Early detection and classification of malignant lung nodules from CT images: An optimal ensemble learning

Malignant pulmonary nodules must be identified promptly to improve the life chances of lung disorder patients. Lung cancer is the most severe type of cancer, and early detection directly impacts the chance of recovery. Despite multimodal imaging techniques for diagnosis, the accuracy of malignant nodule determination remains unreliable. To increase the accuracy of lung cancer prediction, this paper proposes a novel approach called the Multipopulational Neighborhood Particle Swarm Optimized Modified Ensemble Faster Learning (MNPS-MEFL). The Lung Image Database Consortium-Image Database Resource Initiative (LIDC-IDRI) dataset, which contains 1018 thoracic computed tomography (CT) cases, is used to evaluate the detection ability of the proposed system. However, the raw CT images with large intensity variations and data noises affect classification accuracy. Therefore, preprocessing steps, including image enhancement and denoising, are performed before accurately determining benign and malignant nodules using the proposed MNPS-MEFL approach. To enhance the accuracy of the classifier, the performance influencing factors of the ensemble support vector machine are adaptively tuned using a multi-populational particle swarm neighborhood learning optimizer (MPSNLO). The effectiveness of the proposed approach is evaluated using various evaluation measures, including accuracy, sensitivity, specificity, precision, f-score, false alarm rate, and execution time latency. The simulation results demonstrate the superior performance of the proposed MNPS-MEFL approach over existing methods, achieving a classifica-tion accuracy of 98.53%. In conclusion, the proposed MNPS-MEFL approach improves the accuracy of lung cancer prognosis and has the potential to benefit lung disorder patients. Further research could investigate the application of this approach to other medical imaging tasks and evaluate its performance in real-world clinical settings.