{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/forecasting-future-humphrey-visual-fields","title":"Forecasting Future Humphrey Visual Fields Using Deep Learning","arxiv_id":"1804.04543","date":"2018-04-02","proceeding":null,"authors":["Joanne C. Wen","Cecilia S. Lee","Pearse A. Keane","Sa Xiao","Yue Wu","Ariel Rokem","Philip P. Chen","Aaron Y. Lee"],"abstract":"Purpose: To determine if deep learning networks could be trained to forecast\na future 24-2 Humphrey Visual Field (HVF).\n  Participants: All patients who obtained a HVF 24-2 at the University of\nWashington.\n  Methods: All datapoints from consecutive 24-2 HVFs from 1998 to 2018 were\nextracted from a University of Washington database. Ten-fold cross validation\nwith a held out test set was used to develop the three main phases of model\ndevelopment: model architecture selection, dataset combination selection, and\ntime-interval model training with transfer learning, to train a deep learning\nartificial neural network capable of generating a point-wise visual field\nprediction.\n  Results: More than 1.7 million perimetry points were extracted to the\nhundredth decibel from 32,443 24-2 HVFs. The best performing model with 20\nmillion trainable parameters, CascadeNet-5, was selected. The overall MAE for\nthe test set was 2.47 dB (95% CI: 2.45 dB to 2.48 dB). The 100 fully trained\nmodels were able to successfully predict progressive field loss in glaucomatous\neyes up to 5.5 years in the future with a correlation of 0.92 between the MD of\npredicted and actual future HVF (p < 2.2 x 10 -16 ) and an average difference\nof 0.41 dB.\n  Conclusions: Using unfiltered real-world datasets, deep learning networks\nshow an impressive ability to not only learn spatio-temporal HVF changes but\nalso to generate predictions for future HVFs up to 5.5 years, given only a\nsingle HVF.","url_abs":"http://arxiv.org/abs/1804.04543v1","url_pdf":"http://arxiv.org/pdf/1804.04543v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"forecasting-future-humphrey-visual-fields","repo_url":"https://github.com/uw-biomedical-ml/hvfProgression","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"forecasting-future-humphrey-visual-fields","repo_url":"https://github.com/mohaEs/VFPrediction/tree/main/Competitor1","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"deep-learning","task_name":"Deep Learning"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}