AI-Assisted identification of sex-specific patterns in diabetic retinopathy using retinal fundus images
Bibliographic record
Abstract
Diabetic retinopathy (DR) is a microvascular complication of diabetes that can lead to blindness if left untreated. Regular monitoring is crucial for detecting early signs of referable DR, and the progression to moderate to severe non-proliferative DR, proliferative DR (PDR), and macular edema (ME), the most common cause of vision loss in DR. Currently, aside from considerations during pregnancy, sex is not factored into DR diagnosis, management or treatment. Here we examine whether DR manifests differently in male and female patients, using a dataset of retinal images and leveraging convolutional neural networks (CNN) integrated with explainable artificial intelligence (AI) techniques. To minimize confounding variables, we curated 2,967 fundus images from a larger dataset of DR patients acquired from EyePACS, matching male and female groups for age, ethnicity, severity of DR, and hemoglobin A1C levels. Next, we fine-tuned two pre-trained VGG16 models-one trained on the ImageNet dataset and another on a sex classification task using healthy fundus images-achieving AUC scores of 0.72 and 0.75, respectively, both significantly above chance level. To uncover how these models distinguish between male and female retinas, we used the Guided Grad-CAM technique to generate saliency maps, highlighting critical retinal regions for correct classification. Saliency maps showed CNNs focused on different retinal regions by sex: the macula in females, and the optic disc and peripheral vasculature along the arcades in males. This pattern differed noticeably from the saliency maps generated by CNNs trained on healthy eyes. These findings raise the hypothesis that DR may manifest differently by sex, with women potentially at higher risk for developing ME, as opposed to men who may be at greater risk for PDR.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".