Lung Cancer Detection on CT Scan Images with Deep Learning Methods: Sugeno Fuzzy Integral-Based CNN Ensemble Method

Lung cancer causes the highest mortality rate among all cancers. Existing methods for detecting Lung cancer are not optimal, as they rely mostly on a single model or multiple models employing probability averaging. This study is proposed to differentiate malignant, benign and normal cells of lung by employing the Sugeno Ensemble method in conjunction with four Convolutional Neural Networks (CNN): VGG11, SqueezeNet, GoogLeNet, and Wide ResNet-50-2 to improve accuracy of Lung cancer detection. Each model will be individually trained with the dataset before Sugeno ensemble is performed to produce the final accuracy. This study evaluates the accuracy of applying Sugeno fuzzy logic techniques to CT scan images of Lung cancer to determine if it outperforms existing methods and investigates the limitations of earlier research. This study examines the image of cancer morphologies from various patients. The confusion matrix is then utilized to evaluate the performance of the framework. The ultimate achieved accuracy rate is 98.47%. This study found that integrating the Sugeno ensemble technique with the four CNN models increases Lung cancer detection performance. These discoveries could aid in the development of systems to identify Lung cancer patients and provide implications for artificial intelligence and the medical industry.