Author: Leila Sara Eppenberger (Switzerland)
Co-authors: Bart Liefers, Livia Faes, Shana Sturla, Martin K. Schmid, Konstantinos Balaskas, Lucas M. Bachmann
Purpose
Patients suffering from chronic retinal pathologies, require individually tailored, long-term monitoring and therapy. Usually patient management consists of frequent visits to the ophthalmologist and relatively cost-intensive diagnostic imaging. Both in view of the current pandemic and the availability of long-acting treatment the need for reliable home monitoring strategies are gaining importance. Besides the well-established optical coherence tomography (OCT) imaging, further methods are necessary to quantify visual function. Recently, two self-examination and self-monitoring apps to assess visual distortions that run on mobile devices have become available. The Alleye® mobile hyperacuity task (Ouculocare medical Inc. Zurich, Switzerland) is the first application allowing to quantitatively measure and assess the location of visual impairment within the central retina. Using data from a previously published cohort, this study explored a new method of matching morphological features present on OCT scans, with corresponding functional impairment on a spatial level. When the collection time was near, OCT data were matched with corresponding functional data from Alleye®. By studying the interplay between morphological and functional features on the central retina, this study set out to provide new information on the underlying mechanistic relation, a topic that remains ill understood.
Setting/Venue
Originally, patients attending the Medical Retina Service of the Eye Clinic of the Cantonal Hospital Lucerne between March and June 2016 were eligible to be included in the prospective implementation study of the Alleye® application. Primary criterion for the inclusion of patients in the present retrospective study, was the availability of high-quality data from self-examinations via self-monitoring Alleye® test, as well as OCT images acquired during clinical visits in the same time period. Finally, a total of 77 eyes of 53 patients were included and their data extracted from the clinical electronic medical records and imaging databases.
Methods
307 OCT scans of the 77 eyes were transmitted to the Moorfields Ophthalmic Reading Center. Consecutively, a deep learning algorithm was used for segmentation of four morphological features: intraretinal and subretinal fluid (IRF and SRF), drusen and pigment epithelial detachment (PED) and subretinal hyperreflective material (SRHM). From the segmented results 2D enface maps for each morphological feature were created. Using the Gullstrand’s reduced eye model the macular areas associated with the twelve Alleye® tasks (p1-p12) were projected onto the near infrared image acquired with the OCT scans. To create Alleye® heat maps the achieved score values (0-1) were interpolated using an inverse distance weighting function. Temporal matching of hyperacuity tests and OCT scans for every patients was conducted by applying the Hungarian algorithm. Only matching examinations with a maximum of 2 days between the two examination entities were considered. Correlation analysis with pairwise Spearman’s and Pearson’s tests between Alleye® score values (0-1) of the twelve tasks (p1-p12) and volumes of the four morphological features (SRHM, SRF, IRF and PED), as well as pixel-by-pixel comparison for the Alleye® heat maps and the enface images of the four morphological features were conducted.
Results
Overall, mean age of the 53 study participants was 75 years. The large majority (n=57, 74%) of eyes were diagnosed with age-related macular degeneration (AMD). A maximum time interval of two days between OCT acquisition and the Alleye® self-test date was found for 208 cases. Mean best-corrected visual acuity was 71 (ETDRS, number of letters), mean total Alleye® score (0-100) was 51 (range: 2-100). Considering the temporally first best matching observation per eye pairwise comparisons of Alleye® score values and volumes of the four morphological features showed a small negative correlation for SRHM (rs = -0.164, p <.001 for n=77) and very low, statistically insignificant, but still inverse correlations for PED, SRF and IRF, respectively. Juxtaposition of Alleye® heat maps and enface images of morphological features demonstrated varying patterns with main observations being: (1) When higher volumes of PED, SRF, IRF and above all SRHM were present, the Alleye® heat maps also showed more areas with low score values. (2) While the morphological volumes summarized in the enface images remained relatively stable in a certain case, the corresponding Alleye® heat maps showed changes in appearance, i.e. location change of areas with lowest score values.
Conlusions
In summary, a novel approach for the graphical juxtaposition of morphological and functional data on a two dimensional level is introduced, which allows for a side-by-side comparison of results acquired through these two different diagnostic entities. Due to the retrospective design of the study the preliminary findings require further investigation and confirmation in more controlled future settings. Nevertheless, the available data showed that often high volumes of morphological features PED, SRF, IRF and SRHM go in parallel to reduced hyperacuity function. At the same time, it is determined that there is a lively dynamic, especially on the part of the function, and that repeated measurements are necessary to better describe a patient’s disease state and progression. Overall, this study offers a first step for a system’s approach to improve an individual patient’s outcome by integrating information about morphology and function. Hence, the importance of a complementary application of both imaging diagnostic tools and functional testing procedures for precise and earlier monitoring of disease onset and progression needs to be emphasized.
Financial Disclosure
None.
Comments
-
