Sepsis Detection in Sparse Clinical Data Using Long Short-Term Memory Network with Dice Loss

conference paper

English

This paper aims to present a methodology for sepsis prediction from clinical time-series data. Sepsis is one of the most threatening states which could occur while treating a patient at the intensive care unit. Therefore its prediction could significantly improve the quality of the patient treatment. In this work, we address the problem of sepsis predictionwith Long Short-Term Memory (LSTM) network with spe-cialized deep architecture with residual connections. The output of the network is sepsis prediction score at eachpoint in time. Feature normalization into the fixed range of values isapplied including replacing missing values with numericalrepresentation from outside the normalized range. Therefore, the LSTM network is able to include missing values inthe learning process. Also, the rarity of sepsis occurrence in the provided dataset is a challenging problem. This problem is addressed by the application of dice loss providing automatically weighted classes by the occurrence of the feature. The proposed method leads to 0.372 normalized utility score as the best official PhysioNet/Computing in Cardiology (CinC) Challenge 2019 entry of ECGuru10 team.

sepsis, detection, intensive care unit, long short-term memory network, dice loss, computing in cardiology, physionet challenge

VIČAR, T.; NOVOTNÁ, P.; HEJČ, J.; RONZHINA, M.; SMÍŠEK, R.

30. 9. 2019

Computing in Cardiology 2019

Singapore

0276-6574

Computers in Cardiology

United States of America

1

4

4