Lung Cancer Detection System Using Medical Image Processing, Machine Learning, and Deep Learning Approaches رسالة ماجستير

Abstract

Early detection of lung cancer stands as a pivotal turning point in the realm of medical care, where the timely identification of this condition empowers healthcare professionals to prescribe the most effective treatments, ultimately leading to a reduction in mortality rates and the preservation of precious human lives. Countless dedicated researchers have explored this subject, employing a multifaceted approach to diagnosing lung cancer. In response to the formidable challenge of selecting optimal methodologies to ensure consistently high performance, this thesis introduces two distinctive models, one based on machine learning (ML) and the other on deep learning (DL). Both models are designed to facilitate the diagnosis of lung cancer, employing clinical data and medical images. In the domain of ML, a multitude of machine learning models, including eXtreme Gradient Boosting (XGB), Light Gradient Boosting (LGB), Random Forest (RF), Support Vector Machine (SVM), Decision Tree (DT), Gradient Boosting Machine (GBM), and Adaptive Boosting (AdaBoost), have been used. This model culminated in a remarkable achievement, with the XGB and Adaboost models achieving an accuracy of 99.07%. Conversely, the DL was trained on CT scan images categorized under three distinctive labels: Normal, Benign, and Malignant. These images underwent meticulous preprocessing and training across eight deep learning algorithms, namely Convolutional Neural Network (CNN), MobileNet, Xception, DenseNet121, Visual Geometry Group 16 (VGG16), Visual Geometry Group 19 (VGG19), ResNet50, and EfficientNetB0. The best performance, which was achieved by the custom CNN model, boasts an extraordinary classification rate of 99.70% and perfect precision, recall, and f1-score.