Automated Diagnosis of Lung Cancer with the Use of Deep Convolutional Neural Networks on Chest CT

For the past several decades, machine learning has greatly enhanced the process of image analysis. With the development of deep learning technologies in the 21st century, image recognition has gained applicability to various technologies such as automated driving system, face recognition and medical data processing. This research attempts to diagnose lung cancer using chest CT of patients and non-patients. A modified form of Deep Convolutional Neural Network is introduced, which involves using multiple 2D convolutional neural networks on individual slices and combining the results to diagnose patients and non-patients. Each patient/non-patient's chest CT data were first segmented for the lung features and stored into three-dimensional arrays. The preprocessed three-dimensional arrays were fed into the CNN framework, and the parameters of the network were trained. Many iterations of the process with enough data modified network parameters in a way that the network was able to diagnose CT scans of patients with accuracy between 70~80%.