Neural-signature methods for structured EHR prediction

BMC Medical Informatics and Decision Making

Research

Paidi Creed¹  Spiros Denaxas²  Andre Vauvelle² 

[1] BenevolentAI, 4-8 Maple St, London, UK;Institute of Health Informatics, University College London, 222 Euston Road, London, UK;

关键词: Machine learning;    Electronic healthcare records;    Signature methods;   

DOI  :  10.1186/s12911-022-02055-6

 received in 2022-01-24, accepted in 2022-11-17,  发布年份 2022

来源: Springer

【 摘 要 】

Models that can effectively represent structured Electronic Healthcare Records (EHR) are central to an increasing range of applications in healthcare. Due to the sequential nature of health data, Recurrent Neural Networks have emerged as the dominant component within state-of-the-art architectures. The signature transform represents an alternative modelling paradigm for sequential data. This transform provides a non-learnt approach to creating a fixed vector representation of temporal features and has shown strong performances across an increasing number of domains, including medical data. However, the signature method has not yet been applied to structured EHR data. To this end, we follow recent work that enables the signature to be used as a differentiable layer within a neural architecture enabling application in high dimensional domains where calculation would have previously been intractable. Using a heart failure prediction task as an exemplar, we provide an empirical evaluation of different variations of the signature method and compare against state-of-the-art baselines. This first application of neural-signature methods in real-world healthcare data shows a competitive performance when compared to strong baselines and thus warrants further investigation within the health domain.

【 授权许可】

CC BY   
© The Author(s) 2022

【 图 表 】

Fig. 2

12864_2022_9026_Article_IEq246.gif

Fig. 1

Fig. 6

42004_2022_782_Tabh_HTML.png

Fig. 5

Fig. 2

12902_2022_1242_Article_IEq1.gif

Fig. 4

Fig. 3

Fig. 2

Fig. 4

Fig. 5

Fig. 3

12888_2022_4137_Article_IEq2.gif

12888_2022_4137_Article_IEq1.gif

Fig. 6

Fig. 2

Fig. 2

12864_2022_9026_Article_IEq225.gif

Fig. 2

Fig. 1

Fig. 1

Fig. 7

13068_2022_2229_Article_IEq4.gif

13068_2022_2229_Article_IEq3.gif

13068_2022_2229_Article_IEq2.gif

Fig. 1

Fig. 2

Fig. 4

Fig. 3

12951_2022_1749_Article_IEq8.gif

Fig. 2

12951_2022_1749_Article_IEq6.gif

Fig. 2

Fig. 5

Fig. 5

Fig. 4

Fig. 6

Fig. 1

Fig. 1

Fig. 5

12951_2022_1749_Article_IEq3.gif

Fig. 4

Fig. 3

Fig. 3

12864_2022_9026_Article_IEq187.gif

Fig. 1

Fig. 1

Fig. 2

Fig. 7

Fig. 1

12951_2022_1749_Article_IEq1.gif

Fig. 2

12888_2022_4322_Article_IEq11.gif

12888_2022_4322_Article_IEq10.gif

Fig. 3

Fig. 3

Fig. 6

Fig. 1

Fig. 1