Block Expanded DINORET: Adapting Natural Domain Foundation Models for Retinal Imaging Without Catastrophic Forgetting

Integrating deep learning into medical imaging is poised to greatly advance diagnostic methods but it faces challenges with generalizability. Foundation models, based on self-supervised learning, address these issues and improve data efficiency. Natural domain foundation models show promise for medical imaging, but systematic research evaluating domain adaptation, especially using self-supervised learning and parameter-efficient fine-tuning, remains underexplored. Additionally, little research addresses the issue of catastrophic forgetting during fine-tuning of foundation models. We adapted the DINOv2 vision transformer for retinal imaging classification tasks using self-supervised learning and generated two novel foundation models termed DINORET and BE DINORET. Publicly available color fundus photographs were employed for model development and subsequent fine-tuning for diabetic retinopathy staging and glaucoma detection. We introduced block expansion as a novel domain adaptation strategy and assessed the models for catastrophic forgetting. Models were benchmarked to RETFound, a state-of-the-art foundation model in ophthalmology. DINORET and BE DINORET demonstrated competitive performance on retinal imaging tasks, with the block expanded model achieving the highest scores on most datasets. Block expansion successfully mitigated catastrophic forgetting. Our few-shot learning studies indicated that DINORET and BE DINORET outperform RETFound in terms of data-efficiency. This study highlights the potential of adapting natural domain vision models to retinal imaging using self-supervised learning and block expansion. BE DINORET offers robust performance without sacrificing previously acquired capabilities. Our findings suggest that these methods could enable healthcare institutions to develop tailored vision models for their patient populations, enhancing global healthcare inclusivity.

Common and Rare Fundus Diseases Identification Using Vision-Language Foundation Model with Knowledge of Over 400 Diseases

The current retinal artificial intelligence models were trained using data with a limited category of diseases and limited knowledge. In this paper, we present a retinal vision-language foundation model (RetiZero) with knowledge of over 400 fundus diseases. Specifically, we collected 341,896 fundus images paired with text descriptions from 29 publicly available datasets, 180 ophthalmic books, and online resources, encompassing over 400 fundus diseases across multiple countries and ethnicities. RetiZero achieved outstanding performance across various downstream tasks, including zero-shot retinal disease recognition, image-to-image retrieval, internal domain and cross-domain retinal disease classification, and few-shot fine-tuning. Specially, in the zero-shot scenario, RetiZero achieved a Top5 score of 0.8430 and 0.7561 on 15 and 52 fundus diseases respectively. In the image-retrieval task, RetiZero achieved a Top5 score of 0.9500 and 0.8860 on 15 and 52 retinal diseases respectively. Furthermore, clinical evaluations by ophthalmology experts from different countries demonstrate that RetiZero can achieve performance comparable to experienced ophthalmologists using zero-shot and image retrieval methods without requiring model retraining. These capabilities of retinal disease identification strengthen our RetiZero foundation model in clinical implementation.

VAFO-Loss: VAscular Feature Optimised Loss Function for Retinal Artery/Vein Segmentation

Estimating clinically-relevant vascular features following vessel segmentation is a standard pipeline for retinal vessel analysis, which provides potential ocular biomarkers for both ophthalmic disease and systemic disease. In this work, we integrate these clinical features into a novel vascular feature optimised loss function (VAFO-Loss), in order to regularise networks to produce segmentation maps, with which more accurate vascular features can be derived. Two common vascular features, vessel density and fractal dimension, are identified to be sensitive to intra-segment misclassification, which is a well-recognised problem in multi-class artery/vein segmentation particularly hindering the estimation of these vascular features. Thus we encode these two features into VAFO-Loss. We first show that incorporating our end-to-end VAFO-Loss in standard segmentation networks indeed improves vascular feature estimation, yielding quantitative improvement in stroke incidence prediction, a clinical downstream task. We also report a technically interesting finding that the trained segmentation network, albeit biased by the feature optimised loss VAFO-Loss, shows statistically significant improvement in segmentation metrics, compared to those trained with other state-of-the-art segmentation losses.

Forecasting Future Humphrey Visual Fields Using Deep Learning

Purpose: To determine if deep learning networks could be trained to forecast a future 24-2 Humphrey Visual Field (HVF). Participants: All patients who obtained a HVF 24-2 at the University of Washington. Methods: All datapoints from consecutive 24-2 HVFs from 1998 to 2018 were extracted from a University of Washington database. Ten-fold cross validation with a held out test set was used to develop the three main phases of model development: model architecture selection, dataset combination selection, and time-interval model training with transfer learning, to train a deep learning artificial neural network capable of generating a point-wise visual field prediction. Results: More than 1.7 million perimetry points were extracted to the hundredth decibel from 32,443 24-2 HVFs. The best performing model with 20 million trainable parameters, CascadeNet-5, was selected. The overall MAE for the test set was 2.47 dB (95% CI: 2.45 dB to 2.48 dB). The 100 fully trained models were able to successfully predict progressive field loss in glaucomatous eyes up to 5.5 years in the future with a correlation of 0.92 between the MD of predicted and actual future HVF (p < 2.2 x 10 -16 ) and an average difference of 0.41 dB. Conclusions: Using unfiltered real-world datasets, deep learning networks show an impressive ability to not only learn spatio-temporal HVF changes but also to generate predictions for future HVFs up to 5.5 years, given only a single HVF.

Deep learning for predicting refractive error from retinal fundus images

Refractive error, one of the leading cause of visual impairment, can be corrected by simple interventions like prescribing eyeglasses. We trained a deep learning algorithm to predict refractive error from the fundus photographs from participants in the UK Biobank cohort, which were 45 degree field of view images and the AREDS clinical trial, which contained 30 degree field of view images. Our model use the "attention" method to identify features that are correlated with refractive error. Mean absolute error (MAE) of the algorithm's prediction compared to the refractive error obtained in the AREDS and UK Biobank. The resulting algorithm had a MAE of 0.56 diopters (95% CI: 0.55-0.56) for estimating spherical equivalent on the UK Biobank dataset and 0.91 diopters (95% CI: 0.89-0.92) for the AREDS dataset. The baseline expected MAE (obtained by simply predicting the mean of this population) was 1.81 diopters (95% CI: 1.79-1.84) for UK Biobank and 1.63 (95% CI: 1.60-1.67) for AREDS. Attention maps suggested that the foveal region was one of the most important areas used by the algorithm to make this prediction, though other regions also contribute to the prediction. The ability to estimate refractive error with high accuracy from retinal fundus photos has not been previously known and demonstrates that deep learning can be applied to make novel predictions from medical images. Given that several groups have recently shown that it is feasible to obtain retinal fundus photos using mobile phones and inexpensive attachments, this work may be particularly relevant in regions of the world where autorefractors may not be readily available.