Cargando…

Geographic Atrophy Segmentation Using Multimodal Deep Learning

PURPOSE: To examine deep learning (DL)–based methods for accurate segmentation of geographic atrophy (GA) lesions using fundus autofluorescence (FAF) and near-infrared (NIR) images. METHODS: This retrospective analysis utilized imaging data from study eyes of patients enrolled in Proxima A and B (NC...

Descripción completa

Detalles Bibliográficos
Autores principales: Spaide, Theodore, Jiang, Jiaxiang, Patil, Jasmine, Anegondi, Neha, Steffen, Verena, Kawczynski, Michael G., Newton, Elizabeth M., Rabe, Christina, Gao, Simon S., Lee, Aaron Y., Holz, Frank G., Sadda, SriniVas, Schmitz-Valckenberg, Steffen, Ferrara, Daniela
Formato: Online Artículo Texto
Lenguaje:English
Publicado: The Association for Research in Vision and Ophthalmology 2023
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10341294/
https://www.ncbi.nlm.nih.gov/pubmed/37428131
http://dx.doi.org/10.1167/tvst.12.7.10
_version_ 1785072228093657088
author Spaide, Theodore
Jiang, Jiaxiang
Patil, Jasmine
Anegondi, Neha
Steffen, Verena
Kawczynski, Michael G.
Newton, Elizabeth M.
Rabe, Christina
Gao, Simon S.
Lee, Aaron Y.
Holz, Frank G.
Sadda, SriniVas
Schmitz-Valckenberg, Steffen
Ferrara, Daniela
author_facet Spaide, Theodore
Jiang, Jiaxiang
Patil, Jasmine
Anegondi, Neha
Steffen, Verena
Kawczynski, Michael G.
Newton, Elizabeth M.
Rabe, Christina
Gao, Simon S.
Lee, Aaron Y.
Holz, Frank G.
Sadda, SriniVas
Schmitz-Valckenberg, Steffen
Ferrara, Daniela
author_sort Spaide, Theodore
collection PubMed
description PURPOSE: To examine deep learning (DL)–based methods for accurate segmentation of geographic atrophy (GA) lesions using fundus autofluorescence (FAF) and near-infrared (NIR) images. METHODS: This retrospective analysis utilized imaging data from study eyes of patients enrolled in Proxima A and B (NCT02479386; NCT02399072) natural history studies of GA. Two multimodal DL networks (UNet and YNet) were used to automatically segment GA lesions on FAF; segmentation accuracy was compared with annotations by experienced graders. The training data set comprised 940 image pairs (FAF and NIR) from 183 patients in Proxima B; the test data set comprised 497 image pairs from 154 patients in Proxima A. Dice coefficient scores, Bland–Altman plots, and Pearson correlation coefficient (r) were used to assess performance. RESULTS: On the test set, Dice scores for the DL network to grader comparison ranged from 0.89 to 0.92 for screening visit; Dice score between graders was 0.94. GA lesion area correlations (r) for YNet versus grader, UNet versus grader, and between graders were 0.981, 0.959, and 0.995, respectively. Longitudinal GA lesion area enlargement correlations (r) for screening to 12 months (n = 53) were lower (0.741, 0.622, and 0.890, respectively) compared with the cross-sectional results at screening. Longitudinal correlations (r) from screening to 6 months (n = 77) were even lower (0.294, 0.248, and 0.686, respectively). CONCLUSIONS: Multimodal DL networks to segment GA lesions can produce accurate results comparable with expert graders. TRANSLATIONAL RELEVANCE: DL-based tools may support efficient and individualized assessment of patients with GA in clinical research and practice.
format Online
Article
Text
id pubmed-10341294
institution National Center for Biotechnology Information
language English
publishDate 2023
publisher The Association for Research in Vision and Ophthalmology
record_format MEDLINE/PubMed
spelling pubmed-103412942023-07-14 Geographic Atrophy Segmentation Using Multimodal Deep Learning Spaide, Theodore Jiang, Jiaxiang Patil, Jasmine Anegondi, Neha Steffen, Verena Kawczynski, Michael G. Newton, Elizabeth M. Rabe, Christina Gao, Simon S. Lee, Aaron Y. Holz, Frank G. Sadda, SriniVas Schmitz-Valckenberg, Steffen Ferrara, Daniela Transl Vis Sci Technol Retina PURPOSE: To examine deep learning (DL)–based methods for accurate segmentation of geographic atrophy (GA) lesions using fundus autofluorescence (FAF) and near-infrared (NIR) images. METHODS: This retrospective analysis utilized imaging data from study eyes of patients enrolled in Proxima A and B (NCT02479386; NCT02399072) natural history studies of GA. Two multimodal DL networks (UNet and YNet) were used to automatically segment GA lesions on FAF; segmentation accuracy was compared with annotations by experienced graders. The training data set comprised 940 image pairs (FAF and NIR) from 183 patients in Proxima B; the test data set comprised 497 image pairs from 154 patients in Proxima A. Dice coefficient scores, Bland–Altman plots, and Pearson correlation coefficient (r) were used to assess performance. RESULTS: On the test set, Dice scores for the DL network to grader comparison ranged from 0.89 to 0.92 for screening visit; Dice score between graders was 0.94. GA lesion area correlations (r) for YNet versus grader, UNet versus grader, and between graders were 0.981, 0.959, and 0.995, respectively. Longitudinal GA lesion area enlargement correlations (r) for screening to 12 months (n = 53) were lower (0.741, 0.622, and 0.890, respectively) compared with the cross-sectional results at screening. Longitudinal correlations (r) from screening to 6 months (n = 77) were even lower (0.294, 0.248, and 0.686, respectively). CONCLUSIONS: Multimodal DL networks to segment GA lesions can produce accurate results comparable with expert graders. TRANSLATIONAL RELEVANCE: DL-based tools may support efficient and individualized assessment of patients with GA in clinical research and practice. The Association for Research in Vision and Ophthalmology 2023-07-10 /pmc/articles/PMC10341294/ /pubmed/37428131 http://dx.doi.org/10.1167/tvst.12.7.10 Text en Copyright 2023 The Authors https://creativecommons.org/licenses/by/4.0/This work is licensed under a Creative Commons Attribution 4.0 International License.
spellingShingle Retina
Spaide, Theodore
Jiang, Jiaxiang
Patil, Jasmine
Anegondi, Neha
Steffen, Verena
Kawczynski, Michael G.
Newton, Elizabeth M.
Rabe, Christina
Gao, Simon S.
Lee, Aaron Y.
Holz, Frank G.
Sadda, SriniVas
Schmitz-Valckenberg, Steffen
Ferrara, Daniela
Geographic Atrophy Segmentation Using Multimodal Deep Learning
title Geographic Atrophy Segmentation Using Multimodal Deep Learning
title_full Geographic Atrophy Segmentation Using Multimodal Deep Learning
title_fullStr Geographic Atrophy Segmentation Using Multimodal Deep Learning
title_full_unstemmed Geographic Atrophy Segmentation Using Multimodal Deep Learning
title_short Geographic Atrophy Segmentation Using Multimodal Deep Learning
title_sort geographic atrophy segmentation using multimodal deep learning
topic Retina
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10341294/
https://www.ncbi.nlm.nih.gov/pubmed/37428131
http://dx.doi.org/10.1167/tvst.12.7.10
work_keys_str_mv AT spaidetheodore geographicatrophysegmentationusingmultimodaldeeplearning
AT jiangjiaxiang geographicatrophysegmentationusingmultimodaldeeplearning
AT patiljasmine geographicatrophysegmentationusingmultimodaldeeplearning
AT anegondineha geographicatrophysegmentationusingmultimodaldeeplearning
AT steffenverena geographicatrophysegmentationusingmultimodaldeeplearning
AT kawczynskimichaelg geographicatrophysegmentationusingmultimodaldeeplearning
AT newtonelizabethm geographicatrophysegmentationusingmultimodaldeeplearning
AT rabechristina geographicatrophysegmentationusingmultimodaldeeplearning
AT gaosimons geographicatrophysegmentationusingmultimodaldeeplearning
AT leeaarony geographicatrophysegmentationusingmultimodaldeeplearning
AT holzfrankg geographicatrophysegmentationusingmultimodaldeeplearning
AT saddasrinivas geographicatrophysegmentationusingmultimodaldeeplearning
AT schmitzvalckenbergsteffen geographicatrophysegmentationusingmultimodaldeeplearning
AT ferraradaniela geographicatrophysegmentationusingmultimodaldeeplearning