A Comparative Study of Classification Algorithms for Enhanced Lung Cancer Prediction Using Deep Learning and SOM-Based Microscopic Image Analysis

Abstract

Lung cancer is one of the top causes of cancer-related fatalities worldwide, necessitating the development of efficient early detection techniques. This study explores a hybrid approach combining deep learning and a Self-Organizing Map (SOM) for the classification of three lung cancer subtypes: adenocarcinoma, squamous cell carcinoma, and neuroendocrine tumors, using microscopic images. A pre-trained MobileNet model is employed for feature extraction, while the SOM is used for dimensionality reduction and visualization of high-dimensional data. The extracted features are then classified using various machine learning algorithms, including Random Forest, LightGBM and Decision Tree. A comparative analysis of these classifiers is conducted to assess their performance in predicting cancer types. Additionally, thresholding is applied to highlight cancerous regions in the images, enhancing the visual detection of malignant cells. Results indicate that the hybrid model provides competitive classification accuracy, with the Random Forest and Decision Tree classifiers showing particular promise. This research demonstrates the potential of combining deep learning with traditional machine learning techniques for lung cancer detection, offering a pathway toward more accurate and efficient diagnostic tools.