Mode
Text Size
Log in / Sign up

Retrospective study identifies imaging biomarkers for AMD progression using machine learning modelsCan an eye scan predict who will lose central vision from macular degeneration?

AI-generated summary of the cited source, checked by automated accuracy review. How we work

Key Takeaway
Consider imaging biomarkers for AMD progression cautiously; models require prospective validation.

A retrospective cohort study analyzed 324 patients with age-related macular degeneration (AMD) to develop machine learning models predicting progression from early to late AMD stages over 3 years. The study used multimodal fundus images but did not report specific interventions, comparators, or study setting. The analysis identified pigmentary abnormalities, total drusen area in the macular region, and ellipsoid zone characteristics as independent risk factors for progression (P < 0.05), while subfoveal choroidal thickness and choroidal capillary blood flow density were protective factors (P < 0.05).

Four machine learning models were evaluated: Random Forest achieved an AUC of 0.779 in training and 0.700 in validation; XG BOOST showed 0.702 in training and 0.762 in validation; Support Vector Machine reached 0.768 in training and 0.646 in validation; and K-Nearest Neighbor performed at 0.717 in training and 0.596 in validation. The study did not report effect sizes, absolute numbers, or confidence intervals for the identified factors.

Safety and tolerability data were not reported. Key limitations include the retrospective design, single-center nature, absence of external validation, and modest model performance in the validation set. The study did not report funding sources or conflicts of interest. These findings represent associations rather than causal relationships and should be interpreted cautiously until validated in prospective, multicenter studies with diverse populations.

Imagine slowly losing the sharp, central vision you use to read, drive, or see a loved one's face. That's the reality for people whose early age-related macular degeneration (AMD) progresses to its late, vision-threatening stages. A new study tried to answer a critical question: can we spot who is most at risk before the damage is done?

The research team analyzed detailed eye scans from 324 patients with AMD over three years. They found that certain features visible in the scans were linked to a higher chance of the disease getting worse. These 'risk factors' included pigmentary abnormalities (changes in the retina's color) and the total area of drusen (tiny yellow deposits under the retina). They also found two features linked to a lower risk: a thicker layer of blood vessels beneath the retina (the choroid) and better blood flow in those tiny vessels.

Using this data, the team built several computer models to try to predict which patients would progress. The best-performing model, called a Random Forest model, showed modest accuracy. It was correct about 70% of the time in a separate validation test. It's important to understand what this study does and doesn't tell us. Because it looked back at old patient records, it can only show associations, not prove that one thing causes another. The model was developed and tested on data from a single group of patients, so we don't know how well it would work for other people. The results are a step toward better prediction, but this tool isn't ready to help doctors make decisions for individual patients yet.

What this means for you:
An eye scan model shows modest accuracy in predicting which macular degeneration patients may progress.

Study Details

Study typeCohort
EvidenceLevel 3
PublishedMar 2026
View Original Abstract ↓
ObjectiveTo develop an integrated model that combines multimodal fundus image features for accurately predicting the individualized risk of progression from early to late stages of age-related macular degeneration (AMD).MethodsA retrospective analysis was conducted on the data of 324 patients with AMD. The patients were randomly divided into a training set (n = 227) and a validation set (n = 97) at a ratio of 7:3. The follow-up period was 3 years, and patients with disease progression were defined as the progression group. In the training set, indicators related to prognosis were screened through univariate analysis. After variable compression by LASSO regression, independent influencing factors for poor prognosis were determined using multivariate logistic regression. Random Forest, Support Vector Machine, XG BOOST, and K-Nearest Neighbor algorithm prediction models were constructed using Python software. The performance of the models was evaluated by the area under the receiver operating characteristic curve (AUC), and the optimal model was selected.ResultsThere were no significant differences in the baseline characteristics between the training set and the validation set patients (P > 0.05), indicating comparability. In the training set, multivariate logistic regression analysis showed that pigmentary abnormalities, the total area of drusen in the macular area, and the ellipsoid zone were independent risk factors for disease progression (P < 0.05), while subfoveal choroidal thickness and choroidal capillary blood flow density were independent protective factors (P < 0.05). The AUC values of the Random Forest model (0.779 in the training set and 0.700 in the validation set) were significantly higher than those of the K-Nearest Neighbor algorithm (0.717 in the training set and 0.596 in the validation set), the Support Vector Machine model (0.768 in the training set and 0.646 in the validation set), and XG BOOST (0.702 in the training set and 0.762 in the validation set), making it the optimal prediction model.ConclusionIn this study, an AMD progression prediction model based on multimodal fundus images was successfully developed, which can effectively identify patients at high risk of progression and provide a new paradigm for clinical individualized precision medicine.
Free Newsletter

Clinical research that matters. Delivered to your inbox.

Join thousands of clinicians and researchers. No spam, unsubscribe anytime.