Semi-supervised Learning in Computer-aided Diagnosis

Computer-aided diagnosis techniques have significant potential in assisting pathologists in diagnosis, especially in the field of medical image processing. Many supervised learning-based approaches have been successfully used on computerised tomography, ultrasound, or magnetic resonance imaging images recently. Meanwhile, since some pathology disciplines cannot pro-vide the amount of labelled data required by these conventional methods for training, semi-supervised learning (SSL) methods have recently attracted attention. This paper provides the basic introduction of SSL and computer-aided-diagnosis (CAD) and reviews the current experiments and prospects of the application of SSL in CAD systems, including data acquisition, image pre-processing, feature extraction, classification and validation, etc. This paper also reports and highlights the strategy and performance of SSL combined with CAD according to some findings of researchers to date and provides some approaches for model validation.