Analysing Feature Extraction Methods for Enhanced Accuracy in Lung Cancer Detection

Accurate detection and diagnosis of lung cancer from computed tomography (CT) images play a critical role in improving patient outcomes and reducing mortality rate. Feature extraction techniques are vital in extracting discriminative information from lung CT images for improving the accuracy and efficiency of diagnosis. This paper presents a comprehensive methodology and analysis for evaluating and comparing different feature extraction techniques for the accurate the diagnosis of lung cancer using lung input CT images. The methodology involves the collection and preprocessing of a dataset of lung cancer images using a median filter and the CT input images is segmented to identify a region of interest by applying Thresholding segmentation methods, followed by the analysis of each feature extraction techniques like Gray-Level Co-occurrence Matrix (GLCM), Speed Up Robust Features (SURF), Scale Invariant Feature Transform (SIFT) and Principal component analysis (PCA) to extract relevant features. A dataset comprising lung CT images, including cancerous and non-cancerous cases, is utilized for the evaluation. The feature extraction methodologies are employed to lung CT images for extracting relevant features. The extracted features are quantitatively analysed and evaluated based on statistical measures and performance metrics such as accuracy, precision, recall and F1 score. This helps to find an optimum feature extraction technique which is used for accurate diagnosis of lung cancer. The results demonstrate that all four feature extraction techniques show a promising outcome in the evaluation of CT images for the identification of lung cancer. The SIFT-based features combined with SVM achieve the highest accuracy of 96%, outperforming the other combinations.