Evaluation of prediction errors in nine intraocular lens calculation formulas using an explainable machine learning model
2024

Evaluating Intraocular Lens Calculation Formulas with Machine Learning

Sample size: 1430 publication 10 minutes Evidence: moderate

Author Information

Author(s): Oh Richul, Oh Joo Youn, Choi Hyuk Jin, Kim Mee Kum, Yoon Chang Ho

Primary Institution: Seoul National University Hospital

Hypothesis

The study aims to evaluate the relationship between prediction errors and ocular biometric variables in cataract surgery using machine learning.

Conclusion

The study found that the predictive power of current ocular biometric variables for prediction errors in new-generation intraocular lens formulas is low, suggesting these formulas are already optimized.

Supporting Evidence

  • The SRK/T formula had the highest R2 value of 0.231.
  • The Kane formula had the lowest R2 value of 0.021.
  • New-generation formulas like BUII and Cooke K6 showed low predictive power for ocular biometric variables.

Takeaway

This study looked at how well different formulas predict vision after cataract surgery and found that some formulas are already doing a good job, but we might need new measurements to improve them even more.

Methodology

The study retrospectively analyzed medical records of patients who underwent cataract surgery and used a machine learning model to evaluate prediction errors.

Potential Biases

The machine learning models could be overfitted to the dataset, which may introduce bias.

Limitations

The sample size was relatively small for machine learning, and the study only included one type of intraocular lens.

Participant Demographics

Mean age: 69.71 years, 62.8% female.

Statistical Information

P-Value

p<0.05

Confidence Interval

95% CI of R2 ranged from [-0.019, 0.297]

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.1186/s12886-024-03801-2

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication