Efficacy of Machine Learning Models in Lung Cancer Detection: An Emphasis on Bees with ICA Hybrid Feature Extraction

Abstract

The rapid and precise diagnosis of specific lung cancer types, including adenocarcinoma of the left lower lobe, large cell carcinoma of the left hilum, and squamous cell carcinoma of the left hilum, has become paramount in the realm of medical imaging. This research aims to harness advanced image processing techniques to extract disease-specific features and employ a hybrid algorithm for feature reduction, ultimately facilitating the accurate classification of these diseases. Features were meticulously extracted from medical images capturing the diseases above. A novel hybrid algorithm, which fuses the strengths of the Bees Algorithm and Independent Component Analysis (ICA), was introduced to address the challenge of high dimensionality. Following feature reduction, a battery of machine learning classifiers—including k-nearest Neighbours (kNN), Support Vector Machines (SVM), Logistic Regression, Linear Regression, and Random Forest—was applied to the curated features. The classifiers' performance metrics were rigorously evaluated, including accuracy, time complexity, precision, recall, and F1 score. Preliminary findings underscore the efficacy of the hybrid feature reduction technique in preserving salient disease markers, thus amplifying the classifiers' accuracy and computational efficiency. This study propounds a methodological advancement in detecting specific lung cancer types through image processing. The synergistic application of the hybrid feature reduction algorithm and machine learning classifiers offers promise in reshaping contemporary diagnostic paradigms, laying the groundwork for the next generation of diagnostic tools in lung cancer care.