eduzhai > Applied Sciences > Engineering >

Adversarial Multi-Source Transfer Learning in Healthcare Application to Glucose Prediction for Diabetic People

  • Save

... pages left unread,continue reading

Document pages: 12 pages

Abstract: Deep learning has yet to revolutionize general practices in healthcare,despite promising results for some specific tasks. This is partly due to databeing in insufficient quantities hurting the training of the models. To addressthis issue, data from multiple health actors or patients could be combined bycapitalizing on their heterogeneity through the use of transfer learning.To improve the quality of the transfer between multiple sources of data, wepropose a multi-source adversarial transfer learning framework that enables thelearning of a feature representation that is similar across the sources, andthus more general and more easily transferable. We apply this idea to glucoseforecasting for diabetic people using a fully convolutional neural network. Theevaluation is done by exploring various transfer scenarios with three datasetscharacterized by their high inter and intra variability.While transferring knowledge is beneficial in general, we show that thestatistical and clinical accuracies can be further improved by using of theadversarial training methodology, surpassing the current state-of-the-artresults. In particular, it shines when using data from different datasets, orwhen there is too little data in an intra-dataset situation. To understand thebehavior of the models, we analyze the learnt feature representations andpropose a new metric in this regard. Contrary to a standard transfer, theadversarial transfer does not discriminate the patients and datasets, helpingthe learning of a more general feature representation.The adversarial training framework improves the learning of a general featurerepresentation in a multi-source environment, enhancing the knowledge transferto an unseen target.The proposed method can help improve the efficiency of data shared bydifferent health actors in the training of deep models.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...