Stacked Denoising Autoencoders for mortality risk prediction using imbalanced clinical data

Alhassan, Z; McGough, AS; Alshammari, R; Daghstani, T; Budgen, D; Al-Moubayed, N

doi:10.1109/ICMLA.2018.00087

Stacked Denoising Autoencoders for mortality risk prediction using imbalanced clinical data

Lookup NU author(s): Dr Stephen McGough

Downloads

Accepted version [.pdf]

Licence

This is the authors' accepted manuscript of a conference proceedings (inc. abstract) that has been published in its final definitive form by IEEE, 2018.

For re-use rights please refer to the publisher's terms and conditions.

Abstract

Clinical data, such as evaluations, treatments, vital sign and lab test results, are usually observed and recorded at hospital systems. Making use of such data to help physicians evaluating the mortality risk of in-hospital patients provides an invaluable source of information which ultimately help improving the health-care services. Therefore, quick and accurate prediction of mortality can be critical for physicians to make intervention decisions. In this work, we introduce a predictive Deep Learning model aiming to evaluate the mortality risk of in-hospital patients. Stacked Desoising Autoencoder (SDA) is trained using a unique time-stamped dataset (King Abdullah International Research Center - KAIMRC) which is naturally imbalanced. The work is compared to common deep learning approaches using different methods for data balancing. The proposed model demonstrated here to overcome the problem of imbalanced data and outperform common deep learning approaches with an accuracy of 77.13% for the Recall macro.