Adaptive Transfer Learning to Enhance Domain Transfer in Brain Tumor Segmentation
Yuan Liqiang, Marius Erdt, Wang Lipo
The development in the field of deep learning greatly benefits from the improvements in the network structure and the availability of massive data sets for training. However, in medical imaging, labeled data is often not available due to expensive manual cost. This challenge motivates researchers to explore various methods that can cope with the scarcity of data in the medical domain. TL (Transfer Learning) has demonstrated its potential to improve the training efficiency by transferring knowledge from one machine learning classifier (source domain) to another machine learning classifier (target domain).