Fully Automated End to End Analysis of Non-small-cell Lung Carcinoma using Deep Learning Techniques.

Deep learning in medical imaging analysis has revolutionized the field in areas such as computer-aided detection and segmentation of clinical abnormalities. Several studies have been published on lung cancer screening using deep learning methodologies. Specific to lung cancer screening, algorithms have been trained to automatically detect and diagnose lesions in the lungs in low dose computed tomography (CT) by leveraging longitudinal imaging in combination with biopsy results. Perez et.al [3] proposed a three-dimensional (3D) CNN model to detect lung nodules and predict lung cancer using CT images. The lung is extracted from the entire volume in each patient and the extracted data is used to train the model. To increase the precision, both 2D and 3D convolutions were used. They were able to achieve best results using 3D convolutions suggesting there is information between slices that is relevant for lung cancer analysis.

Faculty Supervisor:

Eran Ukwatta

Student:

Jenita Priya Rajamanickam Manokaran

Partner:

Altis Labs Inc.

Discipline:

Engineering

Sector:

Other

University:

University of Guelph