Remote Imaging of the Eye : Solutions and Research

Part
01
of three
Part
01

Remote Imaging of the Eye - Companies

Seven companies that have hardware or software for remote imaging of the eye and transmit images for assessment include Next Sight, Optomed, D-Eye, Welch Allyn, Phoenix Technology Group, Volk, and Essilor Instruments. Details regarding the solutions provided by these companies have been provided below.

Next Sight

Optomed

D-Eye

Welch Allyn

  • The Welch Allyn iExaminer system enables one to take images of the retinal nerve and fundus using an iPhone.
  • The images can be stored to a patient file then printed or sent via email to an eye specialist.
  • The iExaminer system consists of a PanOptic Ophthalmoscope, iExaminer Adapter, and the iExaminer App.

Phoenix Technology Group

  • Phoenix Connect by Phoenix Technology Group is a telemedicine platform for retinopathy of prematurity (ROP) screening.
  • The system remotely connects teams in nursing, ophthalmology, and neonatology to facilitate the professional screening of babies in the neonatal intensive care unit (NICU) swiftly.
  • Using this platform, reports can be generated and sent to the NICU.

Volk

  • Pictor Plus by Volk is a lightweight ophthalmic camera that can take high definition retinal images in daylight and in the dark.
  • Due to its high resolution images, Pictor Plus is effective for follow-up treatments and initial screenings.
  • Image sharing is one of the features if Pictor Plus. Images can be transferred through a USB port or WiFi.

Essilor Instruments

  • The RETINA400E Fundus Camera is Essilor Instruments' solution to high resolution retinal imagery. The device has automated operations and uses intuitive software.
  • RETINA400E uses ANAEYES software, which increases the speed of processing and analysis.
  • Results and images from the device can be analyzed remotely.
Part
02
of three
Part
02

Remote Imaging of the Eye - Papers

Advanced multimodal retinal imaging will facilitate accurate prediction of dry age-related macular degeneration with detailed features. Deep Convolutional Neural Networks automated AMD grading has shown a range of 88.4% (0.5%) to 91.6% accuracy compared to human expert grading.

1. A DEEP-LEARNING APPROACH FOR PROGNOSIS OF AGE-RELATED MACULAR DEGENERATION DISEASE USING SD-OCT IMAGING BIOMARKERS

  • PURPOSE: The purpose of the study is to diagnose dry age-related macular degeneration using SD-OCT Imaging Biomarkers.
  • SUMMARY: Stanford University proposed a hybrid sequential deep learning model using OCT scan dimension with 13,954 dry AMD observations and a 10-fold cross validation setting. They compared 671 longitudinal OCT scan datasets with 13,954 non-neovascular observations. They also used utilized genetic information to estimate the risk of dry AMD.
  • RESULTS: This proposed model achieved high accuracy (0.96 AUCROC) in predicting long-term and short term dry AMD progression.

2. MULTIMODAL IMAGING OF NONNEOVASCULAR AGE-RELATED MACULAR DEGENERATION

  • PURPOSE: This paper presents the importance of multi model imaging in predicting dry age-related macular degeneration.
  • SUMMARY:Macular Foundation, Inc., and To Prevent Blindness, Inc., states that unlike traditional methods, multi model imaging is featured with fundus autofluorescence, near-infrared reflectance (NIR), NIR-AF, fluorescence angiography, optical coherence tomography angiography, OCT (Optical Coherence Tomography Angiography detects the blood flow in retina and choroid), and others. These features help in developing high-contrast retinal images to detect exact areas of atrophy with depth-resolved segmentation.
  • RESULTS: Advanced multimodal retinal imaging will facilitate accurate prediction of dry age-related macular degeneration with detailed features.

3. USE OF DEEP LEARNING FOR DETAILED SEVERITY CHARACTERIZATION AND ESTIMATION OF 5-YEAR RISK AMONG PATIENTS WITH AGE-RELATED MACULAR DEGENERATION.

  • PURPOSE: To describe deep learning techniques for age-related eye diseases and detailed severity scale for dry age-related macular degeneration patients to estimate 5-year risk using color fundus images.
  • SUMMARY: The Johns Hopkins University conducted a study by collecting age-related eye disease datasets of 4613 participants using a multiclass classification setting. Further, they compared AMD severity scales using human grader and fundus photograph reading center graders using 4 step AMD severity scales and 9 step AMD severity scales based on DL regression.
  • RESULTS: The weighted κ scores for 4 step AMD severity scales is 0.77 and 9 step AMD severity scales is 0.74. Therefore, using deep learning technique in estimating risk for dry age-related macular degeneration patients assists physicians in providing advanced care.

4. PREDICTION OF INDIVIDUAL DISEASE CONVERSION IN EARLY AMD USING ARTIFICIAL INTELLIGENCE.

  • PURPOSE : Use of artificial intelligence to predict dry type AMD progression individually.
  • SUMMARY: Genentech, Inc., USA conducted a diagnostic study using artificial intelligence for the prediction of dry age-related macular degeneration. The study used standardized monthly optical coherence tomography (OCT) images given by independent graders. Further, they obtained automated volumetric segmentation of retinal pigment epithelium, drusen, outer neurosensory layers, and hyperreflective foci using OCT image analysis.
  • RESULTS: The study predicted accuracy of 0.68 for choroidal neovascularization and accuracy of 0.8 for geographic atrophy. This study suggests that artificial intelligence along with automated analysis of imaging biomarkers enables personalized estimation of AMD progression.

5. A NEW AND IMPROVED METHOD FOR AUTOMATED SCREENING OF AGE-RELATED MACULAR DEGENERATION USING ENSEMBLE DEEP NEURAL NETWORKS

  • PURPOSE: To develop automated screening method for individuals with dry age-related macular degeneration using Ensemble Deep Neural Networks.
  • SUMMARY: iHealthScreen. Inc., and Icahn School of Medicine used age-related eye disease study of 150,000 images qualitatively graded by experts. Further, they graded by using deep neural networks such as Inception-ResNet-V2 and Xception into 4 groups such as No AMD, early AMD, Intermediate AMD, and Advanced AMD. The gradings were compared to predict the accuracy using Ensemble Deep Neural Networks for estimating the risk of dry age-related macular degeneration.
  • RESULTS: Ensemble Deep Neural Networks has estimated the risk of dry age-related macular degeneration with an accuracy of 86% to 95.3% compared to highly qualified expert grading.

6. AUTOMATED GRADING OF AGE-RELATED MACULAR DEGENERATION FROM COLOR FUNDUS IMAGES USING DEEP CONVOLUTIONAL NEURAL NETWORKS.

  • PURPOSE: To develop automated screening method for individuals with dry age-related macular degeneration using Deep Convolutional Neural Networks.
  • SUMMARY: The Johns Hopkins University conducted another study with Age-related Eye Disease Study data set using Deep Convolutional Neural Networks for automated grading. In this study automated AMD grading with Deep Convolutional Neural Networks is compared with AMD grading given by trained clinical grader (Human grader) using transfer learning and universal features for over 130 000 images.
  • RESULTS: Deep Convolutional Neural Networks automated AMD grading has shown range of 88.4% (0.5%) and 91.6% accuracy when compared with Human expert grading.

7. IDENTIFICATION OF CANDIDATE GENES RESPONSIBLE FOR AGE-RELATED MACULAR DEGENERATION USING MICROARRAY DATA

  • PURPOSE: To analyze candidate genes accountable for dry age-related macular degeneration using micro array data.
  • SUMMARY: Fordham University and NYU School of Medicine conducted a study for identifying genes responsible for dry age-related macular degeneration using AMD micro array dataset. This study followed 4 methods such as Random Forest, Random Lasso, Naïve Bayes, and Ensemble Feature Selection for choosing the exact genes. Further, PCA was used to examine the relevance of the selected features. IMP was used to analyze the top five protein-coding genes.
  • RESULTS: After the study, it is concluded that JNK cascade, Non-canonical Wnt Pathwayten/PI3K/Akt pathway, and NF-kappaB pathway plays significant role in progression of dry age-related macular degeneration.

RELEVANT INSIGHTS

Part
03
of three
Part
03

Remote Imaging - Trends

  • Grady Health System introduced cameras into endocrinology, primary care clinics, and satellite neighborhood health clinics in order to allow diabetic patients to follow the schedule of their exams in an easier way. The solution helped in notifying the primary care doctor when the patient was in their clinic.
  • IRIS Telemedicine solution helped with Grady Health System's backlogged queue of eye center patients, most for diabetic screenings, by implementing its unique technology solutions. The queue dropped from 14,000 to 3,000 within a year.
  • In the United States, use of teleophthalmology is in its early stages but has the potential to improve access to care, decrease cost of care, and improve adherence to evidence-based protocols.
  • Banner Health introduced a teleophthalmology program which allows emergency room doctors to capture the image of a serious eye injury, or other vision disorders, and digitally transfer that image to an ophthalmology expert, improving the overall patient experience with efficiency and timely care.
  • The Paxos technology uses a specialized adapter that takes pictures of the front of the eye and retina using a smartphone. It includes a cloud and mobile phone-based HIPAA-compliant software to allow secure transmission of various vision and ocular tests. It was introduced by DigiSight Technologies and EyeNet.
  • A new technology system was introduced by Digital Optometrics that includes proprietary software, remotely-operated equipment, high-definition video conferencing, and Internet accessibility to enable optometrists to perform comprehensive eye exams remotely.

Research Strategy:

While there was a lot of information related to the ophthalmology market, including its market size and growth, we were not able to find information specific to the field of remote imaging services and solutions.

We began by looking for information in industry reports about the ophthalmology market in United States. We visited sites like Grandviewresearch, Mordor Intelligence, IBIS World, and Euromonitor but most of the information pertained to the market size, growth, and key players of the ophthalmology market.

We also looked through articles and media publications sites like Medicaleconomics and Healthcareitnews in order to find comments by industry experts on the trends driving the teleophtalmology market. However, we were only able to find information on the use of technology, solutions in eye treatment, ways to improve retinal tests through ophthalmology

As a last resort, we tried to gather information by going through Ophthalmology associations like the American Academy of Ophthalmology and the American Ophthalmological Society. However, we only found information on the clinical uses of ophthalmology.



Sources
Sources

From Part 02
Quotes
  • "We propose a hybrid sequential deep learning model to predict the risk of AMD progression in non-exudative AMD eyes at multiple timepoints, starting from short-term progression (3-months) up to long-term progression (21-months). Proposed model combines radiomics and deep learning to handle challenges related to imperfect ratio of OCT scan dimension and training cohort size. "
Quotes
  • "Nonneovascular AMD is a disease that has remarkable propensity to cause significant and progressive anatomic retinal disruption and represents a major burden of vision loss in the elderly population worldwide. Effective diagnosis and management relies heavily on a complete understanding of the broad multimodal imaging features of this disorder that are discussed in this review."
Quotes
  • "While millions of individuals show early age-related macular degeneration (AMD) signs, yet have excellent vision, the risk of progression to advanced AMD with legal blindness is highly variable. We suggest means of artificial intelligence to individually predict AMD progression."
Quotes
  • "The authors address this deficiency by applying modern data mining and machine learning feature selection algorithms to the AMD microarray dataset. In this paper four methods are utilized to perform feature selection: Naïve Bayes, Random Forest, Random Lasso, and Ensemble Feature Selection. "
From Part 03
Quotes
  • "The EHR integration allowed for best practice alerts to fire when a diabetic patient was due for an exam, thus notifying the primary care doctor while the patient was in their clinic to go ahead and capture the photos," he added. "
  • "In the end, the work queue dropped from 14,000 to 3,000 within a year of implementation. The number of preventative lasers and injections went up in the clinics and the outcomes for the patients have been much better, according to Khalifa."
Quotes
  • "Ophthalmic telemedicine in the United States is in its infancy but has the potential to improve access to care, decrease cost of care, and improve adherence to evidence-based protocols. Clinicians will have to reconsider and reevaluate traditional care delivery models as teleophthalmology and remote consultations become more readily available."
Quotes
  • "Banner Health introduces the teleophthalmology program, bringing real-time, remote medical eye care expertise to our emergency rooms. In partnership with DigiSight Technologies, a San Francisco-based digital health company, and EyeNet, a group of Phoenix-based community ophthalmologists, we have launched Paxos - a telehealth solution that enables healthcare teams to collaborate and make informed decisions at the point-of-care – at thirteen of our facilities across Arizona."
Quotes
  • "DigitalOptometrics introduced technology at Vision Expo East that enables optometrists to perform comprehensive eye exams remotely. "