Quantifying Reproducibility of Common Laboratory Tests in a Tertiary Care Hospital for More Realistic Synthetic Data in Machine Learning
Year: 2026
Type: preprint
Abstract: Data from medical records are often limited in size, creating challenges for Artificial Intelligence training, and an opportunity for data augmentation techniques to create synthetic data. A natural question is how much data can be randomized and still be realistic. The main goal of this study is to attempt to measure the reproducibility of labs drawn in practice, so these ranges can be used for generating more realistic synthetic data. A second goal is to inform clinicians of the real error ranges of their laboratory data so these values may be better interpreted.
Source: medRxiv
Author: Dawson Dean
Institution: University of Kentucky
Language: English
Subfield: Medical Laboratory Technology
Field: Health Professions
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7223372252