Lung cancer classification based on enhanced deep learning using gene expression data

Abstract

Lung cancer is one among the life threatening of cancer globally. The estimation of the World Cancer Research Fund International that in 2022, this disease has recorded the diagnosis of 1.8 million fresh cases of this disease. Medical professionals may safely and effectively treat the patient when they make a proactive diagnosis and classification of their condition. The availability of microarray technology has paved the way for investigations of genes and their relationships to many illnesses, including lung cancers. Prior studies have suggested gene selections using IWOA (Improved Whale Optimization Algorithm). After that, an MLSTM (Modified Long Short-Term Memory) Network is taken into account for the categorization of lung cancer. Nevertheless, because the input data is represented in different scales, LSTMs are prone to overfitting and the salient trait (with a decreased scale) may become useless as a result of other features having values on a higher scale. To get over those problems in this research, introduce an enhanced model for lung cancer classification. In this work, initially, data pre-processing is performed to normalization the data scale using min max normalization. Gene selection is done by using IWOA. Finally, ECNN (Enhanced convolution neural networks) is considered for classifying lung cancer. The experimental results are executed on Matlab 2013b using the Kent Ridge Bio-Medical Dataset, the proposed ICNN framework in this study was contrasted with the pre-existing MIMOGA, SMO, and MLSTM techniques. The results of this work's proposed model demonstrate its effectiveness when evaluated using performance measures such as values for recall, precision, accuracy, and f-measure.