Semi-supervised information fusion for medical image analysis: Recent progress and future perspectives
Ying Weng, Yiming Zhang, Wenxin Wang, Tom Dening
Abstract
Supervised machine learning requires training on the dataset with annotation. However, fine-grained annotation is very expensive to acquire. In the medical image analysis domain, the sheer volume of data and lack of annotation limit the performance of the model. To address these limitations, semi-supervised information fusion has recently emerged as an important and promising paradigm owing to its ability to exploit labelled and unlabelled data and combine information from multiple sources to obtain a more robust and accurate performance. In this survey, we review the recent progress of semi-supervised information fusion for medical image analysis. Moreover, we categorize the state-of-the-art information fusion applications of semi-supervised learning with in-depth analysis. Finally, we discuss the challenges and outline the future perspective.