Related Experiment Video
Updated: May 9, 2026
07:45
Tear-Derived Exosomal miR-15a as New Diagnostic Tool for Diabetic Retinopathy
Published on: December 30, 2025
Tear fluid proteomics multimarkers for diabetic retinopathy screening
Zsolt Torok1, Tunde Peto, Eva Csosz
1Department of Computer Graphics and Image Processing, University of Debrecen, Faculty of Informatics, Debrecen, Hungary. zsolt.torok@astridbio.com
BMC Ophthalmology
|August 8, 2013
Summary
Diabetic retinopathy screening using tear fluid protein biomarkers and machine learning showed limited accuracy. While Recursive Partitioning was the best algorithm, low sensitivity and specificity suggest it is not yet suitable for standalone screening.
Area of Science:
- Biochemistry
- Ophthalmology
- Computational Biology
Background:
- Diabetic retinopathy (DR) screening is crucial for preventing vision loss.
- Novel methods for DR screening are needed, particularly non-invasive approaches.
- Tear fluid biomarkers offer a potential avenue for early DR detection.
Purpose of the Study:
- To develop a novel method for diabetic retinopathy screening using tear fluid biomarkers.
- To evaluate the efficacy of protein biomarkers and machine learning algorithms for DR pre-screening.
Main Methods:
- Collected tear samples from 119 diabetic patients (165 eyes), with 55 healthy and 110 showing signs of DR.
- Utilized nano-HPLC coupled ESI-MS/MS mass spectrometry for protein identification in tear samples.
- Applied six machine learning algorithms (SVM, Recursive Partitioning, Random Forest, Naive Bayes, Logistic Regression, K-NN) to classify biomarkers.
Main Results:
- Recursive Partitioning demonstrated the highest accuracy among the six machine learning algorithms evaluated.
- The system achieved 74% sensitivity and 48% specificity in identifying diabetic retinopathy from tear biomarkers.
Conclusions:
- Protein biomarkers and machine learning alone are currently insufficient for reliable diabetic retinopathy screening due to low sensitivity and specificity.
- This approach may serve as a complementary tool to enhance existing image-based screening methods in automated systems.
