National Eye Institute
National Advisory Eye Council
One Hundred Fifty-Fourth Meeting
January 17, 2020
The National Advisory Eye Council (NAEC) convened for its one hundred and fifty-fourth meeting at 8:30 am on Friday, January 17, 2020 at the Conference Center at 6700B Rockledge Drive, Bethesda, Maryland, 20892. Santa Tumminia, PhD, Acting Director of the National Eye Institute (NEI), presided as Chair of the Council, Anne E. Schaffner, PhD, as Executive Secretary, and Michael Steinmetz, PhD, as Director of Extramural Science Programs. The meeting was open to the public from 8:30 am until 1:00 pm. The meeting was closed to the public for a session from 1:45 pm until 03:30 pm for the review of confidentiality and conflict of interest procedures, grant and cooperative agreement applications.
Council Members Present:
Dr. Eduardo Alfonso
Dr. Jose-Manuel Alonso
Dr. Katia Del Rio-Tsonis
Dr. Thomas Glaser
Dr. Mary Elizabeth Hartnett
Dr. Dennis Levi
Dr. Carol Ann Mason
Dr. Louis Pasquale
Dr. Sylvia Smith
Dr. Mary Ann Stepp
Dr. Benjamin Teller
Dr. Russell Van Gelder
Dr. Marco Zarbin
NEI Staff Present
Mr. Shawn Adolphus
Dr. Houmam Araj
Dr. Neeraj Agarwal
Dr. Steven Becker
Ms. Cindy Best
Dr. Sangeeta Bhargava
Dr. Kapil Bharti
Dr. Rachel Bishop
Ms. Sylvia Braxton
Dr. Brian Brooks
Ms. Monique Clark
Mr. Jay Colbert
Ms. Karen Colbert
Dr. Mary Frances Cotch
Dr. Emily Chew
Ms. Ashley Dash
Ms. Kathryn Demott
Ms. Linda Dingle
Ms. Courtney Dodson
Dr. Lesley Earl
Mr. Donald Everett
Dr. Martha Flanders
Dr. Ashley Fortress
Dr. James Gao
Ms. Kerry Goetz
Dr. Shefa Gordon
Dr. Thomas Greenwell
Mr. Dustin Hays
Ms. Lateefah Hill
Ms. Terri Holmes
Dr. Brian Hoshaw
Dr. Jimmy Le
Dr. Ellen Liberman
Ms. Renee Livshin
Dr. George Ann McKie
Dr. Lisa Ann Neuhold
Dr. Gyan Prakash
Dr. Maryann Redford
Ms. Amber Reed
Ms. Karen Robinson-Smith
Ms. Jessica Ryan
Dr. Anne E. Schaffner
Dr. David Schneeweis
Ms. Shauna Schwartz
Dr. Peter Shan
Dr. Grace Shen
Ms. Karen Robinson Smith
Dr. Michael Steinmetz
Ms. Chantelle Stevenson
Mr. Brian Trent
Ms. Sharon Taylor
Dr. Santa Tumminia
Ms. Keturah Williams
Ms. Leslie West-Bushby
Dr. Cheri Wiggs
Ms. Nora Wong
Dr. Chuck Wright
Dr. Jerome Wujek
Ms. LeAnne Young
Ms. Maria Zacharias
Members of the General Public Present at the Open Session:
Ms. Erin Herins, Science Communications Manager, ARVO (Association for Research in Vision and Ophthalmology)
Mr. James Jorkasky, Executive Director, NAEVR (National Alliance for Eye and Vision Research)
Ms. Alison Manson, AOA (American Optometric Association)
Ms. Iris Rush, Executive Director, ARVO
Dr. Aaron Lee, Assistant Professor, University of Washington
Open Session of the Meeting
Call to Order and Opening Remarks
Dr. Santa Tumminia, Acting NEI Director
The Acting Director announced the retirement of four Council members. She thanked Tom Glaser, Dennis Levi, Sylvia Smith and Louis Pasquale for their service, which had been extended through May of 2020. Dr. Paul Sieving stepped down as Director of the NEI in July of 2019; a search for a new Director was initiated in November of 2019 and closed on December 2. Applicants were currently being reviewed; this will be followed by interviews and then an intense financial and ethics vetting process. The process could take some time. The search for a new Scientific Director is on hold to ensure that their research interests will align with those of the new NEI Director.
Dr. Tumminia briefed the group on staffing updates. The position of DEA (Division of Extramural Activities) Director, is expected to be filled within the next few months. There is an outstanding pool of candidates. Dr. Anne Schaffner was thanked for her capable service as Acting DEA Director. James Gao, PhD, recently joined the DESP (Division of Extramural Science programs) as a program analyst. Prior to coming to the NEI he was a scientific reviewer at the FDA in the Division of Molecular Genetics and Pathology. His area of expertise in genomics and bioinformatics aligns well with the current NIH emphasis on big data. He worked as a contractor at NEI from 2001-2008 on database administration and programming for a genomics project in the Molecular Structure and Function Section of the intramural program. Hongman Song, MD, PhD, also joined the DESP as a program analyst. She was most recently a postdoctoral fellow in Dr. Sieving’s lab and has a background in biochemistry and molecular biology. Ms. Lateefah Hill was recently hired as a program specialist in the DEA. She had been working as a contractor doing office administration. She will be taking over Council duties. The Office of Regenerative Medicine has a new program specialist, Ms. Amberlynn Lee. She has BS degrees in Public Health and Social Psychology and is currently working on a Master’s degree in Public Health. She was a disease intervention specialist at the Department of Defense and a public health technician with the Air Force.
Dr. Tumminia shared news of several accolades awarded to NEI intramural staff. Hendrikje Nienborg, a tenure track investigator in the intramural program, was recently named a 2019 Distinguished Scholar in a program that provides research support, as well as mentoring and professional development activities. The Distinguished Scholar’s program is extremely competitive and focuses on bringing diversity and inclusion to NIH intramural science programs. In terms of diversity, in looking over the Distinguished Scholars program over just the past 2 years, it was determined that the mentoring and networking opportunities were the most critical aspects of the program. Financial support is guaranteed for 5 years with talk of increasing that to 7 years. Dr. Kapil Bharti recently received FDA approval to launch the first clinical trial using autologous (patient-derived) RPE (retinal pigmented epithelial) tissue to generate iPS cell (induced pluripotent stem cell) lines for a cell-based therapy. The iPS-RPE cells will be reinjected into the patients with geographic atrophy to replace and repair dying RPE cells. The first patient will receive the injection this summer.
Some NIH activities of interest included an NIH Leadership Forum comprising NIH IC Directors and a meeting of the ACD (Advisory Council to the Director, Francis Collins). There were several overlapping topics including the HEAL (Helping to End Addiction Long-term) Initiative that addresses the opioid crisis, reproducibility and rigor in animal research, diversity, support of next generation researchers, changing the culture to end sexual harassment, artificial intelligence (AI), foreign influences and the new NIH strategic plan. Those interested in knowing more can go to the NIH ACD website and go to Meetings. AI, machine learning and data management were also big topics. A working group of the ACD made several recommendations to help the NIH and grantees deal more effectively with both prospective and retrospective data. One suggestion was that older data sets be discarded and new ones collected using machine learning. While this is not practical, it would be important for researchers to be mindful of what they are looking for and whether assumptions being made are accurate. [Note: the guest speaker, Dr. Aaron Lee, an AI and deep learning expert, made the comment during his presentation that older data sets should not be discarded] The NIH is also piloting some projects in machine learning, one is the Civic Digital Fellowship and a Graduate Data Summer Science Program. Machine learning is a major focus of these programs. Students who have expertise in computer science but no background in biomedical science are involved with real-time NIH research projects to give them an opportunity to see what they can contribute to the field of data management. The NEI had one student involved in the program who was touted as being a “success.” She was able to help with data mining and missing data, important aspects of data analysis. NIH is taking foreign influences very seriously and is working with other government agencies and academic institutions where violations have occurred. More information is available on the ACD web site. There was also discussion of a trans-NIH working group on geroscience that will involve the NEI. However, it has not been fully developed.
Dr. Tumminia also spoke to the fact that 2020 is an important year for the NEI. There are ongoing activities directed by the NEI Office of Communications. The NEI will also be partnering with NAEVER/AEVR and ARVO. Activities will include videos, educational programs, social media campaigns and invigorating the NEI’s NEHEP (National Eye Health Education Program) with new 2020 activities. The contact person at the NEI is Maria Zacharias, and Council members were encouraged to contact her for information and assistance in getting any 2020 activities out to members of their communities.
Dr. Tumminia mentioned several Council updates being presented at the meeting: Karen Colbert on the budget, Dr. Shefa Gordon on the strategic plan, Dr. Steven Becker on the Office of Regenerative Medicine, and Dr. Bruce Reed at CSR on the ENQUIRE process. Also scheduled to speak were Dr. Jerry Wujek on the SBIR and STTR portfolio and Dr. Aaron Lee from the University of Washington on Artificial Intelligence and deep learning applications in ophthalmology and vision science.
Administrative Issues, Consideration of October Minutes and Other Comments
Dr. Anne Schaffner, Executive Secretary and Acting Director, Division of Extramural Activities
The Executive Secretary asked that Council members sign the pre- and post-conflict of interest statements, at the appropriate time. She also asked for comments and corrections to the October 2019 Council minutes. There were no changes, and the minutes were unanimously approved.
Dr. Schaffner told Council members that an updated spreadsheet of current NEI PARs and RFAs would be posted to the Electronic Council Book with a special notation of those to be re-issued. New PARs or RFAs would be presented to Council as usual.
Dr. Schaffner also presented a potential change to the Council Operating Procedures. She asked if NEI Program staff could administratively handle internally the following issues that come up during peer review: human subject protections, inclusion, and vertebrate animals. Biohazards are already handled by staff, working with principal investigators (PIs) and their institutional officials. Staff would work with PIs and their institutional IRBs to resolve human subject and inclusion issues. Staff would address vertebrate animal concerns with OLAW (Office of Animal laboratory Welfare). She assured Council that applications of special interest (such as Phase I, II and III clinical trials) would be brought to Council, and those with unresolved issues or concerns would not be funded until all matters had been satisfactorily resolved. One Council member asked if there were statutory obligations for Council to review human subject concerns, and if there were clinicians on NEI staff. Dr. Schaffner said that she would look into the responsibilities of Council to look at human subject concerns. Another Council member wanted to continue to review ethics (equipoise) concerns in clinical trials. Dr. Schaffner said that she would reword the statement to make sure that Council would continue to review Phase I, II and III clinical trials and trials that had ethical concerns.
Ms. Karen Colbert, Budget Officer, Office of the Director
Ms. Karen Colbert began her presentation by saying that the Budget community was in a “gray” area, since 2019 spending and the 2020 operating plan had not been submitted to Congress. She would not be able to share specific details for that reason. A recap of FY 2019 showed that about 85% of NEI’s operating budget (minus rescissions and transfers) of just over 794 million dollars ($794M) was spent on extramural research. 11% was spent on intramural research, and 4% was spent on support (administrative costs, staff salaries etc.). NEI competing RPG success rates have always exceeded the NIH rates by about 7 points. For example, the NIH success rate in 2018 was 20%, whereas the NEI success rate was 27%. The numbers for 2019 had not been released. Unfortunately, buying power has not kept pace with increases in the budget, and in 2019 buying power was only about 55% of total funding. The FY 2020 President’s Budget for NEI proposed a 14% reduction compared to FY 2019; that would have cost the NEI 170 awards. However, the final Congressional bill awarded a 3.5% increase or $824.1M to the NEI. The NIH received a 6.7% increase. The difference between the NIH and the NEI budget was due to very targeted research mandated for the NIH in the spending bill. Funding levels for both the NIH and NEI have increased since the 2013 sequestration, although it took almost 4 years (2016) for the NIH and NEI to recover from sequestration. The FY 2020 appropriations bill contained several areas that the NIH was mandated to support including included Alzheimer’s research ($2.8B), BRAIN ($500M), Opioid research (at least $500M), and Firearms research ($12.5M) including gun-related deaths and the relationship between victim and shooter. The text also had text related to sexual harassment. NIH must require institutions to notify them when key personnel on an NIH grant are removed due to sexual harassment concerns. The NIH must also submit to Congress measures that address sexual harassment in extramural settings with the same level of attention and resources as that afforded to other research misconduct issues. NIH must support studies of sexual harassment including the psychology underlying harassment and experiences and outcomes of diverse groups that are subject to harassment. The NIH was directed to collaborate with NASA and the National Academy of Sciences (Engineering and Medicine) to develop best practices for more diverse and inclusive cultures.
The 21st Century Cures Act provided Congressional money for Precision Medicine and BRAIN Initiatives, Cancer Moonshot and Regenerative Medicine. Funds were projected through FY 2026; however, funds are not available until appropriated by Congress each FY and cannot exceed what was originally authorized for each FY. The total cap through 2026 is $4.8B. The BRAIN Initiative will see an increase in FY 2020 from $115M to $140M. FY 2020 will see the final year of funding for Regenerative Medicine.
The Budget community is tasked with managing 3 budget years at the same time, 2019, 2020 and 2021, so the NEI is also formulating the FY 2021 President’s Budget request. By law, on the first Monday of February, the President submits to Congress his detailed budget request for FY 2021.
Ms. Colbert entertained questions from Council members. The first was a question regarding success rate and if some ICs use the mechanism of withdrawing applications to increase their success rate, given what happened to an individual at their institution. Dr. Steinmetz commented that withdrawals occur at CSR not the ICs. He added that the reason for the withdrawal was likely the result of the investigator submitting their application under the wrong FOA for clinical trial versus no clinical trial. CSR was no longer offering a grace period for submissions under the incorrect FOA.
NEI Strategic Planning Update
Dr. Shefa Gordon, Director, Office of Program Planning and Analysis and Associate Director for Science Policy and Legislation, Office of the Director
Dr. Shefa Gordon presented an update of the NEI Strategic Plan entitled, “2020 Vision for the Future.” Historically, planning has been based on 6 NEI core programs, defined anatomically as Retina Disease, Corneal Diseases, Lens and Cataract, Glaucoma and Optic Neuropathies, Strabismus, Amblyopia and Visual Processing, and Low Vision and Blindness Rehabilitation. Three main Areas of Emphasis with subareas were defined: Visual System in Health and Disease (from Genes to Disease Mechanisms, Biology and Neuroscience of Vision, Immune System and Eye Health), Capitalizing on Emerging Fields (Regenerative medicine and Data Science), and Preventing Vision Loss and Enhancing Well-Being (Individual Quality of Life and Public Health and Disparities Research). 252 individual responses were received from the recent RFI. Respondents represented citizen advocates, teachers, medical professionals, academicians, government, professional organizations, and industry. Dr. Gordon broke down the number of responses by area of emphasis, and key words. The information was used to help frame what would be covered in each of the panels and to decide on appropriate panel members. Each subgroup is also represented by a Council member and an NEI staff person. One interesting topic frequently cited was CVI (Cortical Visual Impairment). The NIH is also in the midst of an overall planning process. There were no questions or comments from Council members.
AGI Update and RMIP Progress Report
Dr. Steven Becker, Associate Director, NEI Office of Regenerative Medicine (ORM), Office of the Director
Dr. Becker laid out a rationale for the new office. The NIH established an Office of Regenerative Medicine through its Common Fund over 10 years ago when the use of induced pluripotent stem cells came online. The field of regenerative medicine has unique needs that benefit from coordination of many scientific disciplines including bioinformatics, bioengineering, neuroscience and stem cell biology, just to name a few. The NEI ORM will provide information and resources to support the advancement of regenerative medicine in the vision community including funding sources across the NIH and other agencies, collaboration opportunities, the dissemination of links to data, cell, vector and virus repositories and notices of advancements in the field. One opportunity for collaboration within the NIH is through NCATS (National Center for Advancing Translational Science). They have a stem cell translational lab, and 3 times a year solicit proposals from the extramural community to collaborate on overcoming translational difficulties in iPSC technology. The ORM also provides specific information to support stem cell infrastructure, such as identifying sources of tissues and cells, sharing of methodologies, fee-for-service providers, facilitating tool and cell sharing and identifying potential industry collaborations. The ORM is also engaged in activities to generate cell replacement therapies to further the NEI’s AGI (Audacious Goals Initiative) and catalyze stem-cell based research through a 2020 3D Retinal Organoid Challenge and an AMD Integrative Biology Initiative. The ORM will communicate all of this valuable information via its website, monthly e-newsletters and social media.
Dr. Becker turned his attention to the AGI and regenerative medicine. Two of the 3 NEI-funded consortia (on functional imaging and regenerative factors) were finishing up their work, and the ORM will disseminate their findings to the community. The third consortium on animal models will be coming online in the next few years. Phase 2 of the 3D Organoid Challenge is a prize competition to develop a physiologically relevant retinal organoid model. Entries will be accepted until October 2 of 2020.
The NEI intramural program has a contract with the New York Stem Cell Foundation (NYSTF) to generate 100 patient-derived iPSC lines from AMD and Stargardt patients that were part of the AREDS2 cohort. The first 11 lines have been generated and will available soon. The entire AREDS2 dataset will also be available through dbGAP and an online portal, BRIX.
The NIH Regenerative Medicine Innovation Project (RMIP) received $30M in federal funding over 4 years and is in its last year of funding. A recent set of clinical applications, supporting research on adult stem cells, was reviewed, and there were a few vision-related studies that the RMIP will be considering for funding. Dr. Becker emphasized that the ORM is committed to dissemination of information and resources by hosting listservs and webinars, organizing workshops and publishing their reports and promoting resource sharing via platforms and repositories. Resources on regenerative medicine efforts will also be made available to the public, especially for legitimate NIH/NEI stem cell clinical trials, to differentiate them from recent online claims from unregulated stem cell clinics that prey on vulnerable people experiencing vision loss.
Information on the ORM will be presented at the World Stem Cell Summit, ARVO 2020 and an ISSCR (International Society for Stem Cell Research) exhibitor booth. Dr. Becker ended with a list of planned AGI activities in 2020 that included a workshop on “Understanding Human Retina Biology and Perception” in February and consortia investigator meetings in March and April. A 2020 ARVO Town Hall is planned on the “Immunology of Retinal and Tissue Transplants” to gain some understanding of the immune issues that must be considered.
Council members asked a few questions. One member commented that while stem cells were one way of approaching regenerative medicine, trans-differentiation was another way, e.g., differentiating native Muller cells to photoreceptors or RPE cells, which obviates the need to consider immune or possibly synaptic issues. Dr. Becker responded that it was originally a part of the scope, and while the field is advancing, it is not at the point of immediate translation. However, expertise in that area is represented on the various Regenerative Medicine panels. Another member emphasized the need to keep in mind more basic research in organisms like the salamander that retain the ability to regenerate. Dr. Becker agreed that it is necessary to strike a balance between what systems are closer to translatability with more basic science. A question was posed asking if there was a list of clinical trials that could be linked to University web sites. https://clinicaltrials.gov/ is one. The question of protocol and reagent sharing came up. Dr. Becker said one possibility would be to post protocols online or highlight them in a Webinar. A meeting has also been scheduled to discuss various tools that might be available to the AGI consortia to share data etc., first among collaborators, but then look at the possibility of making that information publicly available.
CSR ENQUIRE Process as Applied to Vision Applications
Dr. Bruce Reed, Deputy Director, NIH Center for Scientific Review
Dr. Reed’s presentation was to address Council concerns regarding a reorganization of vision-centric study sections and dissolution of an anterior segment SEP. The acronym stands for Evaluating Panel Quality in Review. The ENQUIRE process was established to keep study sections aligned with the evolving nature of science. The current process includes more external review by scientific clusters not internal organizational clusters. The process also included a look at a study section’s ability to distinguish strong vs weak science and the unequal distribution of strong vs weak science across study sections. In September of 2018, the vision study sections were reviewed by the Vision Science SRG review panel. They made the recommendation to charter BDCN J81, a special emphasis panel reviewing anterior segment and ocular surface applications. However, the CSR Advisory Council (CSR AC) Meeting in March of 2019 did not concur with that recommendation and asked that the vision science panels, as part of Cluster 11, (Functional Neuroscience) be reviewed by ENQUIRE. The process started with an External Evaluation Panel, and their recommendations were taken to a trans-NIH Internal Evaluation Panel, and then considered by the CSR AC. Dr. Reed went on to describe the process in more detail including the information and charge given to the External Panel (consider workload, scientific scope, emerging science, balanced competitive pool, etc.). He noted that 4 of the 12 external panel members had vision science expertise and had study section experience. With respect to the eye panels they recommended retaining a basic eye biology study section (BVS) and creating twin study sections to review all clinical translational disease science (DPVS 1 and 2) from both the front and the back of the eye. The Internal Evaluation Panel, which included Michael Steinmetz, did not agree with the External Panel’s recommendation and were in favor of retaining an ocular surface/anterior segment study section. However, the CSR AC sided with the External Panel. Dr. Reed went on to assure the NAEC that CSR and NEI share the same goals, evaluation is a continuing process and can be revisited, and twin study sections can be handled properly including that referral will not be random. CSR will be looking at the quality of discussion and how specific types of applications are faring.
One Council member asked a number of questions about what he called ground truth issues. First, how “good” science is determined prospectively; he made the point that the current CSR structure has actually produced good vision science, but only from a retrospective point of view. A second concern was the range of expertise on the external panel, and the fact that 5 (vision) panelists would not have had the range of expertise necessary to judge all of the vison study sections. A third issue was the method of analysis used to define good science. The number of citations is useful but incomplete. The last issue was the complex nature of judging scientific applications, especially when the science is not generic. As an example, a cornea expert would not have the knowledge to judge an application in retina. Given these concerns it would be difficult to know if things were not working, and even if that was evident, what changes could be made? Dr. Reed responded that problematic scoring patterns and the development of camps are signs of a dysfunctional study section. Remedies would include changing or reeducating reviewers and working with the SROs or IRG Chiefs. Judging the quality of the output would be based on several indicators including publications coming out of awarded grants, citations, journal impact factor, patents and new technologies. This same Council member remarked that bringing together the groups that disagree on the reorganization might provide a better solution, since there was no evidence presented that the original vision study sections were not working. Dr. Reed responded with the comment that only considering internal views was too insular, and it was necessary to consider multiple voices. Another Council member applauded CSR’s effort and said that if the changes were going forward, then the best chance for success was imbuing, in both the SROs and reviewers of those study sections, the philosophy of what constitutes “best science.” Dr. Reed said that the Director of CSR, Dr. Noni Byrnes understands the critical role of the SRO in the process. He also explained the role of the IRG chief, who makes referrals of applications to study sections, and said that fairness is achieved by percentiling applications within each study section and not across different study sections. Another question posed to Dr. Reed was the problem of an area of vision research that might not be understood or appreciated by the majority of reviewers. Dr. Reed saw that as a “camps” issue that could be alleviated by reviewer education. The use of natural language processing was also suggested by a member to help determine ground truth, good versus bad reviews and good versus bad science. AI (Artificial Intelligence) is something that CSR is considering. A Council member was upset that the advice of people in the NEI, with their historical knowledge, had been discarded in the process. In addition, it appeared to this member that the functionality of the anterior segment SEP (BDCN J81) had been compared to standing study section, but the SEP had never been chartered and thus was a false comparison. Dr. Reed replied that as the former Division Director that oversaw J81, he thought the SEP was working very well, and he was willing to charter it, but the CSR Council said no. He also stressed that CSR did hear the opinions of the all parties. Another question was regarding the current number of twin study sections at CSR. Dr. Reed said there were a few, some of which have been running for many years. He did not accept a Council member’s suggestion that PIs give anonymous feedback on reviewers. He preferred better reviewer training and feedback from surveys of study section reviewers on such aspects as SRO performance.
Applications of Artificial Intelligence in Ophthalmology and Vision Science
Dr. Aaron Lee, University of Washington, Seattle
Dr. Tom Greenwell introduced the guest speaker, Dr. Aaron Lee, a retinal specialist with an interest in Artificial Intelligence (AI) and machine learning.
Dr. Lee introduced his talk by defining terms such as Big Data, that includes data velocity (most clinical data are collected in real time), data volume (IRIS registry) and data variety (different imaging modalities). The registry already has terabytes to petabytes of data. Eye care is Big Data! Examples of tool sets available to analyze these data are various statistical languages such as python, but that requires custom-writing code. The field of AI is actually quite old. The field of machine learning is a subset of AI, and a new type of machine learning is deep learning (DL), an extension of artificial neural networks. Four advances that led to the birth of deep learning were 1) the use of graphics processing cards, 2) spatially aware neural layers that could be used for images, 3) the use of nonlinear activation functions to build deep neural networks, and 4) advanced computer infrastructure to collect and store data.
Dr. Lee gave some examples of what inspired his group to pursue studies in deep learning. He collected 5.5 million OCT (Optical Coherence Tomography) images from EMRs (Electronic Medical Records) representing 16,000 patients and over 10 years of time, all scanned with the same protocol. The images could be linked to other patient data including visual acuity, diagnosis, and OCT expert interpretation. One question they asked was could deep learning distinguish between normal versus AMD OCT images. A neural network was trained on about 100,000 images (VGG16). The results were quite surprising, showing a level of performance >92% at distinguishing between patients with disease and no disease. The algorithm was validated by testing it against OCTs that were not part of the “learning” exercise. Another project involved a DT-based, automated segmentation of macular edema in OCT to reveal sites of intraretinal fluid. The results were as good as clinical expert variability. Dr. Lee showed a movie that depicted the model going through the learning process. With repeated iterations, the model improved over time. Interestingly, the model recognized additional information (demarcation of different retinal layers) it needed to learn before it could determine the location of intraretinal fluid.
Other valuable information gained from the project was that machine learning is limited by ground truth, but it has incredible potential to go even beyond human performance. Another project using an animal model of oxygen-induced retinopathy to study angiogenesis was subjected to DL to see if segmentation results could be improved. The 2 quantifiable parameters were vaso-obliteration ratio and neovascularization ratio. Previously collected retinal images were subjected to DL. When segmentation images produced by humans were compared to images produced by DL, the latter did an amazing job of recapitulating the ratios. All of the work (the model, codes, etc.) was made available through open source and included a web site where other users could upload their images to get the same kind of segmentation data. Within 9 months there were 42 users and 2,142 images had been segmented. Bringing DL to the vision community will accelerate research. Another area of interest was taking OCT-B scans and looking for vessels. Rather than using human assistants to use OCT-B images to map the vessels, Dr. Lee collaborated with Dr. Ricky Wang was using OCT-A scans. Areas that were hyper reflective on OCT-A scans were areas of flow and hence blood vessels. The structural OCT scan was used as input and deep learning was used to try and recapitulate the OCT-A scan (ground truth). After 2 weeks of training, the DL model was able to pick up both large and small vessels that coincided perfectly with the OCT-A scan. An en face picture of the vessels generated by DL was almost equivalent to that generated by OCT-A in patients with central retinal artery occlusion as well as those with retinal vein occlusion. Dr. Lee’s group also hypothesized that flow could be deduced from structure by applying DL to flow B-scans, where speckle (high frequency, spatial) information was embedded in the OCT. The smaller (low flow) vessels did indeed disappear in the DL scans. These methods could be used to generate OCT-A type images from past OCT data sets. MRI (Magnetic Resonance Imaging) scans from the brain were subjected to DL to generate an MRA (Magnetic Resonance Angiogram) map without having to run the MRA scan. Unseen spatial features in conventional structural imaging may be picked up by DL algorithms and may be indicative of functions that have gone unnoticed by human inspection.
AI/DL was also used to forecast disease progression in glaucoma. Pairs of visual fields (VF) could be graded by clinicians to develop the ground truth of whether progression had occurred. Since the arrow of time is undeniable ground truth, the input became the VFs of >30,000 patients (time=T0) and the output was the VF over the following 20-year period. DL was applied to these sets of VFs under a strict protocol where 20% of the images were tested only at the end and the other 80% underwent a cross validation. However, after 5.5 years from T0 the model completely broke down. Early on however, the DL model accurately predicted the state of the VF, and a single input VF could be used to predict future VFs. Future work will incorporate IOP measures and surgical interventions.
The last example was a collaboration with Cynthia Owsley and Christine Curcio, who studied rod-mediated dark adaptation (RMDA). Delayed adaptation is a functional biomarker for age-related macular degeneration (AMD). OCT-B scans were sliced into tiny vertical windows and each subjected to DL to predict RMDA. The ability of the model to learn was different in different scanning locations. The result was a curve with 2 peaks that matched a double gaussian that Curcio had developed previously. It was also possible to determine where in the B-scan the model was looking for RMDA. The model was more interested in the hypo- and opposed the hyper-reflective bands.
Dr. Lee ended his talk with the importance of AI for eye care and vision science. DL models can help with hypothesis generation. DL models have already helped with the classification of diabetic retinopathy (DR) and the delivery of care. DL is non-invasive. Discrete data fields in electronic health records allows for rich annotation. To be able to push AI and DL to the next level, cross domain experts in medicine and data science need to be developed at the student or post-training level. More work is needed in data curation, standardization and sharing. Finally, very specialized computer hardware and infrastructure is needed.
Dr. Lee entertained questions. A Council member had 3 observations. DL could be applied to PET scanning to get functional data. Dl might also give a more accurate picture of DR (diabetic retinopathy) after treatment with anti-VEGF; fundus pictures get better, but the OCT-A images indicate that areas of non-perfusion persist. DL would give a better picture of post-treatment improvement. Photoceptor damage could also be tracked by looking at the shortening of their outer segments and applied to dark adaptometry. Dr. Lee agreed that all 3 areas were feasible. He also emphasized that many of the experiments he did failed, and the null hypothesis held true, but it was worth trying. One member commented that the presentation was brilliant! Another Council member mentioned that he had a number of capillary videos on patients with various pathologies, and AI did not appear to work well with videos and asked how they might proceed. Dr. Lee said that there is a family of algorithms meant to deal with videos. The videos could be broken down into frames to create individual training images while retaining patient level partitioning. The last Council member comment was regarding duplication of effort and cost across many different universities. He felt that the NEI could consider a consortium approach. Dr. Lee said that the work already involved collaboration, since no one institution had enough data to build a training set. He also mentioned that the data scientists he collaborated with were from industry or financial institutions where the work was not as rewarding as when they applied their knowledge to more meaningful questions in public health. However, these people were also hard to find, so a consortium approach would help with that issue.
NEI Small Business Grant Portfolio: From Application to Commercialization
Dr. Jerome Wujek, SBIR/STTR Program Officer, Division of Extramural Science Programs
Dr. Wujek began his presentation by noting that the small business grants are quite different from other grant mechanisms, primarily because they lead to commercialization of a product. The program provides seed money or early-stage capital to small, US businesses for commercialization of innovative, high-risk technologies, areas that venture or angel capitalists do not want to fund. There are 2 programs, Small Business Innovation Research (SBIR) and a smaller program for Small Business Technology Transfer (STTR). The STTR program allows transfer of new technologies from non-profit institutions to small businesses. Both programs are federally mandated and have been re-authorized through FY 2022. The NEI sets aside $21M for SBIRs and $3M for STTRs. The biggest difference is that the STTR requires a partnership between a small business and a research institution to stimulate technology transfer.
Dr. Wujek then focused on SBIRs. In terms of eligibility for a grant, the company 1) must be in the US and organized as for-profit, 2) have fewer than 500 employees, 3) all R&D (Research and Development) work must be done in the US, but the marketplace can be global, and 4) be under individual ownership. The goal is a new product, production method or a new service. Commercial success is defined as sales, licensing agreements or partnerships, mergers and acquisitions, and getting a product to clinical trials. Over 40 companies funded by the NEI have met the definition of success. Dr. Wujek gave several examples of these successes. A company called PlenOptika, Inc. developed an automated mobile tool for accurate, comprehensive refraction (a phoropter called QuickSee) with markets in India and the US. Another company, Biotigen, Inc., developed and sold several innovative OCT systems; that company was purchased by Leica Microsystems in 2015. Parion Sciences Inc. developed epithelia sodium channel blockers for a dry eye clinical trial.
Small businesses apply for a grant, which gets reviewed in special NIH study sections. If funded, the capital is non-dilutive, meaning that the funding agency gets no kickback, and the company can retain intellectual property rights. Both the SBIR and STTR programs have 3 Phases. Phase I is a feasibility study that establishes technical merit and shows commercial potential. Phase II involves full R&D to optimize design, efficacy and manufacturing. Phase III is the commercialization stage. The first 2 Phases have budget caps and time limits. The third phase is funded by the company. Dr. Wujek followed with several other commercial successes. Lumithera, Inc. developed a photobiomodulation device called Valeda that delivers specific wavelengths of light to patients with dry AMD. Clinical trials of the device have shown regression of drusen and improvement in visual acuity and contrast sensitivity by stabilizing cellular metabolic function and stimulating mitochondrial energy production. The device is sold both inside and outside the US and has been approved by the European Union. Mynosys Cellular Devices, Inc. developed a cutting device (Zepto precision Cataract Surgery System) to aid cataract surgery in pediatric cases; the device allows a much smaller circular incision to be made in the lens capsule.
There is a special grant for fast-tracking an idea that has high commercialization potential and involves simultaneous submission and review of both Phase I and II applications. Phase I is awarded first, and if the results are promising, Phase II is awarded quickly. An example of a fast-track SBIR was an award to Bioptigen for development of their intrasurgical OCT machine. Other examples were IrisVision LLC, a company that developed wearable, low vision glasses, and Rebiscam, Inc. that made a device for automated screening of strabismus in children.
Examples of innovative systems developed with SBIR grant funding to investigate alternatives to cataract surgery included the design and synthesis of small molecule chaperones by ViewPoint Therapeutics, Inc. to reverse amyloid deposition in the lens. Plex Pharmaceuticals, Inc. developed similar molecules that activated alpha-crystallin to stop and possibly reverse amyloid deposition.
The program has evolved since inception in 2006 to enhance all stages of the process to commercialization. A Phase II competing renewal for SBIRs has been added, as well as a direct to Phase II grant. A Commercial Readiness Pilot Program (CRP) is another NIH grant mechanism that funds the involvement of expertise needed to take a drug or device to a clinical trial. Applications are peer-reviewed and a CRP grant is paid out of NEI SBR funds. ONL Therapeutics Inc. developed a small peptide that inhibited retinal cell death after retinal detachment; this was important as there is a time delay between when the detachment occurs and surgical re-attachment. The company applied for the CRP in 2017 and is now in the midst of a Phase I clinical trial to test safety and efficacy of the drug.
There are programs and resources available at the NIH to assist SBIR/STTR grantees in making better technical decisions and minimizing risks, especially when anticipating FDA requirements. In 2006 there was the Niche Assessment Program to provide a company with market research information, and the Commercialization Assistance Program (CAP) to help with business development. Currently, the Niche Program is still available, as well as I-Corps at NIH. The latter is an entrepreneurial immersion course for scientists to identify a customer base and discern their needs. CAP is now called the Commercialization Accelerator Program and aids new companies to develop skills in commercialization and identify potential investors (Commercialization Track), to develop a new area (Advanced Track) and to assist businesses in navigating the regulatory path (Regulatory Track). CAP is free to small business, but only Phase II SBIR/STTR grantees are eligible. It is a very competitive program with only about 80 slots available NIH-wide. It is managed by a private sub-contractor to NIH. There are Entrepreneurs in Residence at the NIH who have been involved in small business and with venture capitalists who can advise grantees. There is also Investor Meeting Assistance that funds grantees to go to meetings or visit companies where investors are looking for opportunities in which to invest. Some of the meetings mentioned were BIO (Biotechnology Innovation Organization), ACA (Angel Capital Association), and Morgan Stanley. A final example of a commercialization success was VisionQuest Biomedical’s EyeStar. A hand-held non-mydriatic camera takes a fundus photograph that can be analyzed by a cloud computing AI algorithm to screen for diabetic retinopathy. EyeStar makes rapid, affordable and semi-automated screening available. The EyeStar is currently slated for testing in an FDA clinical trial. Looking to the future, the NIH will work to develop company tracking databases, improve due diligence prior to funding decisions and cooperating with other ICs. There is an NIH-wide office called SEED (Small Business Education and Entrepreneurial Development) that oversees the general SBIR/STTR programs. They provide information to program officers and grantees regarding small business opportunities and resources; they do outreach to investors, tech transfer organizations and state economic development agencies, and provide entrepreneurial training.
Dr. Wujek opened the floor for questions. One Council member how NEI’s commercial successes compare to other institutes. Dr. Wujek said that tracking is a hurdle that hasn’t been worked out, but the National Academy of Sciences has put out several reports indicating that the NIH program is doing well. Another member asked if it was important to have IP (intellectual property) protections in place before applying for an SBIR, and the answer was a decided yes.
General Council Discussion
Dr. Tumminia opened the floor for general Council discussion.
A member asked if there was a timeline for the proposed study section changes. Dr. Steinmetz said that no specific date had been announced. Another member questioned the disappearance of a training course that NEI had supported at Woods Hole and suggested the revival of grant mechanisms like R03s for young investigators. Dr. Steinmetz recognized the deep interest in training opportunities and answered that there are mechanisms like the R13 conference grant, individual (F and K) training grants and institutional training grants (T and K12). There may also be additional training opportunities coming out of the BRAIN Initiative. The NEI budget has increased for Fs and Ks as the NEI shifts its funding focus to more individual as opposed to institutional training grants. Results from both National Academy and NEI studies have shown that the success rate for subsequent R01 funding was much higher (actually double in the NEI study) for trainees who had been on an individual training grant versus an institutional training grant. He continued that the issue with R03s is that they have limited money and time, are not renewable, and are associated with high administrative costs. Dr. Schaffner reminded the group that the NEI would like to see young investigators submit an R01, due to the many advantages of time, money, and special review considerations. A Council member mentioned that she been involved in the Woods Hole course, which had unfortunately gotten caught up with administrative changes at Woods Hole; she has been attempting to reboot the course at Woods Hole or at a different venue. In her words, it did offer students a great opportunity to meet other students, learn science and new techniques in an enjoyable atmosphere. Dr. Tumminia mentioned a new, NIH-wide R01 award, the Steve Katz award, for at-risk early stage investigators who wish to change fields. No preliminary data are required. Another Council member was interested in a breakdown of grantee age in the various programs, to which Dr. Steinmetz answered that age, minority status and gender data were optional for grantees to list, that the OD has such data, but the NEI does not have access to it. A question was raised regarding whether the study section changes had been announced to the various societies that would be affected by it. Dr. Steinmetz replied that the presentation at today’s Council meeting was the first public announcement of the change. The member was of the opinion that the societies should be given the opportunity to comment on the proposed changes, receive a response, and be given an active role in identifying appropriate membership on those study sections. Dr. Steinmetz said that his greatest concern was the lack of expertise on the external panel, and the fact that at the CSR Advisory Council meeting the alternative position was not fairly represented. He was also concerned that twin study sections had been a failure in the past, and there were only 2 still in operation. The other move that had been a proven failure was taking an area of science and adding it as a minority group to another study section. The situation that triggered the creation of the SEP was NEI data that showed that anterior eye applications that were divided between 2 study sections, and represented about 15% of the applications, were not being treated fairly. One study section was treating those applications very negatively, creating a clear disparity in the scores. The issue was brought to the attention of CSR, and the “fix” was the creation of the anterior eye SEP. The internal panel was in 100% agreement regarding the issue and the remedy, even though they did not have applications that were reviewed in the vision study sections. A Council member pressed the point that more political pressure needed to be brought to bear on CSR, perhaps contacting a member of Congress. Dr. Tumminia reminded Council that an action of that type could not be brought by government employees but would need the involvement of the stakeholder community. A Council member suggested bringing the issue up at their respective business meetings, e.g., ARVO. Dr. Steinmetz suggested that since the issue had been brought only at the level of CSR, a letter of appeal could be drafted and sent to Michael Lauer (Director, OER) or Francis Collins (Director, NIH). It was agreed upon that Dr. Schaffner, with Council member Russ Van Gelder, would prepare such a letter and send it to Francis Collins. The final question was about working with other ICs, notably the NIA, NICHD, and NIGMS to obtain human tissue for research. Dr. Steinmetz made several comments. The NEI supports the NDRI (National Disease Research Interchange), and they are willing to work with researchers. ARVO has a current initiative with the Eye Banks to make ocular tissue available. The NIA offers administrative supplements to grantees to work on age-specific issues. The NEI has sent multiple requests for supplements to NIA, and they have funded every request. The Office of Women’s Health and the Office of Dietary Supplements at NIH have also co-funded NEI grantees. A Council member asked if an NEI-funded researcher who wanted to add an age-specific element to their grant would contact the NEI or the NIA. Dr. Steinmetz replied that the more general situation would be to contact NEI first unless it was in response to an FOA or NOSI from another IC. In that case, they should speak first to the contact person on the FOA or NOSI. In the case of Alzheimer-related supplements, the point of contact has been NEI PO Martha Flanders. Those comments ended the open session.
The Open Session was adjourned by Dr. Schaffner at 12:50 PM.
Attachment A: National Advisory Eye Council 2020
Eduardo C. Alfonso, M.D. (2020)
Jose-Manuel Alonso, M.D., Ph.D. (2021)
Katia Del Rio-Tsonis, Ph.D. (2022)
Thomas M. Glaser, M.D., Ph.D. (2019)
Mary Elizabeth Hartnett, M.D., Ph.D. (2022)
Dennis M. Levi, O.D., Ph.D. (2019)
Carol Ann Mason, Ph.D. (2020)
Louis R. Pasquale, M.D. (2019)
Sylvia B. Smith, PhD, FARVO (2019)
Mary Ann Stepp, Ph.D. (2021)
Benjamin Teller, O.D. (2022)
Russell Van Gelder, M.D., Ph.D. (2020)
Marco A. Zarbin, M.D., Ph.D.
DoD Representative position is vacant
Ad Hoc Members
Jane Gwiazda, Ph.D.