An efficient lung cancer detection using optimal SVM and improved weight based beetle swarm optimization

Lung cancer is caused by uncontrolled cell growth in the lung tissue. Early detection of lung cancer is important for correct detection and availability of treatment. Diagnosing lung cancer by radiologists using medical imaging modalities such as computed tomography (CT) images is becoming gradually difficult and time-consuming. This research focused on developing a hybrid framework for lung cancer diagnosis through CT images combining a Discrete Local Binary Pattern (DLBP) and a Hybrid Wavelet Partial Hadamard Transform (Hybrid WPHT). The proposed model is first pre-processed using adaptive median filtering in lung cancer CT images. Then, DLBP and hybrid WPHT-based features are extracted from the pre-processed images. The adaptive Harris-Hawk optimization (AHHO) approach is then used to optimize the feature selection process by reducing the dimensionality of features. The optimal SVM (OSVM)-based classification and the improved weight-based beetle swarm (IW-BS) algorithm are used for parameter tuning in the classifier. The IW-BS approach is used to fine-tune the OSVM’s regularization and kernel parameters, significantly improving classifier performance. The proposed method categorizes input images as normal, benign, or malignant. The results show that the proposed methodology is superior in accuracy, precision, recall, specificity, F-score, running time, AUC (area under the curve) and ROC (receiver operating characteristics).