Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Correspondence
  • Published:

“Overfitting or insight?” Re-evaluating a blood-based machine learning model for high myopia

Eye (2026)Cite this article

Subjects

We read with interest the article by Yang et al. [1] describing a blood-based machine learning (ML) model for high myopia (HM) screening. The authors’ effort to develop an interpretable and low-cost diagnostic tool is commendable. However, several methodological and conceptual limitations call the robustness and clinical applicability of the findings into question.

First, the sample size (313 participants, including 158 with HM) is small relative to the complexity of the analytical approach. The study simultaneously trained eight ML classifiers including XGBoost, random forest, and SVM, on a dataset of modest scale. Such a limited sample heightens the risk of overfitting, as high-capacity algorithms can easily capture noise rather than true signal. Prior methodological analyses have shown that ML models trained on small medical datasets (N ≤ 300) frequently yield overly optimistic cross-validation metrics and fail to generalise beyond the derivation cohort [2]. The reported AUC of 0.954 on the training set therefore likely reflects model memorisation rather than genuine predictive power.

This is a preview of subscription content, access via your institution

Access options

Buy this article

  • Purchase on SpringerLink
  • Instant access to the full article PDF.

USD 39.95

Prices may be subject to local taxes which are calculated during checkout

References

  1. Yang Z, Wang M, Huang X, Luan R, Shao Y, Li X. Interpretable machine learning model based on blood parameters for screening high myopia. Eye. https://doi.org/10.1038/s41433-025-04114-4 (2025).

  2. Zantvoort K, Nacke B, Görlich D, Hornstein S, Jacobi C, Funk B. Estimation of minimal data set sizes for machine learning predictions in digital mental health interventions. npj Digit Med. 2024;7:361. https://doi.org/10.1038/s41746-024-01360-w.

    Article  PubMed  PubMed Central  Google Scholar 

  3. Mitchell WG, Dee EC, Celi LA. Generalisability through local validation: overcoming barriers due to data disparity in healthcare. BMC Ophthalmol. 2021;21:228. https://doi.org/10.1186/s12886-021-01992-6.

    Article  PubMed  PubMed Central  Google Scholar 

  4. Pei Y, Meng J, He W, Zhang K, Guo D, Lu Y, et al. Unique liver function in high myopia: associations with myopic macular degeneration. BMC Ophthalmol. 2025;25:677. https://doi.org/10.1186/s12886-025-04491-0.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  5. Jian L, Huang Z, Du Y, Zhu X. High myopia as a risk factor for severe liver disease in individuals with liver dysfunction: evidence from a prospective cohort. J Clin Med. 2025;14:5860. https://doi.org/10.3390/jcm14165860.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

Download references

Funding

The author received no funding for this work.

Author information

Authors and Affiliations

  1. Department of Statistics and Data Science Cornell University, Ithaca, NY, USA

    Zhihao Lei

Authors
  1. Zhihao Lei
    View author publications

    Search author on:PubMed Google Scholar

Contributions

Zhihao Lei: Conceptualisation, Investigation, Writing—Original Draft, Writing—Review & Editing.

Corresponding author

Correspondence to Zhihao Lei.

Ethics declarations

Competing interests

The author declares no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Lei, Z. “Overfitting or insight?” Re-evaluating a blood-based machine learning model for high myopia. Eye (2026). https://doi.org/10.1038/s41433-025-04234-x

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • Version of record:

  • DOI: https://doi.org/10.1038/s41433-025-04234-x

Search

Advanced search

Quick links