Discovery in Data. The image of the lungs that is obtained from the CT(Computed Tomography) scan images is being pre-processed and the segmentation is being done in the Region Of Interest(ROI), Random Forest method have been applied to classify the distinct features. Saliency Enhancement has been done and using SURF(Speeded Up Robust Features) algorithm features such as entropy, co-relation, energy and variance have been extracted from the Saliency Enhanced images with the help of a SVM (Support Vector Machine) Classifier. Classification identifies whether the image provided is healthy or unhealthy(carcinomic). Minimum objective vs. Number of function evaluation plot is used to study the performance of the method. This entire process has been done using two algorithms - random forest algorithm and SVM classification. The best result is obtained using SVM classification. This procedure is total 94.5% efficient, 74.2% sensitive, 66.3% recall and 77.6% specific.