Classification of Lung Cancer Using Hybrid Convolutional Neural Network Based Bidirectional Long Short-Term Memory

Cancer is a deadly disease caused by a grouping of genetic diseases and several varieties of biochemical abnormalities. Colon and lung cancer appeared as two of the leading causes of death and human disability. The histological diagnosis of such tumors is typically the most critical factor in selecting the optimal course of therapy. The primary aim of this study is to identify malignant lung nodules in a given lung image and to categorize lung cancer and its seriousness. To identify lung nodules, Deep Learning (DL) models are used to categorize the lung cancer stages. A lung nodule is detected using the Convolutional Neural Network (CNN), which identifies a nodule that is precisely segmented and categorized. The proposed approach is divided into two sections: the first is used to categorize normal and aberrant behavior, and the second is used to categorize the various lung cancer stages. In the classification stage, characteristics based on texture and intensity are extracted. The right edge image and morphological thinning operation are applied to improve segmentation performance. The hybrid CNN based Bidirectional Long Short-Term Memory (CNN-Bi-LSTM) model achieves better classification results using performance metrics like accuracy, specificity, sensitivity, and f1-score values of about 99.41%, 99.39%, 99.43% and 99.18% respectively comparing with existing methods.