Date of Degree
Artificial Intelligence and Robotics | Radiology
Medical Imaging, Deep Learning, Cancer detection and diagnosis, Low dose CT
Deep Learning Methods have shown its great effort in the area of Computer Vision. However, when solving the problems of medical imaging, deep learning’s power is confined by limited data available. We present a series of novel methodologies for solving medical imaging analysis problems with limited Computed tomography (CT) scans available. Our method, based on deep learning, with different strategies, including using Generative Adversar- ial Networks, two-stage training, infusing the expert knowledge, voting based or converting to other space, solves the data set limitation issue for the cur- rent medical imaging problems, specifically cancer detection and diagnosis, and shows very good performance and outperforms the state-of-art results in the literature. With the self-learned features, deep learning based techniques start to be applied to the biomedical imaging problems and various structures have been designed. In spite of its simplity and anticipated good performance,
the deep learning based techniques can not perform to its best extent due to the limited size of data sets for the medical imaging problems. On the other side, the traditional hand-engineered features based methods have been studied in the past decades and a lot of useful features have been found by these research for the task of detecting and diagnosing the pulmonary nod- ules on CT scans, but these methods are usually performed through a series of complicated procedures with manually empirical parameter adjustments. Our method significantly reduces the complications of the traditional proce- dures for pulmonary nodules detection, while retaining and even outperforming the state-of-art accuracy. Besides, we make contribution on how to convert low-dose CT image to full-dose CT so as to adapting current models on the newly-emerged low-dose CT data.
Tan, Jiaxing, "Deep Learning Based Medical Image Analysis with Limited Data" (2019). CUNY Academic Works.