Transformative Techniques in Lung Cancer Detection: A Deep Learning Perspective

Integrating deep learning methods with heuristic approaches to enhance accuracy and optimize feature selection holds immense promise for the advancement of medical imaging and cancer diagnosis, particularly in lung cancer detection. Deep learning models often produce high-dimensional feature vectors, but not all extracted features are equally informative for lung cancer detection. Incorporating heuristic feature selection methods like whale optimization algorithm (WOA) and adaptive β-Hill climbing can effectively identify the most relevant and discriminative features. This not only improves the model's accuracy but also enhances interpretability for researchers. The synergy between deep learning and heuristic approaches can significantly enhance the ability of lung cancer detection systems. By fine-tuning hyperparameters and selecting essential features, these hybrid methods can achieve heightened accuracy, sensitivity, and specificity, ultimately leading to more dependable and precise lung cancer diagnoses. The publicly available datasets serve as standardized reference points or test cases for researchers working in the field of lung cancer to evaluate and to do the comparative performance analysis. The current state of research suggests that use of AdBet-WOA enables the support vector machine classifier to achieve remarkable accuracy rates of 99.99% for colon cancer test data, 99.97% for lung cancer test data, and 99.96% for both combined, thereby establishing its potential as one of the highly promising methods. These results align closely with benchmark figures, underscoring the comprehensive matching capabilities of our approach. In this review, we thoroughly compared methods against various independent and hybrid optimization algorithms.