Search

Purpose: The region of growth (ROG) of geographic atrophy (GA) throughout the macular area has an impact on visual outcomes. Here, we developed multiple deep learning models to predict the 1-year ROG of GA lesions using fundus autofluorescence (FAF) images. Design: In this retrospective analysis, 3 types of models were developed using FAF images collected 6 months after baseline to predict the GA lesion area (segmented lesion mask) at 1.5 years, FAF images collected at baseline and 6 months to predict the GA lesion at 1.5 years, and FAF images collected 6 months after baseline to predict the GA lesion at 1 and 1.5 years. The 1-year ROG from the 6-month visit was derived by taking the difference between the GA lesion area (segmented lesion mask) at the 1.5-year and 6-month visits. Participants: Patients enrolled in the following lampalizumab clinical trials and prospective observational studies: NCT02247479, NCT02247531, NCT02479386, and NCT02399072. Methods: Datasets of study eyes from 597 patients were split into model training (310), validation (78), and test sets (209), stratified by baseline or initial lesion area, lesion growth rate, foveal involvement, and focality. Deep learning experiments were performed using the 2-dimensional U-Net; whole-lesion and multiclass models were developed. Main Outcome Measures: The performance of the models was evaluated by calculating the Dice score, coefficient of determination (R2), and the squared Pearson correlation coefficient (r2) between the true and derived GA lesion 1-year ROG. Results: The model using baseline and 6-month FAF images to predict GA lesion enlargement at 1.5 years had the best performance for the derived 1-year ROG. Mean Dice scores were 0.73, 0.68, and 0.70 in the training, validation, and test sets, respectively. The R2 (0.77, 0.53, and 0.79) and r2 (0.83, 0.61, and 0.79) showed similar trends across the 3 sets. Conclusions: These findings show the potential of using baseline and/or 6-month visit FAF images to predict 1-year GA ROG using a deep learning approach. This work could potentially help support decision-making in clinical trials and more informed treatment decisions in clinical practice. Financial Disclosure(s): Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.

Purpose: To determine the interreader agreement for reticular pseudodrusen (RPD) assessment on combined infrared reflectance (IR) and OCT imaging in the early stages of age-related macular degeneration across a range of different criteria to define their presence. Design: Interreader agreement study. Participants: Twelve readers from 6 reading centers. Methods: All readers evaluated 100 eyes from individuals with bilateral large drusen for the following: (1) the presence of RPD across a range of different criteria and (2) the number of Stage 2 or 3 RPD lesions (from 0 to ≥ 5 lesions) on an entire OCT volume scan and on a selected OCT B-scan. Supportive information was available from the corresponding IR image. Main Outcome Measures: Interreader agreement, as assessed by Gwet’s first-order agreement coefficient (AC1). Results: When evaluating an entire OCT volume scan, there was substantial interreader agreement for the presence of any RPD, any or ≥ 5 Stage 2 or 3 lesions, and ≥ 5 definite lesions on en face IR images corresponding to Stage 2 or 3 lesions (AC1 = 0.60–0.72). On selected OCT B-scans, there was also moderate-to-substantial agreement for the presence of any RPD, any or ≥ 5 Stage 2 or 3 lesions (AC1 = 0.58–0.65) and increasing levels of agreement with increasing RPD stage (AC1 = 0.08, 0.56, 0.78, and 0.99 for the presence of any Stage 1, 2, 3, and 4 lesions, respectively). There was substantial agreement regarding the number of Stage 2 or 3 lesions on an entire OCT volume scan (AC1 = 0.68), but only fair agreement for this evaluation on selected B-scans (AC1 = 0.30). Conclusions: There was generally substantial or near-substantial—but not near-perfect—agreement for assessing the presence of RPD on entire OCT volume scans or selected B-scans across a range of differing RPD criteria. These findings underscore how interreader variability would likely contribute to the variability of findings related to the clinical associations of RPD. The low levels of agreement for assessing RPD number on OCT B-scans underscore the likely challenges of quantifying RPD extent with manual grading. Financial Disclosure(s): Proprietary or commercial disclosure may be found after the references.

Purpose: To develop a severity classification for macular telangiectasia type 2 (MacTel) disease using multimodal imaging. Design: An algorithm was used on data from a prospective natural history study of MacTel for classification development. Subjects: A total of 1733 participants enrolled in an international natural history study of MacTel. Methods: The Classification and Regression Trees (CART), a predictive nonparametric algorithm used in machine learning, analyzed the features of the multimodal imaging important for the development of a classification, including reading center gradings of the following digital images: stereoscopic color and red-free fundus photographs, fluorescein angiographic images, fundus autofluorescence images, and spectral-domain (SD)-OCT images. Regression models that used least square method created a decision tree using features of the ocular images into different categories of disease severity. Main Outcome Measures: The primary target of interest for the algorithm development by CART was the change in best-corrected visual acuity (BCVA) at baseline for the right and left eyes. These analyses using the algorithm were repeated for the BCVA obtained at the last study visit of the natural history study for the right and left eyes. Results: The CART analyses demonstrated 3 important features from the multimodal imaging for the classification: OCT hyper-reflectivity, pigment, and ellipsoid zone loss. By combining these 3 features (as absent, present, noncentral involvement, and central involvement of the macula), a 7-step scale was created, ranging from excellent to poor visual acuity. At grade 0, 3 features are not present. At the most severe grade, pigment and exudative neovascularization are present. To further validate the classification, using the Generalized Estimating Equation regression models, analyses for the annual relative risk of progression over a period of 5 years for vision loss and for progression along the scale were performed. Conclusions: This analysis using the data from current imaging modalities in participants followed in the MacTel natural history study informed a classification for MacTel disease severity featuring variables from SD-OCT. This classification is designed to provide better communications to other clinicians, researchers, and patients. Financial Disclosure(s): Proprietary or commercial disclosure may be found after the references.