15 December 2023 in IEEE Access
Diabetic retinopathy (DR) is a leading cause of vision loss. Therefore, screening for early signs and assessment of the DR severity is crucial and extensively studied. To support clinicians, screening could be automated by algorithms that can, for example, refer difficult decisions to specialists for further investigation. However, frequently used neural networks (NNs) typically do not know when they do not know and approximate Bayesian NNs often equally do not suffice to provide well-calibrated uncertainty estimates. Thus, in this work, we investigate whether and how to use deep kernel learning (DKL) which we designed as a hybrid combination of the state-of-the-art EfficientNet-B0 and a Gaussian process (GP) layer to improve the quality of uncertainty estimates in referral-based DR screening. To this end, we first analyze the necessity for recently proposed extensions to the DKL framework to resolve miscalibrated uncertainties, despite the theoretical superiority of GPs to uncertainty quantification. Our subsequent comprehensive comparison of the curated DKL’s performance to that of the most common approximate Bayesian NNs shows our DKL models to particularly improve the detection of near out-of-distribution (OOD) samples containing other eye diseases through epistemic uncertainty information, but also enhance the calibration of aleatoric in-distribution uncertainty and diagnostic performance. Hence, it can provide a substantial benefit for safety-critical medical applications, like automated DR screening, particularly by potentially reducing the risk of missing diseases other than DR due to the improved near-OOD detection performance.
27 September 2023 in Biomedical Signal Processing and Control
1 January 2020 in IFAC-PapersOnLine
1 August 2022 in Current Directions in Biomedical Engineering
The retinal disease Diabetic retinopathy (DR) is one of the most probable causes of blindness. Automatic detection of DR is mostly done using convolutional neural networks (CNNs) on colour retinal images. This work in contrast uses stochastic variational deep kernel learning (SVDKL) for DR grading, combining a deep CNN with Gaussian processes (GPs) into a single end-to-end trainable model, which promises to provide predictions with a reliable uncertainty estimate exploiting approximate Bayesian inference. Evaluating the performance and uncertainty calibration of SVDKL on DR grading compared to a plain CNN, the EfficientNet-B0, preliminary results on a subset of the Kaggle DR dataset show a naturally enhanced uncertainty calibration for SVDKL over the plain CNN as well as a good diagnostic performance. Despite SVDKL achieving a slightly reduced accuracy, incorrect predictions were in closer proximity to the target stages, which is beneficial for clinical diagnosis due to minimizing the cost related to severe misclassifications.