No abstract

Abstract. Obtaining models that capture imaging markers relevant for disease progression and treatment monitoring is challenging. Models are typically based on large amounts of data with annotated examples of known markers aiming at automating detection. High annotation effort and the limitation to a vocabulary of known markers limit the power of such approaches. Here, we perform unsupervised learning to identify anomalies in imaging data as candidates for markers. We propose AnoGAN, a deep convolutional generative adversarial network to learn a manifold of normal anatomical variability, accompanying a novel anomaly scoring scheme based on the mapping from image space to a latent space. Applied to new data, the model labels anomalies, and scores image patches indicating their fit into the learned distribution. Results on optical coherence tomography images of the retina demonstrate that the approach correctly identifies anomalous images, such as images containing retinal fluid or hyperreflective foci.

on behalf of the HAWK and HARRIER Study Investigators* Purpose: Two similarly designed phase 3 trials (HAWK and HARRIER) compared brolucizumab, a singlechain antibody fragment that inhibits vascular endothelial growth factor-A, with aflibercept to treat neovascular age-related macular degeneration (nAMD).Design: Double-masked, multicenter, active-controlled, randomized trials. Participants: Patients (N ¼ 1817) with untreated, active choroidal neovascularization due to age-related macular degeneration in the study eye.Intervention: Patients were randomized to intravitreal brolucizumab 3 mg (HAWK only) or 6 mg or aflibercept 2 mg. After loading with 3 monthly injections, brolucizumab-treated eyes received an injection every 12 weeks (q12w) and were interval adjusted to every 8 weeks (q8w) if disease activity was present; aflibercept-treated eyes received q8w dosing.Main Outcome Measures: The primary hypothesis was noninferiority in mean best-corrected visual acuity (BCVA) change from baseline to Week 48 (margin: 4 letters). Other key end points included the percentage of patients who maintained q12w dosing through Week 48 and anatomic outcomes.Results: At Week 48, each brolucizumab arm demonstrated noninferiority to aflibercept in BCVA change from baseline (least squares [LS] mean, þ6.6 [6 mg] and þ6.1 [3 mg] letters with brolucizumab vs. þ6.8 letters with aflibercept [HAWK]; þ6.9 [brolucizumab 6 mg] vs. þ7.6 [aflibercept] letters [HARRIER]; P < 0.001 for each comparison). Greater than 50% of brolucizumab 6 mgetreated eyes were maintained on q12w dosing through Week 48 (56% [HAWK] and 51% [HARRIER]). At Week 16, after identical treatment exposure, fewer brolucizumab 6 mgetreated eyes had disease activity versus aflibercept in HAWK (24.0% vs. 34.5%; P ¼ 0.001) and HARRIER (22.7% vs. 32.2%; P ¼ 0.002). Greater central subfield thickness reductions from baseline to Week 48 were observed with brolucizumab 6 mg versus aflibercept in HAWK (LS mean À172.8 mm vs. À143.7 mm; P ¼ 0.001) and HARRIER (LS mean À193.8 mm vs. À143.9 mm; P < 0.001). Anatomic retinal fluid outcomes favored brolucizumab over aflibercept. Overall, adverse event rates were generally similar with brolucizumab and aflibercept.Conclusions: Brolucizumab was noninferior to aflibercept in visual function at Week 48, and >50% of brolucizumab 6 mgetreated eyes were maintained on q12w dosing interval through Week 48. Anatomic outcomes favored brolucizumab over aflibercept. Overall safety with brolucizumab was similar to aflibercept (Clin-icalTrials.gov; NCT02307682, NCT02434328).

No abstract

No abstract