An imputation approach using subdistribution weights for deep survival analysis with competing events.

With the popularity of deep neural networks (DNNs) in recent years, many researchers have proposed DNNs for the analysis of survival data (time-to-event data). These networks learn the distribution of survival times directly from the predictor variables without making strong assumptions on the underlying stochastic process. In survival analysis, it is common to observe several types of events, also called competing events. The occurrences of these competing events are usually not independent of one another and have to be incorporated in the modeling process in addition to censoring. In classical survival analysis, a popular method to incorporate competing events is the subdistribution hazard model, which is usually fitted using weighted Cox regression. In the DNN framework, only few architectures have been proposed to model the distribution of time to a specific event in a competing events situation. These architectures are characterized by a separate subnetwork/pathway per event, leading to large networks with huge amounts of parameters that may become difficult to train. In this work, we propose a novel imputation strategy for data preprocessing that incorporates weights derived from a time-discrete version of the classical subdistribution hazard model. With this, it is no longer necessary to add multiple subnetworks to the DNN to handle competing events. Our experiments on synthetic and real-world datasets show that DNNs with multiple subnetworks per event can simply be replaced by a DNN designed for a single-event analysis without loss in accuracy.

Association of Visual Function Measures with Drusen Volume in Early Stages of Age-Related Macular Degeneration.

Purpose: To assess which visual function measures are most strongly associated with overall retinal drusen volume in age-related macular degeneration (AMD). Methods: A total of 100 eyes (16 eyes with early AMD, 62 eyes with intermediate AMD, and 22 eyes from healthy controls) were recruited in this cross-sectional study. All subjects underwent several functional assessments: best-corrected visual acuity (BCVA), low-luminance visual acuity (LLVA), visual acuity (VA) measured with the Moorfields Acuity Chart (MAC-VA), contrast sensitivity with the Pelli-Robson test, reading speed using the International Reading Speed texts, and mesopic and dark-adapted microperimetry. Drusen volume was automatically determined based on optical coherence tomography using an approach based on convolutional neural networks. The relationship between drusen volume and visual function was assessed with linear regressions controlling for confounders. Results: Mean drusen volume and MAC-VA differed significantly among all AMD stages and controls (P < 0.001). In univariate linear regression, LLVA, MAC-VA, contrast sensitivity, and mesopic and dark-adapted microperimetry were significantly negatively associated with the overall drusen volume (all P < 0.006). After controlling for AMD stage, age, and the presence of subretinal drusenoid deposits, MAC-VA and mesopic and dark-adapted microperimetry were still significantly associated with drusen volume (P = 0.008, P = 0.023, and P = 0.022, respectively). Conclusions: Our results suggest that MAC-VA, as well as mesopic and dark-adapted microperimetry, might indicate structural changes related to drusen volume in early stages of AMD.

Replication and Refinement of an Algorithm for Automated Drusen Segmentation on Optical Coherence Tomography.

Here, we investigate the extent to which re-implementing a previously published algorithm for OCT-based drusen quantification permits replicating the reported accuracy on an independent dataset. We refined that algorithm so that its accuracy is increased. Following a systematic literature search, an algorithm was selected based on its reported excellent results. Several steps were added to improve its accuracy. The replicated and refined algorithms were evaluated on an independent dataset with the same metrics as in the original publication. Accuracy of the refined algorithm (overlap ratio 36-52%) was significantly greater than the replicated one (overlap ratio 25-39%). In particular, separation of the retinal pigment epithelium and the ellipsoid zone could be improved by the refinement. However, accuracy was still lower than reported previously on different data (overlap ratio 67-76%). This is the first replication study of an algorithm for OCT image analysis. Its results indicate that current standards for algorithm validation do not provide a reliable estimate of algorithm performance on images that differ with respect to patient selection and image quality. In order to contribute to an improved reproducibility in this field, we publish both our replication and the refinement, as well as an exemplary dataset.

Secondary and Exploratory Outcomes of the Subthreshold Nanosecond Laser Intervention Randomized Trial in Age-Related Macular Degeneration: A LEAD Study Report.

PURPOSE: To evaluate the secondary and exploratory outcomes of the Laser Intervention in Early Stages of Age-Related Macular Degeneration (LEAD) study, a 36-month trial of a subthreshold nanosecond laser (SNL) treatment for slowing the progression to late age-related macular degeneration (AMD) in its early stages. DESIGN: Multicenter, randomized, sham-controlled trial. PARTICIPANTS: Two-hundred ninety-two patients with bilateral large drusen. METHODS: Participants were randomly assigned to receive SNL or sham treatment to the study eye at 6-month intervals. MAIN OUTCOME MEASURES: The secondary outcome measure of the LEAD study was the time to development of late AMD, defined by multimodal imaging in the non-study eye. The exploratory outcome measures were the rate of change in best-corrected visual acuity (BCVA), low-luminance visual acuity, microperimetric mean sensitivity, drusen volume in the study and non-study eyes, and participant-reported outcomes based on the Night Vision Questionnaire and Impact of Vision Impairment questionnaire. RESULTS: Progression to late AMD in the non-study eye was not significantly delayed with SNL treatment (hazard ratio, 0.83; 95% confidence interval, 0.40-1.71; P = 0.611). There was no evidence of effect modification based on the coexistence of reticular pseudodrusen; interaction P = 0.065). There was no significant difference between study groups in the rate of change of low-luminance visual acuity, microperimetric mean sensitivity, and drusen volume in the study or non-study eyes, and Night Vision Questionnaire and Impact of Vision Impairment questionnaire scores (all P ≥ 0.167). The rate of BCVA decline was slightly higher for participants in the SNL group compared with the sham treatment group in the study eye (-0.54 and 0.23 letters/year, respectively; P < 0.001) but not the non-study eye (-0.48 and -0.56 letters/year, respectively; P = 0.628). CONCLUSIONS: Subthreshold nanosecond laser treatment of one eye did not have an effect on delaying progression to late AMD in the fellow eye and did not, in general, have an impact on the exploratory structural, functional, and participant-reported outcomes.