Detection of lung cancer with deep learning using computed tomography images

With the developing technology, the concept of artificial intelligence is an important part of our lives in terms of discourse and technique. By using machine learning, which is a type of artificial intelligence, and its subset, deep learning models and algorithms, existing data can be processed and new and useful data are obtained. Image processing technology, one of the deep learning methods, is a model of presenting new and useful information by processing existing images. Medical imaging technology, on the other hand, increases the rate of early diagnosis of diseases and response to treatment. In addition, the process of reading the medical images of patients by physicians is a matter of discussion due to the fact that the results are subjective as well as a waste of time. Image processing technology is increasing its popularity in this field as it reduces the labor and time loss of radiologists and produces objective results. In this thesis, "What is image processing technology?" the answer to the question has been sought. The contribution of the thesis, which is an answer to this question, to the world of science is emphasized. Within the scope of the thesis, the dataset of computerized tomography images of healthy, lung cancerous malignant and benign tumors, pre-trained convolutional neural networks; AlexNet, DenseNet 201, GoogleNet, MobileNetV2, ResNet50 trained with their architecture. It is aimed to increase the accuracy rate found after this training. In this context, a new hybrid model has been developed. Experimental results prove the success of the proposed model.