Transforming view of medical images using deep learning

Since the last decade, there is a significant change in the procedure of medical diagnosis and treatment. Specifically, when internal tissues, organs such as heart, lungs, brain, kidneys and bones are the target regions, a doctor recommends ‘computerized tomography’ scan and/or magnetic resonance imaging to get a clear picture of the damaged portion of an organ or a bone. This is important for correct examination of the medical deformities such as bone fracture, arthritis, and brain tumor. It ensures prescription of the best possible treatment. But ‘computerized tomography’ scan exposes a patient to high ionizing radiation. These rays make a person more prone to cancer. Magnetic resonance imaging requires a strong magnetic field. Thus, it becomes impractical for patients with implants in their body. Moreover, the high cost makes the above-stated techniques unaffordable for low economy class of society. The above-mentioned challenges of ‘computerized tomography’ scan and magnetic resonance imaging motivate researchers to focus on developing a technique for conversion of 2-dimensional view of medical images into their corresponding multiple views. In this manuscript, the authors design and develop a deep learning model that makes an effective use of conditional generative adversarial network, an extension of generative adversarial network for the transformation of 2-dimensional views of human bone into the corresponding multiple views at different angles. The model will prove useful for both doctors and patients.