Affiliations: [a] Federal Urdu University of Arts Science & Technology, Karachi, Pakistan | [b] National University of Computing and Emerging Sciences, Karachi, Pakistan | [c] Karachi Institute of Economics and Technology, College of Computing and Information Sciences, Karachi, Pakistan | [d] Madni General Hospital, Gujrat City, Pakistan | [e] President Mohammad Ali Jinnah University, Karachi, Pakistan
Correspondence:
[*]
Corresponding author: Uzma Afzal, Department of Computer Science, Federal Urdu University of Arts Science & Technology, Karachi, Pakistan. E-mail:[email protected]
Abstract: BACKGROUND: Configuration of complex Laboratory Developed Tests
(LDTs) is a time-consuming and complicated task, potentially leading to
inconsistent LDTs in which features constraints remain unresolved and
important features could remain unselected.
OBJECTIVE: Our objective is to address these issues by presenting
an automated, health informatics solution which autonomously optimizes
feature selection in complex LDTs through Particle Swarm Optimization
(PSO). The optimization goal is to minimize inconsistencies and configuration
time, and maximize the number of selected features.
METHODS: We implemented our technology in a local, secondary-care
hospital in Pakistan which configures LDT for a local epidemic disease.
First, a list of inconsistent LDT configurations is generated. This is used
to initially estimate optimal PSO parameters, which are then used for
optimization process.
RESULTS: Results show that PSO is able to minimize 91%
inconsistencies between 9 and 11 seconds. The number of selected critical
features also increases by 100% in the optimized LDT configuration.
CONCLUSIONS: We present a novel and the first application of
computational optimization to solve LDT configuration issues.
Keywords: Laboratory developed test, health informatics, inconsistencies, particle swarm optimization
