Ranking Pathology Data in the Absence of a Ground Truth



Qi, Jing ORCID: 0000-0003-2476-7620, Burnside, Girvan ORCID: 0000-0001-7398-1346 and Coenen, Frans ORCID: 0000-0003-1026-6649
(2021) Ranking Pathology Data in the Absence of a Ground Truth. .

[img] Text
jingQi_SGAI2021.pdf - Author Accepted Manuscript

Download (336kB) | Preview

Abstract

Pathology results play a critical role in medical decision making. A particular challenge is the large number of pathology results that doctors are presented with on a daily basis. Some form of pathology result prioritisation is therefore a necessity. However, there is no readily available training data that would support a traditional supervised learning approach. Thus some alternative solutions are needed. There are two approaches presented in this paper, anomaly-based unsupervised pathology prioritisation and proxy ground truth-based supervised pathology prioritisation. Two variations of each were considered. With respect to the first, point and time series based unsupervised anomaly prioritisation; and with respect to the second kNN and RNN proxy ground truth-based supervised prioritisation. To act as a focus, Urea and Electrolytes pathology testing was used. The reported evaluation indicated that the RNN proxy ground truth-based supervised pathology prioritisation method produced the best results.

Item Type: Conference or Workshop Item (Unspecified)
Uncontrolled Keywords: Data ranking, Time series, Deep learning, Pathology data
Divisions: Faculty of Health and Life Sciences
Faculty of Health and Life Sciences > Institute of Population Health
Faculty of Science and Engineering > School of Electrical Engineering, Electronics and Computer Science
Depositing User: Symplectic Admin
Date Deposited: 12 Oct 2021 10:48
Last Modified: 18 Jan 2023 21:27
DOI: 10.1007/978-3-030-91100-3_18
Related URLs:
URI: https://livrepository.liverpool.ac.uk/id/eprint/3140163