Tuesday, November 17, 2026
Science
No Result
View All Result
  • Login
  • HOME
  • SCIENCE NEWS
  • CONTACT US
  • HOME
  • SCIENCE NEWS
  • CONTACT US
No Result
View All Result
Scienmag
No Result
View All Result
Home Science News

Spatial Mixed Model Slashes Error Rates in Estimating Glaucoma Progression

November 17, 2026
in Science News
Ophelia Keating
By Ophelia Keating Scienmag Editorial Profile - Health Services Research
Reading Time: 7 mins read
0
Spatial Mixed Model Slashes Error Rates in Estimating Glaucoma Progression

Spatial Mixed Model Slashes Error Rates in Estimating Glaucoma Progression

65
SHARES
587
VIEWS
Share on FacebookShare on Twitter
ADVERTISEMENT

A new spatial two-fold linear mixed model evaluated on longitudinal visual field records from the Bascom Palmer Glaucoma Repository substantially improves the precision of estimated glaucoma progression rates, achieving lower mean squared error and reduced classification error rates compared to conventional pointwise linear regression and permutation approaches. Developed by researchers Chen Zhao, Swarup S. Swaminathan, and J. Sunil Rao, the statistical framework incorporates both eye-level and within-eye cluster-level random effects while accounting for spatial dependence among perimetric test locations. In retrospective analyses and simulation studies, the method produced mean squared error values for slope estimates of 0.41 in the lower hemifield and 0.50 in the upper hemifield, compared with 1.12 and 1.72 under ordinary least squares. In simulated testing scenarios, the model achieved estimated false-positive and false-negative rates of zero, outperforming standard pointwise regression, which exhibited a false-positive rate of 0.078, and permutation-based pointwise regression, which yielded a false-positive rate of 0.050.

Visual field testing, or perimetry, serves as a fundamental diagnostic tool in ophthalmology to measure peripheral vision loss associated with progressive optic neuropathies such as glaucoma. Glaucoma involves the chronic, progressive loss of retinal ganglion cells, producing subsequent functional vision defects that clinicians identify and monitor through perimetric testing. Standard automated perimetry presents patients with a series of light stimuli of varied brightness at distinct locations, which patients identify by pressing an indicator button. The testing system then calculates pointwise sensitivities and global summary metrics. Beyond raw sensitivity measurements, the Humphrey perimetry system provides total deviation values at each test location, representing sensitivity deviations adjusted for patient age against a normative reference database. These standardized adjustments eliminate the confounding effects of aging across longitudinal testing sequences, establishing a stable foundation for monitoring disease progression over extended clinical follow-up intervals.

Interpreting visual field reports remains clinically complex, prompting the development of diverse summary indices and statistical tools. Among these, the Glaucoma Hemifield Test specifically identifies visual field loss patterns characteristic of glaucomatous neuropathy based on the biological principle that glaucomatous damage typically respects the horizontal midline of the visual field. The test compares the sensitivity of corresponding points in the upper and lower halves of the visual field, where statistically significant disparities serve as strong indicators of glaucomatous injury. Traditionally, focal disease progression has been identified when three or more contiguous test locations show statistically significant sensitivity loss over time, reflecting the clinical understanding that glaucomatous damage tends to occur in localized functional regions rather than at isolated testing points. However, this established criterion produces a binary classification indicating whether focal progression occurred, offering no quantitative rate estimates and obscuring meaningful variations during early or slowly evolving disease stages.

To overcome the substantial noise and arbitrary classification thresholds inherent in unconstrained pointwise linear regression, the investigators structured their two-fold mixed-effects model around the anatomical clustering of the Glaucoma Hemifield Test. This clustering partitions visual field test locations into five distinct functional clusters within each hemifield, designated clusters 1A through 5A for the upper hemifield and 1B through 5B for the lower hemifield, maintaining identical groupings across all eyes. In the formal mathematical specification, measured total deviation values within each cluster for a given eye at a specific follow-up time are modeled through a combination of fixed population effects and nested hierarchical random effects. The fixed parameters comprise an overall population baseline intercept representing the mean initial total deviation and a fixed population slope capturing the overall rate of sensitivity change across time.

Hierarchical random effects operate across multiple structural tiers to capture biological heterogeneity between patients and across distinct regions within each eye. At the eye level, a normally distributed random intercept accounts for patient-specific baseline deviation from the population mean, while an eye-level random slope accounts for individual departures from the population rate of disease change. At the intra-eye cluster level, additional normally distributed random effects characterize localized anatomical heterogeneity. A cluster-specific random intercept captures local deviations from the eye-specific baseline, whereas a cluster-specific random slope models localized differences in progression rates across the five hemifield clusters. By simultaneously modeling mixed effects across individuals and Glaucoma Hemifield Test clusters, the framework systematically borrows statistical strength across fellow eyes and neighboring functional sectors, shrinking noisy individual estimates toward informed group averages while preserving sensitivity to focal disease trajectories.

In addition to hierarchical cluster effects, the model explicitly incorporates a vector of spatial random effects that captures spatial autocorrelation between all test locations across an eye based on their physical visual field coordinates. The spatial covariance between any two coordinate locations is formulated through a continuous Matérn correlation function governed by Euclidean spatial distance. This Matérn covariance structure includes a marginal variance parameter, a specified smoothness parameter, a spatial scale parameter, and a modified Bessel function of the second kind. The spatial random effects are specified to be statistically independent of both the eye-level and cluster-level random intercepts and slopes. A residual error vector accounts for independent unexplained variability across observations within each cluster. The authors highlighted that practitioners fitting nested spatial models should monitor potential boundary estimates under restricted maximum likelihood, where variance components approach zero, and address potential parameter instability between the Matérn smoothness and scale parameters by fixing smoothness at conventional values or applying penalized likelihood methods.

Parameter estimation within the compact linear mixed model matrix formulation is executed using restricted maximum likelihood. The overall variance-covariance structure integrates the random effect design matrices, the block-diagonal covariance matrix of all random effects, and the residual error covariance. Restricted maximum likelihood estimation optimizes the objective function with respect to the complete variance components vector, encompassing eye-level and cluster-level intercept and slope variances and covariances, spatial marginal variance, spatial smoothness and scale parameters, and residual error variance. Once optimal variance components are established, population fixed-effect coefficients are estimated via generalized least squares equations, while best linear unbiased predictions are derived for all eye-level, cluster-level, and spatial random effects.

To quantify total mean squared error for combined cluster-specific slopes and intercepts, the researchers implemented a parametric bootstrap procedure. The true cluster-specific intercept combines the population fixed intercept, the eye-level random intercept, and the cluster-level random intercept, while the true cluster-specific slope combines the population fixed slope, the eye-level random slope, and the cluster-level random slope. In each bootstrap replicate, eye-level and cluster-level random effects are drawn from bivariate normal distributions parameterized by restricted maximum likelihood estimates, and spatial random effects are simulated from a multivariate normal distribution based on the estimated Matérn parameters. Adding simulated errors to fixed and random components generates synthetic responses. Refitting the model across replicates yields the empirical mean squared error as the average squared difference between estimated and true trajectory parameters.

The bootstrap algorithm was additionally configured to construct ninety-five percent confidence intervals for cluster-specific intercepts and slopes using a normal approximation. For each cluster within an eye, the bootstrap standard error is derived as the standard deviation of parameter estimates across all bootstrap replicates. Confidence interval bounds are calculated by adding and subtracting 1.96 times the bootstrap standard error from the point estimate. Individual eyes whose upper confidence bound for the cluster progression slope falls entirely below zero are identified as statistically significant fast progressors. This objective criterion provides a reliable threshold for distinguishing confirmed localized deterioration from random measurement noise during longitudinal perimetric monitoring.

Empirical evaluation utilized clinical data from the Bascom Palmer Glaucoma Repository at the University of Miami, comprising longitudinal testing records from over seventy thousand subjects evaluated between November 1996 and February 2022. The analytical sample included eyes with open-angle glaucoma under International Classification of Diseases code H40.1X. Only reliable perimetric examinations, defined by false-positive rates below fifteen percent and Glaucoma Hemifield Test results other than abnormally high sensitivity, were analyzed. Initial examinations with mean deviation values five decibels or more worse than subsequent tests were excluded to remove artifactual patient learning effects. Similarly, visual fields showing temporary drops and subsequent recoveries of five decibels or more were excluded as poor-quality tests. The final sample was restricted to eyes with at least five reliable examinations conducted over at least two years of follow-up.

To prevent lower-bound sensitivity compression from distorting progression estimates, the researchers developed an objective classification and adjustment framework for the perimetric floor effect. This floor effect arises when sensitivity reaches measurement lower limits, causing subsequent biological damage to present as an artificially flattened slope. Test locations were flagged at the floor if initial sensitivity measured ten decibels or less and subsequent values remained within three decibels of that baseline. Clusters with three to four points were labeled floor clusters if more than two points were flagged, while clusters with five to six points required more than three flagged points. Eye-level severity was categorized as non-floor for under two affected clusters, mild for two to three, moderate for four to five, and severe for six or more. Integrating a floor-specific random effect ensured that baseline variance driven by lower-bound compression did not bias estimated progression rates.

Comparative evaluations demonstrated that the two-fold linear mixed model produced point estimates that were regularly shrunk toward zero and accompanied by substantially shorter ninety-five percent confidence intervals compared with ordinary least squares, reflecting variance reduction achieved by borrowing strength across eyes and clusters. Estimated variance components indicated that floor-level intercept variance was the largest contributor to baseline variability, measuring 40.95 in the upper hemifield and 28.55 in the lower hemifield, while eye and floor effects accounted for most variation in progression rates. Two-dimensional contour plots of predicted random effects reflected glaucomatous biology, showing pronounced peripheral sensitivity loss and characteristic finger-like projections extending toward central regions in fast progressors. In five-hundred-sample simulation experiments, the mixed model achieved false-positive and false-negative rates of zero across all progression scenarios, outperforming pointwise linear regression, which showed a false-positive rate of 0.078, and permutation-based regression, which exhibited a false-positive rate of 0.050. The authors noted that future research should validate these findings in prospective cohorts and examine extensions incorporating non-linear trajectories and clinical covariates such as intraocular pressure fluctuations.

Subject of Research: Statistical modeling of longitudinal visual field progression in glaucoma

Article Title: Improved Estimation of Rates of Glaucoma Progression Using Spatial Two-Fold Linear Mixed Models

Article References: Zhao, C., Swaminathan, S. S., & Rao, J. S. (2026). Improved Estimation of Rates of Glaucoma Progression Using Spatial Two-Fold Linear Mixed Models. Journal of Statistical Theory and Applications, 25(1), Article 36. https://doi.org/10.1007/s44199-026-00184-9

Image Credits: AI Generated

DOI: 10.1007/s44199-026-00184-9

Keywords: glaucoma progression, visual field testing, standard automated perimetry, linear mixed models, spatial statistics, Glaucoma Hemifield Test, total deviation, Matérn correlation, restricted maximum likelihood, parametric bootstrap, Bascom Palmer Glaucoma Repository, open-angle glaucoma, pointwise linear regression, permutation of pointwise linear regression, perimetric floor effect, mean squared error, retinal ganglion cells

Cite Scienmag News

Ophelia Keating. (November 17, 2026). Spatial Mixed Model Slashes Error Rates in Estimating Glaucoma Progression. Scienmag. https://scienmag.com/spatial-mixed-model-slashes-error-rates-in-estimating-glaucoma-progression/

Ophelia Keating. "Spatial Mixed Model Slashes Error Rates in Estimating Glaucoma Progression." Scienmag, 17 November 2026, https://scienmag.com/spatial-mixed-model-slashes-error-rates-in-estimating-glaucoma-progression/. Accessed 17 November 2026.

Ophelia Keating. "Spatial Mixed Model Slashes Error Rates in Estimating Glaucoma Progression." Scienmag. November 17, 2026. https://scienmag.com/spatial-mixed-model-slashes-error-rates-in-estimating-glaucoma-progression/

Tags: Bascom Palmer Glaucoma RepositoryGlaucoma Hemifield Testglaucoma progressionlinear mixed modelsMatérn correlationmean squared erroropen-angle glaucomaparametric bootstrapperimetric floor effectpermutation of pointwise linear regressionpointwise linear regressionrestricted maximum likelihoodretinal ganglion cellsspatial statisticsstandard automated perimetrytotal deviationvisual field testing
Share26Tweet16
Previous Post

Mathematical Model Explains Why Atmospheric Pekeris Waves Attenuate Rapidly After Eruptions

Next Post

Boundary Element Framework Cuts Computation Time in Nonlinear Thermo-Viscoplastic Solids

Related Posts

Combined Cataract and Microhook Glaucoma Surgery Proves Safe in Older Adults
Medicine

Combined Cataract and Microhook Glaucoma Surgery Proves Safe in Older Adults

November 17, 2026
Mature Stone’s sheep reduce home range extent during hunting periods
Biology

Mature Stone’s sheep reduce home range extent during hunting periods

November 17, 2026
Review Finds No Fatigue Scale Meets Key Clinical Needs in Myasthenia Gravis
Medicine

Review Finds No Fatigue Scale Meets Key Clinical Needs in Myasthenia Gravis

November 17, 2026
Persistent Microglial Homeostasis Loss Contributes to Neurodegeneration and Neurological Dysfunction
Medicine

Persistent Microglial Homeostasis Loss Contributes to Neurodegeneration and Neurological Dysfunction

November 17, 2026
Consortium Launches Framework to Standardize Atrial Cardiomyopathy Care Across Specialties
Science News

Consortium Launches Framework to Standardize Atrial Cardiomyopathy Care Across Specialties

November 17, 2026
Comprehensive Review Details Pathophysiology, Fertility Challenges, and Emerging Therapies in Congenital Adrenal Hyperplasia
Science News

Comprehensive Review Details Pathophysiology, Fertility Challenges, and Emerging Therapies in Congenital Adrenal Hyperplasia

November 17, 2026
Next Post
Boundary Element Framework Cuts Computation Time in Nonlinear Thermo-Viscoplastic Solids

Boundary Element Framework Cuts Computation Time in Nonlinear Thermo-Viscoplastic Solids

  • Mothers who receive childcare support from maternal grandparents show more optimized

    Mothers who receive childcare support from maternal grandparents show more parental warmth, finds NTU Singapore study

    27656 shares
    Share 11059 Tweet 6912
  • University of Seville Breaks 120-Year-Old Mystery, Revises a Key Einstein Concept

    1061 shares
    Share 424 Tweet 265
  • Bee body mass, pathogens and local climate influence heat tolerance

    682 shares
    Share 273 Tweet 171
  • Researchers record first-ever images and data of a shark experiencing a boat strike

    546 shares
    Share 218 Tweet 137
  • Groundbreaking Clinical Trial Reveals Lubiprostone Enhances Kidney Function

    531 shares
    Share 212 Tweet 133
Science

Embark on a thrilling journey of discovery with Scienmag.com—your ultimate source for cutting-edge breakthroughs. Immerse yourself in a world where curiosity knows no limits and tomorrow’s possibilities become today’s reality!

RECENT NEWS

  • Combined Cataract and Microhook Glaucoma Surgery Proves Safe in Older Adults
  • Mature Stone’s sheep reduce home range extent during hunting periods
  • Review Finds No Fatigue Scale Meets Key Clinical Needs in Myasthenia Gravis
  • Persistent Microglial Homeostasis Loss Contributes to Neurodegeneration and Neurological Dysfunction

Categories

  • Agriculture
  • Anthropology
  • Archaeology
  • Athmospheric
  • Biology
  • Biotechnology
  • Blog
  • Bussines
  • Cancer
  • Chemistry
  • Climate
  • Earth Science
  • Editorial Policy
  • Marine
  • Mathematics
  • Medicine
  • Pediatry
  • Policy
  • Psychology & Psychiatry
  • Science Education
  • Science News
  • Social Science
  • Space
  • Technology and Engineering

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 5,151 other subscribers

© 2026 Scienmag - Science Magazine

Welcome Back!

Login to your account below

Forgotten Password?

Retrieve your password

Please enter your username or email address to reset your password.

Log In
No Result
View All Result
  • HOME
  • SCIENCE NEWS
  • CONTACT US

© 2026 Scienmag - Science Magazine

Discover more from Science

Subscribe now to keep reading and get access to the full archive.

Continue reading