基于多模态深度学习的地理萎缩分割。

Geographic Atrophy Segmentation Using Multimodal Deep Learning.

机构信息

Roche Personalized Healthcare, Genentech, Inc., South San Francisco, CA, USA.

Clinical Imaging Group, Genentech, Inc., South San Francisco, CA, USA.

出版信息

Transl Vis Sci Technol. 2023 Jul 3;12(7):10. doi: 10.1167/tvst.12.7.10.

DOI:10.1167/tvst.12.7.10

PMID:37428131

原文链接:https://pmc.ncbi.nlm.nih.gov/articles/PMC10341294/

Abstract

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.

摘要

目的

利用眼底自发荧光（FAF）和近红外（NIR）图像，研究基于深度学习（DL）的方法对地图状萎缩（GA）病变的精确分割。

方法

本回顾性分析使用了 Proxima A 和 B（NCT02479386；NCT02399072）中 GA 自然史研究中入组患者的研究眼成像数据。使用两种多模态 DL 网络（UNet 和 YNet）自动分割 FAF 上的 GA 病变；将分割准确性与经验丰富的分级员的注释进行比较。训练数据集由来自 Proxima B 的 183 例患者的 940 对图像（FAF 和 NIR）组成；测试数据集由来自 Proxima A 的 154 例患者的 497 对图像组成。使用 Dice 系数评分、Bland-Altman 图和 Pearson 相关系数（r）来评估性能。

结果

在测试集中，DL 网络与分级员比较的 Dice 评分在筛查就诊时范围为 0.89 至 0.92；分级员之间的 Dice 评分为 0.94。YNet 与分级员、UNet 与分级员以及分级员之间的 GA 病变面积相关性（r）分别为 0.981、0.959 和 0.995。从筛查到 12 个月（n = 53）的 GA 病变面积扩大的纵向相关性（r）分别为 0.741、0.622 和 0.890，与筛查时的横断面结果相比更低。从筛查到 6 个月（n = 77）的纵向相关性（r）甚至更低（0.294、0.248 和 0.686）。