• HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
Monday, November 16, 2026
BIOENGINEER.ORG
No Result
View All Result
  • Login
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
  • HOME
  • NEWS
  • EXPLORE
    • CAREER
      • Companies
      • Jobs
        • Lecturer
        • PhD Studentship
        • Postdoc
        • Research Assistant
    • EVENTS
    • iGEM
      • News
      • Team
    • PHOTOS
    • VIDEO
    • WIKI
  • BLOG
  • COMMUNITY
    • FACEBOOK
    • INSTAGRAM
    • TWITTER
No Result
View All Result
Bioengineer.org
No Result
View All Result
Home NEWS Science News Health

Time-Aware Transformer Forecasts Pediatric Myopia Progression From Irregular School Screening Data

by
November 16, 2026
in Health
Reading Time: 3 mins read
0
Time-Aware Transformer Forecasts Pediatric Myopia Progression From Irregular School Screening Data

Time-Aware Transformer Forecasts Pediatric Myopia Progression From Irregular School Screening Data

Share on FacebookShare on TwitterShare on LinkedinShare on RedditShare on Telegram

A compact time-aware transformer deep learning model can forecast one-year-ahead pediatric refractive error and detect rapid myopia progression directly from irregularly sampled school screening records without requiring data imputation, according to an internal validation study published in the Journal of Translational Medicine. Rising global myopia rates present a major public health challenge due to the risk of irreversible vision loss associated with high myopia, creating an urgent clinical need to identify children on rapidly worsening visual trajectories early. While standard school-based vision screening programs routinely characterise a child’s current refractive state, they do not forecast the future course of that trajectory. To address this limitation, investigators developed an architecture designed to handle uneven clinical screening intervals and evaluated whether incorporating school-level geospatial context provides meaningful prognostic information.

The deep learning model adapts a Time-Aware Transformer framework to evaluate irregularly sampled refractive screening sequences without relying on data imputation techniques. The system combines an Adaptive-Frequency Time Embedding mechanism, which tunes a sinusoidal frequency spectrum to variable clinical sampling intervals using two learnable parameters, with a temperature-scaled Gated Residual Network. This gated network selectively integrates school-level environmental indices alongside individual longitudinal screening measurements.

The researchers developed and evaluated the model using a Gold Standard cohort of 2,874 students from Yecheng County, China, all of whom had complete four-year screening trajectories collected between 2020 and 2023. The study evaluated two primary predictive outcomes: one-year-ahead spherical equivalent refraction as a continuous measure and the binary classification of rapid myopic progression, defined specifically as an annual decrease in spherical equivalent refraction of at least 0.75 diopters per year. The rigorous internal validation design incorporated a held-out test set, five-fold cross-validation, a ten-seed stability evaluation, a school-level grouped split, and leave-one-school-out validation protocols to assess model generalizability.

In test evaluations, the proposed time-aware transformer achieved a mean squared error of 0.168 D² and an R² of 0.929 when predicting one-year-ahead spherical equivalent refraction. For the secondary task of identifying rapid progressors, the model achieved an area under the receiver operating characteristic curve of 0.891. Model risk estimates demonstrated strong calibration, yielding a calibration slope of 0.968 and an intercept of -0.014 diopters. Decision-analytic assessment indicated that the transformer yielded higher net benefit than either referring all children or referring no children across a wide spectrum of referral probability thresholds ranging from 0.10 to 0.55.

Validation experiments demonstrated consistent performance across multiple resampling schemes, with the model recording a mean squared error of 0.171 ± 0.017 across five-fold cross-validation and 0.168 ± 0.004 across ten random seeds. When compared against nine baseline architectures using paired test-set bootstrap tests with post hoc multiplicity control, the model achieved accuracy comparable to the strongest baselines rather than a marked superiority. Specifically, the difference in root mean squared error between the proposed model and ContiFormer was only 0.010 diopters, an increment well below the measurement precision of non-cycloplegic autorefraction. However, the proposed transformer required 42 percent fewer parameters, using 52,000 versus 89,000 parameters, and trained 2.7 times faster than ContiFormer.

The study identified notable variations in predictive difficulty across distinct patient subgroups and geographic validation splits. Predicting refractive trajectories proved most difficult for children who converted to myopia during the follow-up period, where the mean squared error reached 0.243, while forecasting was easiest among persistent non-myopes, who exhibited a mean squared error of 0.105. Furthermore, when entire schools were held out during evaluation, the overall mean squared error increased to 0.191, and the performance benefit attributable to environmental features dropped from 0.027 to 0.005 D². This substantial reduction indicated that the apparent geospatial signal predominantly captures school identity rather than direct, measured environmental exposures.

The authors concluded that the compact time-aware transformer successfully forecasts one-year-ahead refractive error from routinely collected school screening records with accuracy comparable to substantially larger continuous-time models. The primary justification for adopting this deep learning architecture rests on its operational efficiency, parameter stability, and clinically interpretable attention mechanisms rather than on raw predictive superiority over existing complex models. Furthermore, the findings indicate that school-level environmental indices must be understood as geospatial baseline priors rather than direct individual-level causal exposures when modeling refractive progression in pediatric screening programs.

Subject of Research: Deep learning forecasting of pediatric myopia progression from irregular longitudinal school screening records

Article Title: Development and internal validation of a time-aware transformer model for predicting pediatric myopia progression from irregular longitudinal school screening data

Article References: Zhang, M., Wumaier, T., Maimaiti, A., Zhao, Y., Moutari, S., Usama, M., Zhang, Z., Bu, S., Pazo, E. E., & Ren, X. (2026). Development and internal validation of a time-aware transformer model for predicting pediatric myopia progression from irregular longitudinal school screening data. Journal of Translational Medicine. https://doi.org/10.1186/s12967-026-09221-8

Image Credits: AI Generated

DOI: 10.1186/s12967-026-09221-8

Keywords: pediatric myopia, myopia progression, transformer models, deep learning, spherical equivalent refraction, school screening, longitudinal data, irregular sampling, time-aware transformer, ophthalmology, internal validation, refractive error, environmental indices, ContiFormer, gated residual network, adaptive-frequency time embedding, Yecheng County, clinical calibration

News Source: Harold Sullivan. (November 16, 2026). Time-Aware Transformer Forecasts Pediatric Myopia Progression From Irregular School Screening Data. Scienmag.

Tags: adaptive-frequency time embeddingclinical calibrationContiFormerdeep learningenvironmental indicesgated residual networkinternal validationirregular samplinglongitudinal datamyopia progressionOphthalmologyPediatric myopiarefractive errorschool screeningspherical equivalent refractiontime-aware transformerTransformer modelsYecheng County
Share12Tweet7Share2ShareShareShare1

Related Posts

Thirteen Blood Proteins Predict Insulin Resistance and Cardiometabolic Disease Risk

Thirteen Blood Proteins Predict Insulin Resistance and Cardiometabolic Disease Risk

November 16, 2026
Awake Glioma Surgery Language Deficits Mostly Resolve Within Three Months

Awake Glioma Surgery Language Deficits Mostly Resolve Within Three Months

November 16, 2026

AI-Based Clinical Simulation May Support Medical Training, Scoping Review Finds

November 16, 2026

Thirty-Two-Year Study Identifies Seven Scrotal Mesothelioma Cases With Early Asbestos Exposure

November 16, 2026

POPULAR NEWS

  • Information-Based Prescription Review Systems Evolve from Manual Screening to Intelligent Interception

    Information-Based Prescription Review Systems Evolve from Manual Screening to Intelligent Interception

    29 shares
    Share 12 Tweet 7
  • Study Reveals Incidence of Rare Pediatric Hip Disorders in Japanese Children

    29 shares
    Share 12 Tweet 7
  • Children With Nephrotic Syndrome Exhibit Severe Disruption Across Gut Microbial Kingdoms

    29 shares
    Share 12 Tweet 7
  • Automated Platform Maps Genomic Hotspots in CHO Cells for Antibody Production

    29 shares
    Share 12 Tweet 7

About

We bring you the latest biotechnology news from best research centers and universities around the world. Check our website.

Follow us

Recent News

Thirteen Blood Proteins Predict Insulin Resistance and Cardiometabolic Disease Risk

Spin Infection Protocol Enhances Adenoviral Gene Delivery in Mesenchymal Stromal Cells

Aroid Corms Reveal Contrasting Calcium Oxalate Crystal Deposition and Mineral Profiles

Subscribe to Blog via Email

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

Join 88 other subscribers
  • Contact Us

Bioengineer.org © Copyright 2023 All Rights Reserved.

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
  • Homepages
    • Home Page 1
    • Home Page 2
  • News
  • National
  • Business
  • Health
  • Lifestyle
  • Science

Bioengineer.org © Copyright 2023 All Rights Reserved.