Related Experiment Video
Updated: Jun 27, 2026
06:29
Comparing Objective Conjunctival Hyperemia Grading and the Ocular Surface Disease Index Score in Dry Eye Syndrome During COVID-19
Published on: May 25, 2022
2.3K
Classifying Dry Eye Disease Patients from Healthy Controls Using Machine Learning and Metabolomics Data.
Sajad Amouei Sheshkal1,2,3, Morten Gundersen3,4, Michael Alexander Riegler1,2
1Department of Computer Science, Oslo Metropolitan University, 0166 Oslo, Norway.
Diagnostics (Basel, Switzerland)
|December 17, 2024
Summary
Machine learning models can identify dry eye disease in cataract patients using tear film metabolomics. Logistic regression demonstrated superior performance, outperforming complex models for early detection.
Area of Science:
- Ophthalmology
- Biochemistry
- Computational Biology
Background:
- Dry eye disease is a prevalent ocular surface disorder diagnosed via clinical signs and symptoms.
- Metabolomics offers a promising approach for early detection of dry eye disease by identifying unique metabolic profiles.
- This study investigates the novel application of machine learning and metabolomics for identifying dry eye disease in cataract patients.
Purpose of the Study:
- To explore the efficacy of machine learning models in detecting dry eye disease in cataract patients using metabolomics data.
- To compare the performance of various machine learning models for this specific diagnostic challenge.
- To identify the most suitable machine learning model for analyzing metabolomics data in the context of dry eye disease.
Main Methods:
- A comparative analysis of eight machine learning models was performed on metabolomics data from cataract patients.
- Models were evaluated and optimized using nested k-fold cross-validation.
- Performance was assessed using metrics tailored to the dataset's characteristics, including AUC, balanced accuracy, MCC, F1-score, and specificity.
Main Results:
- Logistic regression achieved the highest performance, with an AUC of 0.8378, balanced accuracy of 0.735, and F1-score of 0.8513.
- XGBoost and Random Forest models also showed strong performance.
- The logistic regression model with L2 regularization proved effective on an imbalanced dataset with limited samples and numerous features.
Conclusions:
- Machine learning, particularly logistic regression, can effectively identify dry eye disease in cataract patients from tear film metabolomics data.
- A simpler logistic regression model can outperform complex models in specific scenarios, avoiding overfitting and ensuring consistent results.
- This research highlights the potential of metabolomics and machine learning for early and accurate diagnosis of dry eye disease in complex patient populations.
