Access to large, annotated samples represents a considerable challenge for training accurate deep-learning models in medical imaging. While current leading-edge transfer learning from pre-trained models can help with cases lacking data, it limits design choices, and generally results in the use of unnecessarily large models. We propose a novel, self-supervised training scheme for obtaining high-quality, pre-trained networks from unlabeled, cross-modal medical imaging data, which will allow for creating accurate and efficient models. We demonstrate this by accurately predicting optical coherence tomography (OCT)-based retinal thickness measurements from simple infrared (IR) fundus images. Subsequently, learned representations outperformed advanced classifiers on a separate diabetic retinopathy classification task in a scenario of scarce training data. Our cross-modal, three-staged scheme effectively replaced 26,343 diabetic retinopathy annotations with 1,009 semantic segmentations on OCT and reached the same classification accuracy using only 25% of fundus images, without any drawbacks, since OCT is not required for predictions. We expect this concept will also apply to other multimodal clinical data-imaging, health records, and genomics data, and be applicable to corresponding sample-starved learning problems.
2.
Purpose
To determine the anatomical and functional outcomes of an extended 6-month intravitreal anti-vascular endothelial growth factor (anti-VEGF) upload in choroidal neovascularization (CNV) secondary to chronic central serous chorioretinopathy (CSCR).
Methods
A retrospective database analysis was performed applying the following inclusion criteria: (1) diagnosis of CSCR, (2) diagnosis of secondary CNV, and (3) treatment of at least six consecutive injections of anti-VEGF. Outcome measures included the change of central retinal subfield thickness, remodeling of the pigment epithelium detachments, and change in visual function.
Results
Twenty-one eyes of 21 patients were included. Mean patient age was 65 ± 8.3 years, and 35% of the patients (n = 8) were female. Mean disease duration before diagnosis of CNV was 48 ± 25.3 months. Mean central retinal thickness decreased from 346 ± 61 to 257 ± 57 μm (p < 0.01) after the sixth injection while mean visual acuity improved from 0.65 ± 0.35 to 0.49 ± 0.29 (logMAR; p < 0.01). Of note, an extended upload of six as opposed to three injections yielded an additional mean central retinal thickness reduction (280 ± 46 μm vs. 257 ± 57 μm, p = 0.038). Significant CNV remodeling was observed as a decrease in pigment epithelium detachment (PED) vertical (p = 0.021) and horizontal diameter (p = 0.024) as well as PED height (p < 0.01).
Conclusion
An extended anti-VEGF upload of six consecutive injections seems to be effective in inducing CNV remodeling and fluid resorption in CNV complicating chronic CSCR.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.