Policy-Driven, Multimodal Deep Learning for Predicting Visual Fields from the Optic Disc and OCT Imaging.

PURPOSE: To develop and validate a deep learning (DL) system for predicting each point on visual fields (VFs) from disc and OCT imaging and derive a structure-function mapping. DESIGN: Retrospective, cross-sectional database study. PARTICIPANTS: A total of 6437 patients undergoing routine care for glaucoma in 3 clinical sites in the United Kingdom. METHODS: OCT and infrared reflectance (IR) optic disc imaging were paired with the closest VF within 7 days. EfficientNet B2 was used to train 2 single-modality DL models to predict each of the 52 sensitivity points on the 24-2 VF pattern. A policy DL model was designed and trained to fuse the 2 model predictions. MAIN OUTCOME MEASURES: Pointwise mean absolute error (PMAE). RESULTS: A total of 5078 imaging scans to VF pairs were used as a held-out test set to measure the final performance. The improvement in PMAE with the policy model was 0.485 (0.438, 0.533) decibels (dB) compared with the IR image of the disc alone and 0.060 (0.047, 0.073) dB with to the OCT alone. The improvement with the policy fusion model was statistically significant (P < 0.0001). Occlusion masking shows that the DL models learned the correct structure-function mapping in a data-driven, feature agnostic fashion. CONCLUSIONS: The multimodal, policy DL model performed the best; it provided explainable maps of its confidence in fusing data from single modalities and provides a pathway for probing the structure-function relationship in glaucoma.

Our Experience of Deep Sclerectomy at a Tertiary Center in the United Kingdom Over 14 Years.

Background Deep sclerectomy (DS) is a non-penetrating surgical procedure for glaucoma, reducing the resistance to aqueous outflow and lowering intraocular pressure while maintaining a physiological barrier between the anterior chamber and the sub-scleral space. This offers a lower complication profile than penetrating procedures, though with less intraocular pressure (IOP) reduction. Methods We retrospectively reviewed the electronic record for all DS undertaken at our hospital (a tertiary care center) over 14 years, collecting data on demographics, diagnosis, IOP, visual acuity, complications, medications, and further procedures required. Results Eighty eyes of 69 patients underwent DS, with a mean follow-up period of 53.5 months. The mean pre-operative IOP was 23.55 mmHg (range 11-52, standard deviation 8.46); the mean final IOP was 13.61 mmHg (range 5-35, SD 4.73), with a mean reduction of 42.21%. The mean change in glaucoma medications was -1.64. 78.40% experienced a reduction in glaucoma treatment. Post-operatively, 43.80% had no complications; this improved to 85.0% when numerical hypotony and raised IOP without visual sequelae were excluded. Further procedures required included Nd:YAG goniopuncture (10%), bleb needling (13.75%) or revision (7.5%), iridectomy (3.75%), goniosynechiolysis (1.25%), and autologous blood injection (1.25%). Two eyes were converted to trabeculectomy peri-operatively, with seven overall (8.75%) requiring trabeculectomy over the course of follow-up. 3.75% underwent glaucoma drainage device implantation, and 3.75% underwent cyclodiode laser. Conclusion We have found DS to be a safe, effective procedure for selected patients where trabeculectomy has a high likelihood of failure or where a higher IOP can be tolerated.

Inner Retinal Thinning Comparison between Branch Retinal Artery Occlusion and Primary Open-Angle Glaucoma.

Purpose: to assess the tomographic retinal layers' thickness in eyes affected by branch retinal artery occlusion (BRAO) and to compare it to those of patients affected by primary open angle glaucoma (POAG). Methods: retrospective review of 27 patients; 16 with BRAO (16 eyes) and 11 with POAG (20 eyes) were identified among those who received SD-OCT scans, including analysis of macular retinal nerve fiber layer (mRNFL), ganglion cell layer (GCL), inner plexiform layer (IPL), inner nuclear layer (INL), neuroretinal rim (NRR), circumpapillary RNFL at 3.5 mm and hemisphere asymmetry (HA). Results: the total IPL and INL thinning difference between the two groups was statistically significant (p = 0.0067 and p < 0.0001, respectively). The HA difference for the total macular thinning, mRNFL, GCL, IPL and INL (p < 0.0001) was also statistically significant. The analysis of the average total retinal thinning, total mRNFL and GCL thinning showed no statistically significant difference between the two groups. Conclusions: unilateral inner retinal thinning may represent a sign of temporal BRAO, particularly for INL thinning and HA difference over 17µm in total retinal layer thinning. This information is particularly useful in the diagnosis of previous, undiagnosed BRAO and may help prevent further retinal arterial occlusion and possible cerebrovascular incidents.

Genome-wide association study of primary open angle glaucoma risk and quantitative traits.

PURPOSE: Primary open angle glaucoma (POAG) is a characteristic optic neuropathy which progresses to irreversible vision loss. Few genes have been detected that influence POAG susceptibility and other genes are therefore likely to be involved. We analyzed carefully characterized POAG cases in a genome-wide association study (GWAS). METHODS: We performed a GWAS in 387 POAG cases using public control data (WTCCC2). We also investigated the quantitative phenotypes, cup:disc ratio (CDR), central corneal thickness (CCT), and intra-ocular pressure (IOP). Promising single nucleotide polymorphisms (SNPs), based on various prioritisation criteria, were genotyped in a cohort of 294 further POAG cases and controls. RESULTS: We found 2 GWAS significant results in the discovery stage for association, one of which which had multiple evidence in the gene 'neural precursor cell expressed, developmentally down-regulated 9' (NEDD9; rs11961171, p=8.55E-13) and the second on chromosome 16 with no supporting evidence. Taking into account all the evidence from risk and quantitative trait ocular phenotypes we chose 86 SNPs for replication in an independent sample. Our most significant SNP was not replicated (p=0.59). We found 4 nominally significant results in the replication cohort, but none passed correction for multiple testing. Two of these, for phenotypes CDR (rs4385494, discovery p=4.51x10-5, replication p=0.029) and CCT (rs17128941, discovery p=5.52x10-6, replication=0.027), show the consistent direction of effects between the discovery and replication data. We also assess evidence for previously associated known genes and find evidence for the genes 'transmembrane and coiled-coil domains 1' (TMCO1) and 'cyclin-dependent kinase inhibitor 2B' (CDKN2B). CONCLUSIONS: Although we were unable to replicate any novel results for POAG risk, we did replicate two SNPs with consistent effects for CDR and CCT, though they do not withstand correction for multiple testing. There has been a range of publications in the last couple of years identifying POAG risk genes and genes involved in POAG related ocular traits. We found evidence for 3 known genes (TMCO1, CDKN2B, and S1 RNA binding domain 1 [SRBD1]) in this study. Novel rare variants, not detectable by GWAS, but by new methods such as exome sequencing may hold the key to unravelling the remaining contribution of genetics to complex diseases such as POAG.

A multi-centre interventional case series of 259 ab-interno Xen gel implants for glaucoma, with and without combined cataract surgery.

AIMS: To assess the efficacy of Xen in reducing intraocular pressure (IOP) in varying glaucoma subtypes. To assess the effect of combined phacoemulsification. To determine the frequency of complications and explore further bleb management needed. METHODS: Retrospective case note review of all patients undergoing Xen implantation across four centres from August 2015 to May 2017. RESULTS: In total, 259 consecutive surgeries of 226 patients were reviewed. IOP reduced from 19.3 (SD ± 6.0) mmHg preoperatively to 14.2 (SD ± 4.4) at month 12 and 13.5 (SD ± 3.3) at month 18 (p < 0.0001). Medication usage reduced from 2.6 (±1.1) preoperatively to 0.8 (±1.0) at month 12 (p < 0.0001) and 1.1 (±1.3) medications at month 18 (p < 0.0001). Simultaneous phacoemulsification did not alter outcomes as Xen IOP was 14.3 (SD ± 4.7) mmHg and Phaco-Xen was 13.8 (SD ± 2.6) mmHg at month 12 (p = 0.5367). Xen appears to be effective in previous failed filtration surgery. Adverse events included: IOP spikes of ≥30 mmHg in 33 (12.7%) cases, secondary filtration surgery required in 24 (9.3%) cases; implant exposure in 6 (2.3%) cases; persistent hypotonous maculopathy in 5 (1.9%) cases; persistent choroidal effusions in 4 (1.5%) cases; a cyclodialysis cleft secondary to implant insertion in 1 (0.5%) case; and 1 (0.5%) case of endophthalmitis post-implant bleb resuturing. In all, 40.9% of cases required postoperative bleb needling or antimetabolite injection. CONCLUSIONS: Xen reduces IOP and medications at 18 months. Adverse events are uncommon. Careful postoperative surveillance and low threshold for bleb management is needed. Xen is safe and effective in mild to moderate glaucoma.

Deep sclerectomy versus trabeculectomy: a morphological study with anterior segment optical coherence tomography.

PURPOSE: To investigate the intraocular pressure (IOP) lowering mechanisms of deep sclerectomy (DS) with anterior segment optical coherence tomography (AS-OCT). METHODS: In a prospective cross-sectional study, AS-OCT parameters were compared between DS, trabeculectomy and control cases. Association with IOP and success (IOP≤16 mm Hg without medication) was investigated. RESULTS: 18 DS (15 patients), 17 trabeculectomy (16 patients) and 15 controls (15 patients) were examined. Successful had a taller intrascleral lake (IL) and thicker conjunctival/Tenon's layer (CTL) than non-successful cases (513.3 vs 361.1 µm, p=0.027 and 586.7 vs 251.1 µm, p<0.001, respectively). CTL thickness correlated with IOP (r=-0.6407, p=0.004). CTL thickness was significantly different between controls, DS and trabeculectomy (mean (SD): 203.3 (62.6) vs 418.9 (261.9) vs 604.1 (220.7) µm, p<0.0001). Successful trabeculectomy cases had a taller bleb cavity (BC) than non-successful cases (607.5 vs 176.7 µm, p=0.041). CTL microcysts were detected in 50% of DS and 52.9% of trabeculectomy cases (p=1). CONCLUSIONS: Trans-conjunctival aqueous percolation was identified as a novel DS drainage route. DS had a fluid reservoir below the scleral flap, the IL, in analogy to the trabeculectomy BC. A postoperative tall IL and a thick CTL were associated with good outcome.

Common variants near CAV1 and CAV2 are associated with primary open-angle glaucoma.

We conducted a genome-wide association study for primary open-angle glaucoma (POAG) in 1,263 affected individuals (cases) and 34,877 controls from Iceland. We identified a common sequence variant at 7q31 (rs4236601[A], odds ratio (OR) = 1.36, P = 5.0 × 10⁻¹°). We then replicated the association in sample sets of 2,175 POAG cases and 2,064 controls from Sweden, the UK and Australia (combined OR = 1.18, P = 0.0015) and in 299 POAG cases and 580 unaffected controls from Hong Kong and Shantou, China (combined OR = 5.42, P = 0.0021). The risk variant identified here is located close to CAV1 and CAV2, both of which are expressed in the trabecular meshwork and retinal ganglion cells that are involved in the pathogenesis of POAG.