English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Transfer learning is widely used for training machine learning models. Here,we study the role of transfer learning for training fully convolutionalnetworks (FCNs) for medical image segmentation. Our experiments show thatalthough transfer learning reduces the training time on the target task, theimprovement in segmentation accuracy is highly task/data-dependent. Largerimprovements in accuracy are observed when the segmentation task is morechallenging and the target training data is smaller. We observe thatconvolutional filters of an FCN change little during training for medical imagesegmentation, and still look random at convergence. We further show that quiteaccurate FCNs can be built by freezing the encoder section of the network atrandom values and only training the decoder section. At least for medical imagesegmentation, this finding challenges the common belief that the encodersection needs to learn data/task-specific representations. We examine theevolution of FCN representations to gain a better insight into the effects oftransfer learning on the training dynamics. Our analysis shows that althoughFCNs trained via transfer learning learn different representations than FCNstrained with random initialization, the variability among FCNs trained viatransfer learning can be as high as that among FCNs trained with randominitialization. Moreover, feature reuse is not restricted to the early encoderlayers; rather, it can be more significant in deeper layers. These findingsoffer new insights and suggest alternative ways of training FCNs for medicalimage segmentation.

Critical Assessment of Transfer Learning for Medical Image Segmentation with Fully Convolutional Neural Networks