Topographic Clinical Insights From Deep Learning-Based Geographic Atrophy Progression Prediction.

Purpose: To explore the contributions of fundus autofluorescence (FAF) topographic imaging features to the performance of convolutional neural network-based deep learning (DL) algorithms in predicting geographic atrophy (GA) growth rate. Methods: Retrospective study with data from study eyes from three clinical trials (NCT02247479, NCT02247531, NCT02479386) in GA. The algorithm was initially trained with full FAF images, and its performance was considered benchmark. Ablation experiments investigated the contribution of imaging features to the performance of the algorithms. Three FAF image regions were defined relative to GA: Lesion, Rim, and Background. For No Lesion, No Rim, and No Background datasets, a single region of interest was removed at a time. For Lesion, Rim, and Background Shuffled datasets, individual region pixels were randomly shuffled. For Lesion, Rim, and Background Mask datasets, masks of the regions were used. A Convex Hull dataset was generated to evaluate the importance of lesion size. Squared Pearson correlation (r2) was used to compare the predictive performance of ablated datasets relative to the benchmark. Results: The Rim region influenced r2 more than the other two regions in all experiments, indicating the most relevant contribution of this region to the performance of the algorithms. In addition, similar performance was observed for all regions when pixels were shuffled or only a mask was used, indicating intensity information was not independently informative without textural context. Conclusions: These ablation experiments enabled topographic clinical insights on FAF images from a DL-based GA progression prediction algorithm. Translational Relevance: Results from this study may lead to new insights on GA progression prediction.

Deep Learning Approaches for Detecting of Nascent Geographic Atrophy in Age-Related Macular Degeneration.

Purpose: Nascent geographic atrophy (nGA) refers to specific features seen on OCT B-scans, which are strongly associated with the future development of geographic atrophy (GA). This study sought to develop a deep learning model to screen OCT B-scans for nGA that warrant further manual review (an artificial intelligence [AI]-assisted approach), and to determine the extent of reduction in OCT B-scan load requiring manual review while maintaining near-perfect nGA detection performance. Design: Development and evaluation of a deep learning model. Participants: One thousand eight hundred and eighty four OCT volume scans (49 B-scans per volume) without neovascular age-related macular degeneration from 280 eyes of 140 participants with bilateral large drusen at baseline, seen at 6-monthly intervals up to a 36-month period (from which 40 eyes developed nGA). Methods: OCT volume and B-scans were labeled for the presence of nGA. Their presence at the volume scan level provided the ground truth for training a deep learning model to identify OCT B-scans that potentially showed nGA requiring manual review. Using a threshold that provided a sensitivity of 0.99, the B-scans identified were assigned the ground truth label with the AI-assisted approach. The performance of this approach for detecting nGA across all visits, or at the visit of nGA onset, was evaluated using fivefold cross-validation. Main Outcome Measures: Sensitivity for detecting nGA, and proportion of OCT B-scans requiring manual review. Results: The AI-assisted approach (utilizing outputs from the deep learning model to guide manual review) had a sensitivity of 0.97 (95% confidence interval [CI] = 0.93-1.00) and 0.95 (95% CI = 0.87-1.00) for detecting nGA across all visits and at the visit of nGA onset, respectively, when requiring manual review of only 2.7% and 1.9% of selected OCT B-scans, respectively. Conclusions: A deep learning model could be used to enable near-perfect detection of nGA onset while reducing the number of OCT B-scans requiring manual review by over 50-fold. This AI-assisted approach shows promise for substantially reducing the current burden of manual review of OCT B-scans to detect this crucial feature that portends future development of GA. Financial Disclosures: Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Geographic Atrophy Segmentation Using Multimodal Deep Learning.

Purpose: To examine deep learning (DL)-based methods for accurate segmentation of geographic atrophy (GA) lesions using fundus autofluorescence (FAF) and near-infrared (NIR) images. Methods: This retrospective analysis utilized imaging data from study eyes of patients enrolled in Proxima A and B (NCT02479386; NCT02399072) natural history studies of GA. Two multimodal DL networks (UNet and YNet) were used to automatically segment GA lesions on FAF; segmentation accuracy was compared with annotations by experienced graders. The training data set comprised 940 image pairs (FAF and NIR) from 183 patients in Proxima B; the test data set comprised 497 image pairs from 154 patients in Proxima A. Dice coefficient scores, Bland-Altman plots, and Pearson correlation coefficient (r) were used to assess performance. Results: On the test set, Dice scores for the DL network to grader comparison ranged from 0.89 to 0.92 for screening visit; Dice score between graders was 0.94. GA lesion area correlations (r) for YNet versus grader, UNet versus grader, and between graders were 0.981, 0.959, and 0.995, respectively. Longitudinal GA lesion area enlargement correlations (r) for screening to 12 months (n = 53) were lower (0.741, 0.622, and 0.890, respectively) compared with the cross-sectional results at screening. Longitudinal correlations (r) from screening to 6 months (n = 77) were even lower (0.294, 0.248, and 0.686, respectively). Conclusions: Multimodal DL networks to segment GA lesions can produce accurate results comparable with expert graders. Translational Relevance: DL-based tools may support efficient and individualized assessment of patients with GA in clinical research and practice.

Interreader Agreement and Longitudinal Progression of Incomplete/Complete Retinal Pigment Epithelium and Outer Retinal Atrophy in Age-Related Macular Degeneration.

OBJECTIVE: To analyze the ability to evaluate changes over time of individual lesions of incomplete or complete retinal pigment epithelium (RPE) and outer retinal atrophy (iRORA and cRORA, respectively) in patients with intermediate age-related macular degeneration (iAMD). DESIGN: OCT images from patients enrolled in Proxima B clinical trial (NCT02399072) were utilized. PARTICIPANTS: Patients enrolled in the Proxima B clinical trial, from the cohort with geographic atrophy (GA) in 1 eye and iAMD in the other eye at baseline, were included. METHODS: Junior and senior readers analyzed OCT images for the qualitative presence of 9 distinct early atrophic features (presence of zone of choroidal hypertransmission, attenuation and/or disruption of RPE, disruption of ellipsoid zone [EZ] and external limiting membrane [ELM], outer nuclear layer [ONL] thinning, outer plexiform layer [OPL]/inner nuclear layer [INL] subsidence, and hyporeflective wedge-shaped band). If deemed "present," 7 features were quantified with a predefined tolerance level of 50 µm (diameter for the zone of choroidal hypertransmission, zone of attenuation and/or disruption of the RPE, outer retinal thickness left/right vertical diameter, outer retinal thickness thinnest vertical diameter, annotation of EZ, and ELM disruption). MAIN OUTCOME MEASURES: Interreader agreements for qualitative assessments (κ-type statistics) and quantitative measurements (Bland-Altman statistics) were assessed. Progression of the lesion features over time was described. RESULTS: Moderate agreement was found for presence of choroidal hypertransmission (κ = 0.54), followed by ELM disruption (κ = 0.58), OPL/INL subsidence (κ = 0.46), and a hyporeflective wedge-shaped band (κ = 0.47). Quantification measurements showed that choroidal hypertransmission had the highest agreement, whereas RPE attenuation/disruption had the lowest agreement. Longitudinal adjudicated changes for quantitative measurements of lesion progression showed that choroidal hypertransmission and ELM disruption showed significant progression, whereas EZ disruption and RPE attenuation/disruption did not. CONCLUSIONS: The ability to evaluate changes over time for specific features of iRORA and cRORA was explored. The most robust biomarker was found to be choroidal hypertransmission, followed by ELM disruption and the qualitative markers of OPL/INL subsidence, as well as a wedge-shaped band. Disease progression over time could be assessed by some, but not all, spectral-domain OCT features that were explored. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

The Role of the Complement Pathway in Clinical Progression of Geographic Atrophy: Analysis of the Phase III Chroma and Spectri Trials.

Purpose: To investigate the relationship between complement pathway activities and progression of geographic atrophy (GA) secondary to age-related macular degeneration in samples collected from patients enrolled in the Chroma and Spectri trials. Design: Chroma and Spectri were phase III, double-masked, and sham-controlled, 96-week trials. Participants: Aqueous humor (AH) samples collected at baseline and week 24 visits from 81 patients with bilateral GA across all 3 treatment groups (intravitreal lampalizumab 10 mg every 6 weeks, every 4 weeks, or corresponding sham procedures) were tested, along with patient-matched plasma samples collected at baseline. Methods: Antibody capture assays using the Simoa platform were used to measure the levels of complement factor B, the Bb fragment of complement factor B, intact complement component 3 (C3), processed C3, intact complement C4, and processed C4. Complement factor D levels were measured using enzyme-linked immunosorbent assay. Main Outcome Measures: Correlations of complement levels and activities (i.e., processed:intact ratio of complement component) in AH and plasma with baseline GA lesion size and growth rate. Results: In baseline AH, there were strong correlations (Spearman's rho ≥ 0.80) between intact complement proteins, between processed complement proteins, and between linked processed and intact complement proteins; weak correlations (rho ≤ 0.24) were found between complement pathway activities. There were no strong correlations between complement protein levels and activities measured in AH and plasma at baseline (rho ≤ 0.37). Baseline complement levels and activities in AH and plasma did not correlate with baseline GA lesion size or change from baseline in GA lesion area at week 48 (i.e., annualized growth rate). There were no strong correlations between changes in complement levels/activities in the AH from baseline to week 24 and annualized GA lesion growth rate. Genotype analysis did not reveal a meaningful correlation between complement-related, age-related macular degeneration risk single-nucleotide polymorphisms and complement levels and activities. Conclusions: Complement levels or activities in AH and plasma did not correlate with GA lesion size or growth rate. This suggests that local complement activation as measured in AH does not appear to be related to GA lesion progression. Financial Disclosures: Proprietary or commercial disclosure may be found after the references.

Alternative Complement Pathway Inhibition by Lampalizumab: Analysis of Data From Chroma and Spectri Phase III Clinical Trials.

Purpose: Lampalizumab, an antigen-binding fragment of a humanized monoclonal antibody directed against complement factor D (CFD), is designed to treat geographic atrophy (GA) secondary to age-related macular degeneration. Given the lack of clinical efficacy observed in patients with GA in the phase III Chroma/Spectri trials, we investigated the impact of lampalizumab on the complement system in vivo. We developed 6 novel assays to measure changes in complement pathway activities in aqueous humor samples collected from patients enrolled in these trials. Design: Chroma/Spectri were double-masked, sham-controlled, 96-week trials. Participants: Aqueous humor samples from 97 patients with bilateral GA across all groups (i.e., intravitreous lampalizumab 10 mg every 6 weeks, every 4 weeks, or corresponding sham procedures) were tested. Methods: Novel antibody capture assays were developed on the Simoa platform for complement factor B (CFB), the Bb fragment of CFB, intact complement component 3 (C3), processed C3, intact complement component 4 (C4), and processed C4. Main Outcome Measures: The ratio of processed vs. intact complement factors (i.e., complement activity) in aqueous humor were assessed. Results: Patients treated with either of the lampalizumab regimens demonstrated an increase in CFD level at week 24 compared with baseline, along with a corresponding median reduction in the Bb:CFB ratio of 41% to 43%. There were no strong correlations between lampalizumab concentrations in aqueous humor and change in CFD levels or Bb:CFB ratio over time. No change in downstream C3 processing was observed with lampalizumab treatment. Additionally, there was no change in C4 processing. Conclusions: The collection of aqueous humor samples from patients in Chroma and Spectri trials provided key insights on the effects of lampalizumab, a novel complement inhibitor, on local ocular complement activation. Lampalizumab inhibited the alternative complement pathway in the eyes of patients with GA; however, this did not translate into a measurable reduction in either classical or total complement activity, based on absence of changes in C4 and C3 processing, respectively. Financial Disclosures: Proprietary or commercial disclosure may be found after the references.

Deep Learning to Predict Geographic Atrophy Area and Growth Rate from Multimodal Imaging.

OBJECTIVE: To develop deep learning models for annualized geographic atrophy (GA) growth rate prediction using fundus autofluorescence (FAF) images and spectral-domain OCT volumes from baseline visits, which can be used for prognostic covariate adjustment to increase power of clinical trials. DESIGN: This retrospective analysis estimated GA growth rate as the slope of a linear fit on all available measurements of lesion area over a 2-year period. Three multitask deep learning models-FAF-only, OCT-only, and multimodal (FAF and OCT)-were developed to predict concurrent GA area and annualized growth rate. PARTICIPANTS: Patients were from prospective and observational lampalizumab clinical trials. METHODS: The 3 models were trained on the development data set, tested on the holdout set, and further evaluated on the independent test sets. Baseline FAF images and OCT volumes from study eyes of patients with bilateral GA (NCT02247479; NCT02247531; and NCT02479386) were split into development (1279 patients/eyes) and holdout (443 patients/eyes) sets. Baseline FAF images from study eyes of NCT01229215 (106 patients/eyes) and NCT02399072 (169 patients/eyes) were used as independent test sets. MAIN OUTCOME MEASURES: Model performance was evaluated using squared Pearson correlation coefficient (r2) between observed and predicted lesion areas/growth rates. Confidence intervals were calculated by bootstrap resampling (B = 10 000). RESULTS: On the holdout data set, r2 (95% confidence interval) of the FAF-only, OCT-only, and multimodal models for GA lesion area prediction was 0.96 (0.95-0.97), 0.91 (0.87-0.95), and 0.94 (0.92-0.96), respectively, and for GA growth rate prediction was 0.48 (0.41-0.55), 0.36 (0.29-0.43), and 0.47 (0.40-0.54), respectively. On the 2 independent test sets, r2 of the FAF-only model for GA lesion area was 0.98 (0.97-0.99) and 0.95 (0.93-0.96), and for GA growth rate was 0.65 (0.52-0.75) and 0.47 (0.34-0.60). CONCLUSIONS: We show the feasibility of using baseline FAF images and OCT volumes to predict individual GA area and growth rates using a multitask deep learning approach. The deep learning-based growth rate predictions could be used for covariate adjustment to increase power of clinical trials. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found after the references.

Dose-dependent inactivation of airway tryptase with a novel dissociating anti-tryptase antibody (MTPS9579A) in healthy participants: A randomized trial.

Tryptase is the most abundant secretory granule protein in human lung mast cells and plays an important role in asthma pathogenesis. MTPS9579A is a novel monoclonal antibody that selectively inhibits tryptase activity by dissociating active tetramers into inactive monomers. The safety, tolerability, pharmacokinetics (PKs), and systemic and airway pharmacodynamics (PDs) of MTPS9579A were assessed in healthy participants. In this phase I single-center, randomized, observer-blinded, and placebo-controlled study, single and multiple ascending doses of MTPS9579A were administered subcutaneously (s.c.) or intravenously (i.v.) in healthy participants. In addition to monitoring safety and tolerability, the concentrations of MTPS9579A, total tryptase, and active tryptase were quantified. This study included 106 healthy participants (82 on active treatment). Overall, MTPS9579A was well-tolerated with no serious or severe adverse events. Serum MTPS9579A showed a dose-proportional increase in maximum serum concentration (Cmax ) values at high doses, and a nonlinear increase in area under the curve (AUC) values at low concentrations consistent with target-mediated clearance were observed. Rapid and dose-dependent reduction in nasosorption active tryptase was observed postdose, confirming activity and the PK/PD relationship of MTPS9579A in the airway. A novel biomarker assay was used to demonstrate for the first time that an investigative antibody therapeutic (MTPS9579A) can inhibit tryptase activity in the upper airway. A favorable safety and tolerability profile supports further assessment of MTPS9579A in asthma. Understanding the exposure-response relationships using the novel PD biomarker will help inform clinical development, such as dose selection or defining patient subgroups.

Conversion from Intermediate Age-Related Macular Degeneration to Geographic Atrophy in a Proxima B Subcohort Using a Multimodal Approach.

INTRODUCTION: This retrospective analysis assessed geographic atrophy (GA) progression in fellow eyes from the Proxima B trial intermediate age-related macular degeneration (iAMD) subcohort using high-resolution multimodal imaging anchored on optical coherence tomography (OCT). METHODS: Thirty-two patients from the Proxima B iAMD subcohort were assessed; all had GA with no macular neovascularization (MNV) in the contralateral eye. Imaging data, including color fundus photography, fluorescein angiography, near-infrared reflectance, fundus autofluorescence (FAF), and spectral-domain OCT, were obtained. Features preceding progression/conversion to advanced AMD (drusen, reticular pseudodrusen [RPD], MNV, incomplete/complete retinal pigment epithelium and outer retinal atrophy [iRORA/cRORA]) were assessed. RESULTS: Of 30 fellow eyes with available follow-up images, 12 converted to GA (FAF), 2 converted to MNV, and 16 were nonconverters during the review period (median: 17.8 months). iRORA/cRORA features (present in all converters at baseline) were identified on OCT in eyes that progressed to GA. Median time interval from iRORA to cRORA and from cRORA to GA was 7 months each. GA development/progression was either drusen- or RPD-associated (n = 6 each). Eyes with baseline RPD showed faster GA progression versus eyes with drusen (1.49 vs. 0.38 mm2/year). CONCLUSIONS: RPD presence was associated with rapid GA lesion enlargement and may provide an early indication of faster GA progression.

Survival forests under test: Impact of the proportional hazards assumption on prognostic and predictive forests for amyotrophic lateral sclerosis survival.

We investigate the effect of the proportional hazards assumption on prognostic and predictive models of the survival time of patients suffering from amyotrophic lateral sclerosis. We theoretically compare the underlying model formulations of several variants of survival forests and implementations thereof, including random forests for survival, conditional inference forests, Ranger, and survival forests with L1 splitting, with two novel variants, namely distributional and transformation survival forests. Theoretical considerations explain the low power of log-rank-based splitting in detecting patterns in non-proportional hazards situations in survival trees and corresponding forests. This limitation can potentially be overcome by the alternative split procedures suggested herein. We empirically investigated this effect using simulation experiments and a re-analysis of the Pooled Resource Open-Access ALS Clinical Trials database of amyotrophic lateral sclerosis survival, giving special emphasis to both prognostic and predictive models.

Target engagement in an alzheimer trial: Crenezumab lowers amyloid ß oligomers in cerebrospinal fluid.

OBJECTIVE: Oligomeric forms of amyloid ß protein (oAß) are believed to be principally responsible for neurotoxicity in Alzheimer disease (AD), but it is not known whether anti-Aß antibodies are capable of lowering oAß levels in humans. METHODS: We developed an ultrasensitive immunoassay and used it to measure oAß in cerebrospinal fluid (CSF) from 104 AD subjects participating in the ABBY and BLAZE phase 2 trials of the anti-Aß antibody crenezumab. Patients received subcutaneous (SC) crenezumab (300mg) or placebo every 2 weeks, or intravenous (IV) crenezumab (15mg/kg) or placebo every 4 weeks for 68 weeks. Ninety-eight of the 104 patients had measurable baseline oAß levels, and these were compared to levels at week 69 in placebo (n = 28), SC (n = 35), and IV (n = 35) treated patients. RESULTS: Among those receiving crenezumab, 89% of SC and 86% of IV patients had lower levels of oAß at week 69 versus baseline. The difference in the proportion of patients with decreasing levels was significant for both treatment arms: p = 0.0035 for SC and p = 0.01 for IV crenezumab versus placebo. The median percentage change was -48% in the SC arm and -43% in the IV arm. No systematic change was observed in the placebo group, with a median change of -13% and equivalent portions with negative and positive change. INTERPRETATION: Crenezumab lowered CSF oAß levels in the large majority of treated patients tested. These results support engagement of the principal pathobiological target in AD and identify CSF oAß as a novel pharmacodynamic biomarker for use in trials of anti-Aß agents. ANN NEUROL 2019;86:215-224.