Linxiao Xia

Mr. Linxiao Xia

orcid

Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
19 November 2026 in Bioengineering
Verified
Journal Article
Open access
PDF file
ZZZiqian Zeng
Ziqian Zeng
LXLinxiao Xia
Linxiao Xia
YLYalei Liu
Yalei Liu
JWJianjun Wu
Jianjun Wu
QCQi Chen
Qi Chen
ZTZhifeng Tang
Zhifeng Tang
Wideband laser speckle rheological microscopy (WBSHEAR) enables frequency-resolved characterization of blood viscoelasticity, yet its potential for quantitative blood-composition sensing remains largely unexplored. This study investigated whether different representations of laser-speckle dynamics could support task-specific prediction of blood constituents through machine learning. Four signal representations, including temporal autocorrelation (G2), local scaling exponent (Alpha), viscoelastic spectrum (Spectrum), and manually extracted spectral descriptors (Manual), were evaluated individually and in combination. Using controlled porcine blood suspensions, 378 task-specific measurement records were collected across four compositional perturbations: red blood cell (RBC) fraction, Intralipid, fibrinogen, and albumin. Seven representation pathways and six regression models were systematically compared using acquisition-provenance-grouped nested cross-validation. Record-level out-of-fold predictions achieved R2 values of 0.802, 0.956, 0.949, and 0.936 for RBC fraction, Intralipid, fibrinogen, and albumin, respectively. The most frequently selected representations differed across tasks: Alpha + Spectrum for RBC fraction (75.0% of outer folds), Spectrum for Intralipid (55.6%), and G2 for fibrinogen (91.7%) and albumin (92.3%). These findings demonstrate that different representations of the same laser-speckle measurements retain distinct predictive information associated with blood-composition changes. Task-adaptive representation selection therefore provides a promising computational framework for extending WBSHEAR beyond viscoelastic characterization toward quantitative blood-composition sensing.