Diabetic retinopathy (DR) is simply a starring origin of blindness among working-age adults, affecting much than 130 cardinal group worldwide. Although advances successful ophthalmic imaging person improved illness monitoring, astir patients are diagnosed only aft years of unrecognized retinal damage. Early molecular and cellular changes—including neurodegeneration, inflammation, and vascular dysfunction—often stay undetectable pinch modular methods specified arsenic fundus photography aliases angiography. As a result, patients whitethorn acquisition irreversible imagination impairment earlier diagnosis. Due to these limitations, caller noninvasive imaging biomarkers are urgently needed to detect subclinical retinal alterations astatine nan earliest stages of diabetes.
A investigation squad from nan University of Coimbra, Portugal, has developed a texture-based study of optical coherence tomography (OCT) images tin of detecting early retinal changes successful type 2 diabetes. The study, Eye and Vision connected September 3, 2025, utilized a high-fat-diet and low-dose streptozotocin rat exemplary to show retinal alterations complete 12 weeks. By quantifying microscopic texture variations, nan method revealed early neurovascular abnormalities that occurred good earlier accepted biomarkers aliases vascular leakage.
Using precocious image analysis, nan researchers evaluated complete 80 retinal scans from diabetic and power rats, applying a gray-level co-occurrence matrix (GLCM) attack to quantify texture parameters crossed retinal layers. Among nan 20 features examined, eight—including autocorrelation, cluster prominence, correlation, homogeneity, accusation measurement of relationship II (IMCII), inverse quality infinitesimal normalised (IDN), inverse quality normalised (INN), and sum average—showed important changes successful diabetic retinas, peculiarly successful nan soul plexiform furniture (IPL) and photoreceptor segments (IS/OS). Interestingly, 7 of these metrics had besides been altered successful a erstwhile study utilizing a type 1 glucosuria model, reinforcing their diagnostic consistency. Despite minimal thinning and delayed oscillatory potentials, nan retinas displayed nary awesome inflammation aliases vascular leakage, confirming that texture changes precede overt pathology. The findings item texture study arsenic a sensitive, quantitative method for detecting early structural disorganization successful nan retina—potentially bridging nan spread betwixt biologic alterations and objective diagnosis.
Our results show that texture study tin uncover infinitesimal retinal changes agelong earlier DR becomes clinically visible. By capturing subtle structural signals wrong OCT images, this attack opens a caller diagnostic model into nan earliest illness processes. It offers a measurement to place high-risk patients earlier imperishable imagination harm occurs, supporting earlier curen and amended outcomes. The coherence of these texture metrics crossed glucosuria models strengthens their imaginable arsenic cosmopolitan early biomarkers."
Professor António Francisco Ambrósio, co-senior writer of nan study
This investigation paves nan measurement for processing AI-assisted diagnostic devices that automatically surface for preclinical DR based connected retinal texture signatures. Integrating this study into regular OCT imaging could let ophthalmologists to place patients who show microscopic structural disruption—even erstwhile their imagination appears normal. Such early discovery whitethorn thief tailor personalized care, forestall irreversible retinal damage, and trim nan world load of diabetic blindness. Further objective tests are now needed to validate these findings successful quality subjects and refine algorithms for large-scale screening and teleophthalmology applications.
Source:
Journal reference:
Oliveira, S., et al. (2025). Early retinal changes successful type 2 glucosuria detected by texture-based OCT analysis: imaginable attack for subclinical diabetic retinopathy diagnosis. Eye and Vision. doi: 10.1186/s40662-025-00451-3. https://link.springer.com/article/10.1186/s40662-025-00451-3
English (US) ·
Indonesian (ID) ·