In patients with PD with normal vision, we found a decrease in the electrical activity of the fovea as well as in the thickness of the RNFL. Multifocal electroretinogram and OCT scan objectively detect early subclinical PD-associated visual functional impairment.

No abstract

The timeline of computer vision research is marked with advances in learning and utilizing efficient contextual representations. Most of them, however, are targeted at improving model performance on a single downstream task. We consider a multi-task environment for dense prediction tasks, represented by a common backbone and independent task-specific heads. Our goal is to find the most efficient way to refine each task prediction by capturing cross-task contexts dependent on tasks' relations. We explore various attention-based contexts, such as global and local, in the multi-task setting and analyze their behavior when applied to refine each task independently. Empirical findings confirm that different source-target task pairs benefit from different context types. To automate the selection process, we propose an Adaptive Task-Relational Context (ATRC) module, which samples the pool of all available contexts for each task pair using neural architecture search and outputs the optimal configuration for deployment. Our method achieves state-of-the-art performance on two important multi-task benchmarks, namely NYUD-v2 and PASCAL-Context. The proposed ATRC has a low computational toll and can be used as a drop-in refinement module for any supervised multi-task architecture.

Purpose: To report the clinical outcomes of the use of a novel specially designed scleral fixated intraocular lens, the Carlevale intraocular lens (carlevale IOL, Soleko, Italy) for the correction of aphakia in the absence of capsular support of variable etiology. Methods: This retrospective, non-comparative study included 169 eyes of 169 consecutive patients who underwent 3-port pars plana vitrectomy and scleral fixation on Carlevale IOL. Inclusion criteria were at least 6 months’ follow-up period, patients > 18 years old who underwent vitrectomy and Carlevale IOL placement for aphakia and inadequate capsular support Results: The median follow up period of 9 months (range 6–18 months). Mean post-operative BCVA at the last follow-up visit was 20/25 (0.09 ± 0.1 LogMAR), improving from a mean baseline BCVA of 20/80 (0.58 ± 0.49 LogMAR), a statistically significant change ( p = 0.0001). Regarding the post-operative complications, a transient rise in the IOP was observed in 28 patients (16.5%) and mild vitreous hemorrhage was observed in the immediate post-operative period in eight eyes (4.7%) and it spontaneously resolved within 3 weeks. All patients demonstrated good IOL position at the end of the follow-up without IOL capture. None of the patients required re-operation. Conlcusion: The present study represents the largest to date in evaluating the use of carlevale IOL in patients with aphakia and inadequate capsular support. The technique is safe and provides excellent post-operative IOL fixation without IOL capture in any of the patients studied.

Multi-task networks are commonly utilized to alleviate the need for a large number of highly specialized single-task networks. However, two common challenges in developing multi-task models are often overlooked in literature. First, enabling the model to be inherently incremental, continuously incorporating information from new tasks without forgetting the previously learned ones (incremental learning). Second, eliminating adverse interactions amongst tasks, which has been shown to significantly degrade the single-task performance in a multi-task setup (task interference). In this paper, we show that both can be achieved simply by reparameterizing the convolutions of standard neural network architectures into a non-trainable shared part (filter bank) and taskspecific parts (modulators), where each modulator has a fraction of the filter bank parameters. Thus, our reparameterization enables the model to learn new tasks without adversely affecting the performance of existing ones. The results of our ablation study attest the efficacy of the proposed reparameterization. Moreover, our method achieves state-of-the-art on two challenging multi-task learning benchmarks, PASCAL-Context and NYUD, and also demonstrates superior incremental learning capability as compared to its close competitors. The code and models are made publicly available 1 .

BackgroundBeta-thalassemia is a severe genetic blood disorder caused by a mutation in the gene encoding for the beta chains of hemoglobin. Individuals with beta-thalassemia major require regular lifelong Red Blood Cell transfusions to survive. Ocular involvement is quite common and may have serious implications.MethodsExtensive review of observational studies on beta-thalassemia, to determine the prevalence and spectrum of ocular abnormalities, by clinical examination and multimodal imaging, and to investigate risk factors for their development.ResultsFrequency of ocular involvement differs among various studies (41.3–85 %, three studies). Ocular findings in beta-thalassemia may correlate to the disease itself, iron overload or the chelating agents used. Beta-thalassemia ocular manifestations include ocular surface disease, as demonstrated by tear function parameters (two studies). Lens opacities are present in 9.3–44 % (five studies). Lenticular opacities and RPE degeneration correlated positively with use of desferrioxamine and deferriprone respectively (two studies). Ocular fundus abnormalities characteristic of pseudoxanthoma elasticum (PXE), including peau d’orange, angioid streaks, pattern dystrophy-like changes, and optic disc drusen are a consistent finding in seven studies. Patients with PXE-like fundus changes were older than patients without these fundus changes (two studies). Age (two studies) and splenectomy (one study) had the strongest association with presence of PXE-like fundus changes. Increased retinal vascular tortuosity independently of the PXE-like fundus changes was found in 11–17.9 % (three studies), which was associated with aspartate amino transferase, hemoglobin and ferritin levels (two studies). Fundus autofluorescence and electrophysiological testing (ERG and EOG) may indicate initial stages or more widespread injury than is suggested by fundus examination (two studies).ConclusionsBeta-thalassemia may present with various signs, both structural and functional. Pseudoxanthoma elasticum like fundus changes are a frequent finding in patients with b-thalassemia. These changes increase with duration or severity of the disease. Retinal vascular tortuosity may be an additional disease manifestation related to the severity and duration of anemia and independent of the PXE-like syndrome. Patients with long-standing disease need regular ophthalmic checkups because they are at risk of developing PXE-like fundus changes and potentially of subsequent choroidal neovascularization.

Nowadays, the increasingly growing number of mobile and computing devices has led to a demand for safer user authentication systems. Face anti-spoofing is a measure towards this direction for biometric user authentication, and in particular face recognition, that tries to prevent spoof attacks. The state-of-the-art anti-spoofing techniques leverage the ability of deep neural networks to learn discriminative features, based on cues from the training set images or video samples, in an effort to detect spoof attacks. However, due to the particular nature of the problem, i.e. large variability due to factors like different backgrounds, lighting conditions, camera resolutions, spoof materials, etc., these techniques typically fail to generalize to new samples. In this paper, we explicitly tackle this problem and propose a class-conditional domain discriminator module, that, coupled with a gradient reversal layer, tries to generate live and spoof features that are discriminative, but at the same time robust against the aforementioned variability factors. Extensive experimental analysis shows the effectiveness of the proposed method over existing image-and video-based anti-spoofing techniques, both in terms of numerical improvement as well as when visualizing the learned features.

We present an approach for encoding visual task relationships to improve model performance in an Unsupervised Domain Adaptation (UDA) setting. Semantic segmentation and monocular depth estimation are shown to be complementary tasks; in a multi-task learning setting, a proper encoding of their relationships can further improve performance on both tasks. Motivated by this observation, we propose a novel Cross-Task Relation Layer (CTRL), which encodes task dependencies between the semantic and depth predictions. To capture the cross-task relationships, we propose a neural network architecture that contains task-specific and cross-task refinement heads. Furthermore, we propose an Iterative Self-Learning (ISL) training scheme, which exploits semantic pseudo-labels to provide extra supervision on the target domain. We experimentally observe improvements in both tasks' performance because the complementary information present in these tasks is better captured. Specifically, we show that: (1) our approach improves performance on all tasks when they are complementary and mutually dependent; (2) the CTRL helps to improve both semantic segmentation and depth estimation tasks performance in the challenging UDA setting;(3) the proposed ISL training scheme further improves the semantic segmentation performance. The implementation is available at https://github.com/susaha/ctrl-uda.