Objective: The advent of Electronic Medical Records (EMR) with large electronic imaging databases along with advances in deep neural networks with machine learning has provided a unique opportunity to achieve milestones in automated image analysis. Optical coherence tomography (OCT) is the most commonly obtained imaging modality in ophthalmology and represents a dense and rich dataset when combined with labels derived from the EMR. We sought to determine if deep learning could be utilized to distinguish normal OCT images from images from patients with Age-related Macular Degeneration (AMD). Design: EMR and OCT database study Subjects: Normal and AMD patients who had a macular OCT. Methods: Automated extraction of an OCT imaging database was performed and linked to clinical endpoints from the EMR. OCT macula scans were obtained by Heidelberg Spectralis, and each OCT scan was linked to EMR clinical endpoints extracted from EPIC. The central 11 images were selected from each OCT scan of two cohorts of patients: normal and AMD. Cross-validation was performed using a random subset of patients. Receiver operator curves (ROC) were constructed at an independent image level, macular OCT level, and patient level. Main outcome measure: Area under the ROC. Results: Of a recent extraction of 2.6 million OCT images linked to clinical datapoints from the EMR, 52,690 normal macular OCT images and 48,312 AMD macular OCT images were selected. A deep neural network was trained to categorize images as either normal or AMD. At the image level, we achieved an area under the ROC of 92.78% with an accuracy of 87.63%. At the macula level, we achieved an area under the ROC of 93.83% with an accuracy of 88.98%. At a patient level, we achieved an area under the ROC of 97.45% with an accuracy of 93.45%. Peak sensitivity and specificity with optimal cutoffs were 92.64% and 93.69% respectively. Conclusions: Deep learning techniques achieve high accuracy and is effective as a new image classification technique. These findings have important implications in utilizing OCT in automated screening and the development of computer aided diagnosis tools in the future.

PurposeTo characterize the ocular surface microbiome of healthy volunteers using a combination of microbial culture and high-throughput DNA sequencing techniques.MethodsConjunctival swab samples from 107 healthy volunteers were analyzed by bacterial culture, 16S rDNA gene deep sequencing (n = 89), and biome representational in silico karyotyping (BRiSK; n = 80). Swab samples of the facial skin (n = 42), buccal mucosa (n = 50), and environmental controls (n = 27) were processed in parallel. 16S rDNA gene quantitative PCR was used to calculate the bacterial load in each site. Bacteria were characterized by site using principal coordinate analysis of metagenomics data. BRiSK data were analyzed for presence of fungi and viruses.ResultsCorynebacteria, Propionibacteria, and coagulase-negative Staphylococci were the predominant organisms identified by all three techniques. Quantitative 16S PCR demonstrated approximately 0.1 bacterial 16S rDNA/human actin copy on the ocular surface compared with greater than 10 16S rDNA/human actin copy for facial skin or the buccal mucosa. The conjunctival bacterial community structure is distinct compared with the facial skin (R = 0.474, analysis of similarities P = 0.0001), the buccal mucosa (R = 0.893, P = 0.0001), and environmental control samples (R = 0.536, P = 0.0001). 16S metagenomics revealed substantially more bacterial diversity on the ocular surface than other techniques, which appears to be artifactual. BRiSK revealed presence of torque teno virus (TTV) on the healthy ocular surface, which was confirmed by direct PCR to be present in 65% of all conjunctiva samples tested.ConclusionsRelative to adjacent skin or other mucosa, healthy ocular surface microbiome is paucibacterial. Its flora are distinct from adjacent skin. Torque teno virus is a frequent constituent of the ocular surface microbiome. (ClinicalTrials.gov number, NCT02298881.)

Evaluation of clinical images is essential for diagnosis in many specialties. Therefore the development of computer vision algorithms to help analyze biomedical images will be important. In ophthalmology, optical coherence tomography (OCT) is critical for managing retinal conditions. We developed a convolutional neural network (CNN) that detects intraretinal fluid (IRF) on OCT in a manner indistinguishable from clinicians. Using 1,289 OCT images, the CNN segmented images with a 0.911 cross-validated Dice coefficient, compared with segmentations by experts. Additionally, the agreement between experts and between experts and CNN were similar. Our results reveal that CNN can be trained to perform automated segmentations of clinically relevant image features.

Increased AD risk was found for recent glaucoma diagnoses, established AMD diagnoses, and both recent and established DR. People with certain ophthalmic conditions may have increased AD risk.

Optical coherence tomography angiography (OCTA) allows for the evaluation of functional retinal vascular networks without a need for contrast dyes. For sophisticated monitoring and diagnosis of retinal diseases, OCTA capable of providing wide-field and high definition images of retinal vasculature in a single image is desirable. We report OCTA with motion tracking through an auxiliary real-time line scan ophthalmoscope that is clinically feasible to image functional retinal vasculature in patients, with a coverage of more than 60 degrees of retina while still maintaining high definition and resolution. We demonstrate six illustrative cases with unprecedented details of vascular involvement in retinal diseases. In each case, OCTA yields images of the normal and diseased microvasculature at all levels of the retina, with higher resolution than observed with fluorescein angiography. Wide-field OCTA technology will be an important next step in augmenting the utility of OCT technology in clinical practice.