Predicting Outcomes for Cancer Patients with Transformer-Based Multi-task Learning

Gerrard, L; Peng, X; Clarke, A; Schlegel, C; Jiang, J

Predicting Outcomes for Cancer Patients with Transformer-Based Multi-task Learning

Gerrard, L Peng, X

Clarke, A Schlegel, C Jiang, J

Permalink

Publisher:: Springer International Publishing
Publication Type:: Conference Proceeding
Citation:: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2022, 13151 LNAI, pp. 381-392
Issue Date:: 2022-01-01

In Progress

	Filename	Description	Size
	AJCAI21_Transformer_based_multi_task_for_cancer.pdf	Accepted version	394.28 kB	Adobe PDF	View/Open

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is being processed and is not currently available.

Full metadata record

Field	Value	Language
dc.contributor.author	Gerrard, L
dc.contributor.author	Peng, X https://orcid.org/0000-0002-8901-1472
dc.contributor.author	Clarke, A
dc.contributor.author	Schlegel, C
dc.contributor.author	Jiang, J
dc.date.accessioned	2022-05-26T06:05:53Z
dc.date.available	2022-05-26T06:05:53Z
dc.date.issued	2022-01-01
dc.identifier.citation	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2022, 13151 LNAI, pp. 381-392
dc.identifier.isbn	9783030975456
dc.identifier.issn	0302-9743
dc.identifier.issn	1611-3349
dc.identifier.uri	http://hdl.handle.net/10453/157722
dc.description.abstract	Cancer patients often experience numerous hospital admissions as a result of their cancer and treatment, which can negatively impact treatment progress and quality of life. Accurately predicting outcomes for cancer patients is therefore crucial in providing personalised care and improving patient outcomes. Existing models leveraging deep learning with Electronic Health Record (EHR) data to predict outcomes for cancer patients are limited, despite the demonstrated success of these approaches with cancer imaging data and non-cancer EHR applications. Additionally, current methods focus on single-task predictions, and increasing evidence suggests jointly training a model on two related tasks can improve predictive performance. To address these limitations, we propose a Transformer-based Multi-Task (TransMT) model that captures relationships between diagnosis codes and sequential hospital visits to simultaneously predict related outcomes for hospitalised cancer patients. Experiments conducted on two public datasets show the proposed model outperforms both single-task and recurrent neural network approaches in predicting future diagnosis and hospital readmission, and demonstrates the benefits of using deep learning with EHR data for cancer-related research.
dc.language	en
dc.publisher	Springer International Publishing
dc.relation	Commonwealth Department of Health
dc.relation.ispartof	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
dc.relation.isbasedon	10.1007/978-3-030-97546-3_31
dc.rights	info:eu-repo/semantics/restrictedAccess
dc.rights	This is a post-peer-review, pre-copyedit version of an article published in [Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2022, 13151 LNAI, pp. 381-392]. Publication date Published: 01 Jan 2022] The final authenticated version is available online at: [ https://link.springer.com/chapter/10.1007/978-3-030-97546-3_31]”
dc.subject.classification	Artificial Intelligence & Image Processing
dc.title	Predicting Outcomes for Cancer Patients with Transformer-Based Multi-task Learning
dc.type	Conference Proceeding
utslib.citation.volume	13151 LNAI
pubs.organisational-group	/University of Technology Sydney
pubs.organisational-group	/University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	/University of Technology Sydney/Strength - AAII - Australian Artificial Intelligence Institute
utslib.copyright.status	in_progress	*
dc.date.updated	2022-05-26T06:05:52Z
pubs.publication-status	Published
pubs.volume	13151 LNAI

Abstract:

Cancer patients often experience numerous hospital admissions as a result of their cancer and treatment, which can negatively impact treatment progress and quality of life. Accurately predicting outcomes for cancer patients is therefore crucial in providing personalised care and improving patient outcomes. Existing models leveraging deep learning with Electronic Health Record (EHR) data to predict outcomes for cancer patients are limited, despite the demonstrated success of these approaches with cancer imaging data and non-cancer EHR applications. Additionally, current methods focus on single-task predictions, and increasing evidence suggests jointly training a model on two related tasks can improve predictive performance. To address these limitations, we propose a Transformer-based Multi-Task (TransMT) model that captures relationships between diagnosis codes and sequential hospital visits to simultaneously predict related outcomes for hospitalised cancer patients. Experiments conducted on two public datasets show the proposed model outperforms both single-task and recurrent neural network approaches in predicting future diagnosis and hospital readmission, and demonstrates the benefits of using deep learning with EHR data for cancer-related research.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/157722