Enhanced Cancer Classification Using Optimized Deep Learning Approaches: A Novel Framework Integrating RNA Sequence Analysis and AI Techniques

Abstract

Lung cancer is one of the most aggressive diseases globally, responsible for over 9 million deaths annually. Early staging is critical for improving recovery rates, and RNA sequence analysis has emerged as a vital technique in this process. Recent advancements in AI have significantly enhanced the efficiency and accuracy of human genomics analysis. This study proposes Lung Cancer Classification Using Optimized Deep Learning Approaches(CC-ODLA) to classify various lung cancer typesspecifically kidney renal clear cell carcinoma (KIRC), breast invasive carcinoma (BRCA), lung squamous cell carcinoma (LUSC), lung adenocarcinoma (LUAD), and uterine corpus endometrial carcinoma (UCEC)—using deep learning methods. The first approach combines BPSO-DT and CNN to analyzetumor RNA-seq gene expression data. The proposed methodology consists of three phases: (1) preprocessing, which optimizes high-dimensional RNA-seq data to select optimal features and converts them into 2D images; (2) data augmentation, which increases the dataset size from 2,086 samples to five times larger to mitigate overfitting; and (3) deep CNN architecture, employing a two-layer convolutional framework for feature extraction and classification, achieving a testing accuracy of 96.90%. The second approach introduces an AFOECNN, designed to enhance classification accuracy in lung cancer datasets. This approach also involves three phases: (1) preprocessing using K-Means Clustering (KMC) to reduce noise and handle missing data; (2) feature subset selection via the AFO algorithm to identify significant features based on fitness values; and (3) classification using the Enhanced CNN (ECNN) algorithm, which demonstrates superior precision, recall, F-measure, and accuracy compared to existing algorithms while maintaining lower time complexity. The comparative results indicate that the proposed methods not only improve classification performance but also offer efficient computational resource management, underscoring the transformative potential of AI in smart healthcare solutions.Using Adaptive Firefly Optimisation, the Enhanced CNN improves classification performance with 98% testing accuracy, 93.21% precision, 98% recall, and 91.82% F1-score. However, the combination BPSO-DT and CNN method achieves 96.90% testing accuracy.