Closing the loop for AI-ready radiology.

BACKGROUND: In recent years, AI has made significant advancements in medical diagnosis and prognosis. However, the incorporation of AI into clinical practice is still challenging and under-appreciated. We aim to demonstrate a possible vertical integration approach to close the loop for AI-ready radiology. METHOD: This study highlights the importance of two-way communication for AI-assisted radiology. As a key part of the methodology, it demonstrates the integration of AI systems into clinical practice with structured reports and AI visualization, giving more insight into the AI system. By integrating cooperative lifelong learning into the AI system, we ensure the long-term effectiveness of the AI system, while keeping the radiologist in the loop.  RESULTS: We demonstrate the use of lifelong learning for AI systems by incorporating AI visualization and structured reports. We evaluate Memory Aware-Synapses and Rehearsal approach and find that both approaches work in practice. Furthermore, we see the advantage of lifelong learning algorithms that do not require the storing or maintaining of samples from previous datasets. CONCLUSION: In conclusion, incorporating AI into the clinical routine of radiology requires a two-way communication approach and seamless integration of the AI system, which we achieve with structured reports and visualization of the insight gained by the model. Closing the loop for radiology leads to successful integration, enabling lifelong learning for the AI system, which is crucial for sustainable long-term performance. KEY POINTS: · The integration of AI systems into the clinical routine with structured reports and AI visualization.. · Two-way communication between AI and radiologists is necessary to enable AI that keeps the radiologist in the loop.. · Closing the loop enables lifelong learning, which is crucial for long-term, high-performing AI in radiology..

Opportunity Cost of Vitreoretinal Surgeries.

Purpose: To compare physician reimbursements for vitreoretinal surgeries with office-based patient care. Methods: A theoretical model was performed comparing physician work reimbursements for the 10 most common vitreoretinal surgeries with office-based work relative value units (wRVUs) that could have been generated during the same global time period. The reference physician was modeled at 40 patients per 8-hour workday. A lower volume physician and higher volume physician were modeled at 30 patients/day and 50 patients/day, respectively. The reimbursement rates and allocated times for surgery were based on the 2021 values set by Medicare, and the average wRVU per office visit was based on 2021 real-world data from the Vestrum Retinal Healthcare Database. Results: In the reference case, performing any of the 10 most common vitreoretinal surgeries was associated with an opportunity cost with a weighted mean of 49% (range, 40%-68%) relative to lost office productivity. The Centers for Medicare & Medicaid Services (CMS) allocated a weighted mean intraservice time of 73 minutes; however, the reference physician would have to complete the surgery with a weighted average of 5 minutes (range, -31-12 minutes) for surgical wRVUs to equal office-based reimbursements. Performing these 10 surgeries was associated with a 25% opportunity cost even for the lower volume physician and 61% for the higher volume physician. Probability sensitivity analysis with a range of conditions identified opportunity costs from surgery in over 99% of simulated scenarios. Conclusions: Medicare reimbursements for the physician work component of vitreoretinal surgeries represented a significant opportunity cost for the physician relative to office-based patient care of equivalent time, especially for busier physicians. The model did not explore practice overhead and professional liability insurance, which are factored separately by CMS and may influence the opportunity cost depending on utilization. The average threshold surgery times for surgical reimbursements to equal office-based reimbursements may be difficult to achieve.

Evolving Treatment Patterns in Diabetic Macular Edema Between 2015 and 2020.

Purpose: To explore the recent evolution of diabetic macular edema (DME) treatment practice patterns over 5 years among retina specialists in the United States. Methods: This retrospective analysis assessed 306 700 eyes with newly diagnosed DME from the Vestrum Health database between January 2015 and October 2020. The year-over-year and cumulative 5-year distributions of eyes treated with antivascular endothelial growth factor (anti-VEGF) agents, steroids, focal laser, or any combination and those of untreated eyes were calculated. Changes from baseline visual acuity were assessed. Results: Yearly treatment patterns changed significantly from 2015 (n = 18056) to 2020 (n = 11042). The proportion of untreated patients declined over time (32.7% vs 27.7%; P < .001), the use of anti-VEGF monotherapy increased (43.5% vs 61.8%; P < .001), the use of focal laser monotherapy declined (9.7% vs 3.0%; P < .001), and the use of steroid monotherapy remained stable (0.9% vs 0.7%; P = 1.000). Of eyes that maintained follow-up for 5 years (from 2015 to 2020), 16.3% were untreated while 77.5% were treated with anti-VEGF agents (as monotherapy or combination therapy). Vision gains in treated patients remained approximately stable from 2015 (3.6 letters) to 2020 (3.5 letters). Conclusions: From 2015 to 2020, treatment patterns for DME evolved toward greater anti-VEGF monotherapy, stable steroid monotherapy, less laser monotherapy, and fewer untreated eyes.

The Broader Economic Value of Treatment for Diabetic Macular Edema.

OBJECTIVE: To simulate economic outcomes for individuals with diabetic macular edema (DME) and estimate the economic value of direct and indirect benefits associated with DME treatment. RESEARCH DESIGN AND METHODS: Our study pairs individual and cohort analyses to demonstrate the value of treatment for DME. We used a microsimulation model to simulate self-reported vision (SRV) and economic outcomes for individuals with DME. Four scenarios derived from clinical trial data were simulated and compared for a lifetime horizon: untreated, anti-VEGF therapy, laser, and steroid. To quantify the relative magnitude of costs and benefits of DME treatment in the U.S., we used a cohort-level analysis based on real-world treatment parameters derived from published data. RESULTS: In the model, excellent/good SRV roughly corresponded to 20/40 or better visual acuity. A representative 51-year-old treated for DME would spend 30-35% additional years with excellent/good SRV and 29-32% fewer years with fair/poor SRV relative to being untreated. A treated individual would experience 4-5% greater life expectancy and 9-13% more quality-adjusted life-years. Indirect benefits from treatment included 6-9% more years working, 12-19% greater lifetime earnings, and 8-16% fewer years with disability. For the U.S. DME cohort (1.1. million people), total direct benefit was $63.0 billion over 20 years, and total indirect benefit was $4.8 billion. Net value (benefit - cost) of treatment ranged from $28.1 billion to $52.8 billion. CONCLUSIONS: Treatment for DME provides economic value to patients and society through improved vision, life expectancy, and quality of life and indirectly through improved employment and disability outcomes.

Cohort-Specific Serological Recognition of SARS-CoV-2 Variant RBD Antigens.

OBJECTIVE: Estimating the response of different population cohorts to new SARS-CoV-2 variants is important to customize measures of control. Our goal was to evaluate how antibodies from sera of infected and vaccinated people recognize antigens expressed by different SARS-CoV-2 variants. METHODS: We compared sera from vaccinated donors and four patient/donor cohorts: Sera from critically ill patients collected 2-7 days and more than 10 days after admission to an intensive care unit, a NIBSC/WHO reference panel of SARS-CoV-2 positive individuals, and ambulatory or hospitalized (but not critically ill) positive donors. Samples were tested with an anti-SARS-CoV-2 ELISA kit coated with SARS-CoV-2 RBD recombinant antigens including mutations present in eleven of the most widespread variants. RESULTS: Sera from vaccinated individuals exhibited higher antibody binding (P<0.001) than sera from infected (but not critically ill) individuals when tested against the wild type (WT) and each of 11 variants' receptor binding domain (RBD). Antibodies' binding to the SARS-CoV-2 antigens of at least 6 variants, including Variants of Concern (VOCs), was reduced in comparison to the WT in vaccinated and non-critically ill convalescence individuals. CONCLUSION: Understanding differences between population cohorts in the antibody titers against WT vs variant RBD antigens can help design variant-specific immunoassays for surveillance and evaluation of the epidemiology of new variants.

The Impact of COVID-19 on US and International Retina Specialists, Their Practices, and Their Patients.

Purpose: This work analyzes data from a series of surveys developed by the American Society of Retina Specialists (ASRS) that assesses the impact of COVID-19 on physicians, their practices, and their patients. Methods: Five surveys were sent by the ASRS between March and July 2020 to more than 2600 US and international retina specialists. Data and trends from these surveys were analyzed. Results: Most responding retina specialists (87%-95% in the United States and internationally) reported having no known COVID-related symptoms despite reported limitations in personal protective equipment. Clinic volumes globally were drastically reduced in March 2020 with only partial recovery through July 2020, which was slower internationally than in the United States. Practices were compelled to reduce staff and physician employment levels. Most respondents estimated some degree of delay in patient treatment with corresponding declines in vision and/or anatomy that were attributed most frequently to patients' fears of the pandemic and least frequently to office unavailability. Conclusions: The reported impact of COVID-19 on retina specialists, their practices, and their patients has been substantial. Although retina specialists were quickly resilient in optimizing delivery of patient care in a manner safe for patients and providers, the reduction in clinic volume has been devastating in the United States and internationally, with negative impacts on patient outcomes, reductions in practices' volume and employment, and risk to practices' financial health. Future studies will be required to quantify losses associated with these unprecedented and ongoing circumstances caused by the pandemic.

Opportunity Cost of Retinal Detachment Surgery vs Office-Based Patient Care.

Purpose: This work compares physician reimbursements for retinal detachment (RD) surgery with office-based patient care. Methods: A theoretical model was constructed from the physician's perspective for performing a 90-minute uncomplicated RD surgery with its associated perioperative work in the global period (Current Procedural Terminology code 67108) compared with managing 40 patients per 8-hour clinic day in the equivalent time period. The reimbursement rates were based on the 2019 values set by the US Centers for Medicare and Medicaid Services (CMS). Sensitivity analyses were performed varying the perioperative times, clinical productivity, and postoperative visits. Results: The CMS physician reimbursement rate for 67108 surgery was 17.13 work relative value units (wRVUs); meanwhile, the physician in the reference case could have generated 40.89 wRVUs in the office. CMS reimbursement therefore represented a 58% opportunity cost relative to lost office productivity for the physician. A significant disparity was still present even when modeling 30 patients per day. In sensitivity analyses, clinical productivity exceeded surgical compensation in 99% of modeled scenarios. In threshold analyses, the surgeon in the reference case would have to complete the surgery and all immediate perioperative care within 18 minutes to equal the total CMS valuation. Conclusions: CMS reimbursement for RD surgery resulted in a significant opportunity cost for the physician relative to office-based patient care, which was more pronounced for more efficient clinicians in the office. The sensitivity analyses supported the robustness of the model. Reductions in surgery reimbursements relative to office-based patient care might disincentivize busy clinicians.

Brolucizumab-associated intraocular inflammation in eyes without retinal vasculitis.

PURPOSE: To analyze a series of eyes with brolucizumab-associated intraocular inflammation (IOI) without retinal vasculitis reported to the American Society of Retina Specialists (ASRS). METHODS: The ASRS Research and Safety in Therapeutics (ReST) Committee analyzed clinical characteristics from submitted reports of IOI after brolucizumab. Eyes with retinal vasculitis or that received intraocular antibiotics were excluded. RESULTS: Forty-nine eyes of 45 patients were collected. Mean visual acuity (VA) at baseline was 20/49 (range 20/20 - 5/200). Patients presented with IOI a mean of 24 (range 3-63) days after most recent brolucizumab injection; 61% presented for an unscheduled visit while 39% presented at routine follow-up. Mean VA at IOI presentation was 20/67 (range 20/20 - 3/200). Most common symptoms were floaters (78%) and blurry vision (76%). Pain (20%) and redness (16%) were less common; 3 (6%) eyes were asymptomatic. IOI was anterior only in 18%, posterior only in 31%, and both anterior and posterior in 51% of eyes. Treatment included topical steroids alone in 67% eyes, while 10% eyes received no treatment. Mean VA at last follow-up was 20/56 (range 20/20 - 1/200). Three (6%) eyes lost 3 or more lines and 1 (2%) eye lost 6 or more lines. CONCLUSIONS: Brolucizumab-associated IOI without retinal vasculitis typically presented with a delayed onset of a few weeks. Often, visual acuity decline was relatively mild. Most symptoms resolved and nearly all had a return to baseline VA, but a small percentage of patients had a significant decrease in VA at last follow-up.

Occlusive Retinal Vasculitis Following Intravitreal Brolucizumab.

PURPOSE: To analyze a case series of retinal vasculitis reported to the American Society of Retina Specialists (ASRS) following Food and Drug Administration approval of brolucizumab for treatment of neovascular age-related macular degeneration. METHODS: The ASRS Research and Safety in Therapeutics Committee analyzed clinical and imaging characteristics from submitted reports of retinal vasculitis after brolucizumab. RESULTS: Retinal vasculitis was reported in 26 eyes of 25 patients (22 [88%] female) after treatment with brolucizumab. Imaging studies were available for 24 of 26 eyes. Most cases (92%) were associated with intraocular inflammation, which presented at a mean of 25 days (range, 3-63 days) after the most recent brolucizumab injection. Mean visual acuity (VA) was 20/52 (range, 20/25-4/200) before the adverse event, 20/151 (range, 20/25-hand motion) at presentation of the adverse event, and 20/243 (range, 20/30-light perception) at last follow-up. Twelve eyes (46%) had a greater than 3-line decrease in VA at final follow-up, and 12 eyes (46%) had a final VA of 20/200 or worse. Analysis of retinal imaging identified vasculopathy that involved retinal arteries (91%), retinal veins (79%), and choroidal vessels (48%). Occlusive disease was apparent on imaging in 83% of eyes. Treatment approaches were varied. CONCLUSIONS: Retinal vasculitis has been identified in a series of eyes following brolucizumab. Although a few eyes in this series were asymptomatic or minimally symptomatic, some eyes had significant vision loss. A careful examination for signs of active inflammation prior to brolucizumab injection is recommended. Once vasculopathy is suspected, angiographic imaging may help define the spectrum of involvement. Optimal treatment strategies remain unknown.

Orally Administered Alpha Lipoic Acid as a Treatment for Geographic Atrophy: A Randomized Clinical Trial.

PURPOSE: Alpha lipoic acid (ALA) is a nutraceutical and potent antioxidant that has shown efficacy in the retina light damage mouse model and in humans for multiple sclerosis. Our objective was to evaluate the efficacy and safety of oral ALA for the treatment of geographic atrophy (GA). DESIGN: Randomized, controlled, double-masked, multicenter phase 2 clinical trial of ALA versus placebo. PARTICIPANTS: Participants with unilateral or bilateral GA from age-related macular degeneration. METHODS: Participants were randomized to 1200 mg daily of ALA or placebo. Fundus autofluorescence, fundus color photography, and spectral-domain OCT were conducted and best-corrected visual acuity (BCVA) was obtained at baseline and every 6 months through month 18. MAIN OUTCOME MEASURES: Annual rate of change over 18 months in square root-transformed area of GA in study eyes as measured on fundus autofluorescence. Secondary outcomes included the number of adverse events (AEs), change in BCVA, and annual rate of change in area of GA measured on color photographs. RESULTS: Fifty-three participants (mean age, 80 years) were randomized (April 2016-August 2017). Twenty-seven participants (37 eyes) were in the placebo group, and 26 participants (36 eyes) were in the ALA group. Unadjusted mean (standard error) annual change in GA area was 0.28 (0.02) mm and 0.31 (0.02) mm for the placebo and ALA groups, respectively (difference, 0.04 mm; 95% confidence interval [CI], -0.03 to 0.11 mm; P = 0.30). Adjusting for baseline GA area, number of GA lesions, and presence of subfoveal GA, the mean annual change in GA area was 0.27 (0.04) mm and 0.32 (0.05) mm for the placebo and ALA groups, respectively (difference, 0.05 mm; 95% CI, -0.02 to 0.12 mm; P = 0.14). At 18 months, the percent of eyes losing 15 letters or more of BCVA was 22% (8 of 36) and 14% (5 of 36) in the placebo and ALA groups, respectively (P = 0.54). No difference was found in the percentage of participants with nonserious AEs (P = 0.96) or serious AEs (P = 0.28) between the placebo and ALA groups. CONCLUSIONS: Results do not support ALA having beneficial effects on GA or BCVA. This trial design may be useful for other GA repurposing drug trials.

Feeding-dependent VIP neuron-ILC3 circuit regulates the intestinal barrier.

The intestinal mucosa serves both as a conduit for the uptake of food-derived nutrients and microbiome-derived metabolites, and as a barrier that prevents tissue invasion by microorganisms and tempers inflammatory responses to the myriad contents of the lumen. How the intestine coordinates physiological and immune responses to food consumption to optimize nutrient uptake while maintaining barrier functions remains unclear. Here we show in mice how a gut neuronal signal triggered by food intake is integrated with intestinal antimicrobial and metabolic responses that are controlled by type-3 innate lymphoid cells (ILC3)1-3. Food consumption rapidly activates a population of enteric neurons that express vasoactive intestinal peptide (VIP)4. Projections of VIP-producing neurons (VIPergic neurons) in the lamina propria are in close proximity to clusters of ILC3 that selectively express VIP receptor type 2 (VIPR2; also known as VPAC2). Production of interleukin (IL)-22 by ILC3, which is upregulated by the presence of commensal microorganisms such as segmented filamentous bacteria5-7, is inhibited upon engagement of VIPR2. As a consequence, levels of antimicrobial peptide derived from epithelial cells are reduced but the expression of lipid-binding proteins and transporters is increased8. During food consumption, the activation of VIPergic neurons thus enhances the growth of segmented filamentous bacteria associated with the epithelium, and increases lipid absorption. Our results reveal a feeding- and circadian-regulated dynamic neuroimmune circuit in the intestine that promotes a trade-off between innate immune protection mediated by IL-22 and the efficiency of nutrient absorption. Modulation of this pathway may therefore be effective for enhancing resistance to enteropathogens2,3,9 and for the treatment of metabolic diseases.

ADVERSE EVENTS OF THE ARGUS II RETINAL PROSTHESIS: Incidence, Causes, and Best Practices for Managing and Preventing Conjunctival Erosion.

PURPOSE: To analyze and provide an overview of the incidence, management, and prevention of conjunctival erosion in Argus II clinical trial subjects and postapproval patients. METHODS: This retrospective analysis followed the results of 274 patients treated with the Argus II Retinal Prosthesis System between June 2007 and November 2017, including 30 subjects from the US and European clinical trials, and 244 patients in the postapproval phase. Results were gathered for incidence of a serious adverse event, incidence of conjunctival erosion, occurrence sites, rates of erosion, and erosion timing. RESULTS: Overall, 60% of subjects in the clinical trial subjects versus 83% of patients in the postapproval phase did not experience device- or surgery-related serious adverse events. In the postapproval phase, conjunctival erosion had an incidence rate of 6.2% over 5 years and 11 months. In 55% of conjunctival erosion cases, erosion occurred in the inferotemporal quadrant, 25% in the superotemporal quadrant, and 20% in both. Sixty percent of the erosion events occurred in the first 15 months after implantation, and 85% within the first 2.5 years. CONCLUSION: Reducing occurrence of conjunctival erosion in patients with the Argus II Retinal Prosthesis requires identification and minimization of risk factors before and during implantation. Implementing inverted sutures at the implant tabs, use of graft material at these locations as well as Mersilene rather than nylon sutures, and accurate Tenon's and conjunctiva closure are recommended for consideration in all patients.

OCTA: A Practical Method of Image Averaging Using Adobe Photoshop Software.

BACKGROUND AND OBJECTIVE: To describe a method for averaging ocular coherence tomography angiography (OCTA) images using a commercially available photo editing software: Adobe Photoshop CC 2017. PATIENTS AND METHODS: This single-center study assessed the feasibility of using Adobe Photoshop as an OCTA image averaging tool. Three 3.0 mm × 3.0 mm OCTA images from each eye were obtained using commercially available OCTA devices. Captured OCTA images were exported in high-resolution TIFF format, imported as an image stack, aligned using an automated function, and averaged by creating a Smart Object using Photoshop CC 2017 software. In conjunction with qualitative assessment, the main outcome of the study was image grader preferences with respect to clarity of the foveal avascular zone (FAZ), blood vessel delineation, and the ability to identify abnormal vasculature. RESULTS: After removing OCTA scans with significant image distortion, 25 sets of images were included in the analysis. Adobe Photoshop CC 2017 successfully aligned and averaged all images of the superficial and deep retinal plexuses that contained a minimum 40% overlap. Three independent retinal specialists found the averaged images to be slightly or definitely preferable to the original 87%, 89%, and 69% of the time with respect to clarity of the FAZ, clarity of blood vessel delineation, and the ability to identify abnormal vasculature, respectively. CONCLUSIONS: Adobe Photoshop CC 2017 is an excellent tool for image averaging, producing high-quality resulting OCTA images. As an easily accessible software, Photoshop has the potential for use in a diversity of pathological conditions. [Ophthalmic Surg Lasers Imaging Retina. 2019;50:802-807.].

Retinal Anatomy and Electrode Array Position in Retinitis Pigmentosa Patients After Argus II Implantation: An International Study.

PURPOSE: To assess the retinal anatomy and array position in Argus II retinal prosthesis recipients. DESIGN: Prospective, noncomparative cohort study. METHODS: Setting: International multicenter study. PATIENTS: Argus II recipients enrolled in the Post-Market Surveillance Studies. PROCEDURES: Spectral-domain optical coherence tomography images collected for the Surveillance Studies (NCT01860092 and NCT01490827) were reviewed. Baseline and postoperative macular thickness, electrode-retina distance (gap), optic disc-array overlap, and preretinal membrane presence were recorded at 1, 3, 6, and 12 months. MAIN OUTCOME MEASURES: Axial retinal thickness and axial gap along the array's long axis (a line between the tack and handle); maximal retinal thickness and maximal gap along a B-scan near the tack, midline, and handle. RESULTS: Thirty-three patients from 16 surgical sites in the United States and Germany were included. Mean axial retinal thickness increased from month 1 through month 12 at each location, but reached statistical significance only at the array midline (P = .007). The rate of maximal thickness increase was highest near the array midline (slope = 6.02, P = .004), compared to the tack (slope = 3.60, P < .001) or the handle (slope = 1.93, P = .368). The mean axial and maximal gaps decreased over the study period, and the mean maximal gap size decrease was significant at midline (P = .032). Optic disc-array overlap was seen in the minority of patients. Preretinal membranes were common before and after implantation. CONCLUSIONS: Progressive macular thickening under the array was common and corresponded to decreased electrode-retina gap over time. By month 12, the array was completely apposed to the macula in approximately half of the eyes.

Real-Time Volumetric Imaging of Vitreoretinal Surgery with a Prototype Microscope-Integrated Swept-Source OCT Device.

PURPOSE: To evaluate the images produced in an initial surgical series of intraoperative near-real-time volumetric swept-source (SS) OCT imaging. DESIGN: Prospective translational study. PARTICIPANTS: Forty-one consecutive adult patients undergoing vitreoretinal surgery between July 22, 2014, and July 1, 2015, at the Duke University Eye Center who agreed to participate. METHODS: A novel microscope-integrated SS-OCT prototype captured volumetric renderings of imaging of macular surgery in near-real-time and showed them to the surgeon via a heads-up display through the microscope oculars. Then the images were analyzed formally after surgery. MAIN OUTCOME MEASURES: Image quality, successful capture of surgical instruments, maneuvers and associated retinal deformation volumetrically over time, and qualitative image analysis. RESULTS: Volumetric SS-OCT images were graded as acceptable in 92% of patients. Volumetric imaging of scraping and peeling procedures was achieved in 75% and 78% of patients in whom it was performed, respectively. Imaging provided the surgeon with near-real-time volumetric visualization of the position of the instrument relative to the retinal surface, flap initiation, flap removal, and retinal deformation during instrumentation via a heads-up display. CONCLUSIONS: This volumetric, microscope-integrated SS-OCT prototype seems to provide high-detail, near-real-time volumetric imaging of delicate maneuvers during macular surgery.