Predicting risk of late age-related macular degeneration using deep learning

Abstract

By 2040, age-related macular degeneration (AMD) will affect~288 million people worldwide. Identifying individuals at high risk of progression to late AMD, the sight-threatening stage, is critical for clinical actions, including medical interventions and timely monitoring. Although deep learning has shown promise in diagnosing/screening AMD using color fundus photographs, it remains difficult to predict individuals' risks of late AMD accurately. For both tasks, these initial deep learning attempts have remained largely unvalidated in independent cohorts. Here, we demonstrate how deep learning and survival analysis can predict the probability of progression to late AMD using 3298 participants (over 80,000 images) from the Age-Related Eye Disease Studies AREDS and AREDS2, the largest longitudinal clinical trials in AMD. When validated against an independent test data set of 601 participants, our model achieved high prognostic accuracy (5-year C-statistic 86.4 (95% confidence interval 86.2-86.6)) that substantially exceeded that of retinal specialists using two existing clinical standards (81.3 (81.1-81.5) and 82.0 (81.8-82.3), respectively). Interestingly, our approach offers additional strengths over the existing clinical standards in AMD prognosis (e.g., risk ascertainment above 50%) and is likely to be highly generalizable, given the breadth of training data from 82 US retinal specialty clinics. Indeed, during external validation through training on AREDS and testing on AREDS2 as an independent cohort, our model retained substantially higher prognostic accuracy than existing clinical standards. These results highlight the potential of deep learning systems to enhance clinical decision-making in AMD patients.

Multi-task deep learning-based survival analysis on the prognosis of late AMD using the longitudinal data in AREDS

“…This may be due to the fact that the EfficientNet models were shown to better capture fine image details within an image 17 . Further, we improved upon the standard CoxPH model that is commonly used in previous literature 11 using deep learning methods. Last, while incorporating longitudinal data we do see improvements in short-term prediction performance and in clinical feature performance as compared to using a single time point data, but do not see improved long-term performance.…”

“…In addition, multi-task models have been developed to characterize these eye characteristics simultaneously 10 . Subsequently, researchers have used these image features derived from a CNN model in a survival setting to predict patients who are at risk of developing late-stage AMD 11 . However, it is well known that the rate of progression for patients within the early-stage AMD category is heterogeneous.…”

Deep survival analysis with longitudinal X-rays for COVID-19

“…As mentioned in section 3.4, the negative partial likelihood function loss function is computationally expensive and almost infeasible (in terms of GPU memory) to compute when we have a large patient group with unstructured image data. [30] adopted an alternative two-step training strategy which requires first training a feature extraction network on image data with expert labelling, then uses the extracted features as covariates of the Cox model. Using mini-batch sampling, our training process is end-to-end and does not require any expert labelling.…”

LONGL-Net: temporal correlation structure guided deep learning model to predict longitudinal age-related macular degeneration severity

“…Nevertheless, they are not capable of predicting progression to advanced AMD, which is more clinically desirable because it enables clinicians to start preventive actions such as prescribing AREDS nutritional supplements to make the progression of vulnerable subjects to late AMD slow, considering that late AMD is currently incurable and irreversible. The recent 2 studies ( 15 , 34 ) have used fundus images w/o AMD-associated independent genetic variants of the subjects to predict whether the AMD progression time for them will be longer than a predetermined duration or not (2–7 years in ( 15 ) and 1–5 years in ( 34 )). Bridge et al ( 35 ), inspired by advances in sequence modeling architectures ( 36 , 37 ), have developed a model which uses a sequence of longitudinal images to predict the AMD risk of the subject in the future.…”