Hemelings Ruben, Elen Bart, Blaschko Matthew B, Jacob Julie, Stalmans Ingeborg, De Boever Patrick
Research Group Ophthalmology, KU Leuven, Herestraat 49, 3000 Leuven, Belgium; VITO NV, Boeretang 200, 2400 Mol, Belgium.
VITO NV, Boeretang 200, 2400 Mol, Belgium.
Comput Methods Programs Biomed. 2021 Feb;199:105920. doi: 10.1016/j.cmpb.2020.105920. Epub 2020 Dec 28.
Pathological myopia (PM) is the seventh leading cause of blindness, with a reported global prevalence up to 3%. Early and automated PM detection from fundus images could aid to prevent blindness in a world population that is characterized by a rising myopia prevalence. We aim to assess the use of convolutional neural networks (CNNs) for the detection of PM and semantic segmentation of myopia-induced lesions from fundus images on a recently introduced reference data set.
This investigation reports on the results of CNNs developed for the recently introduced Pathological Myopia (PALM) dataset, which consists of 1200 images. Our CNN bundles lesion segmentation and PM classification, as the two tasks are heavily intertwined. Domain knowledge is also inserted through the introduction of a new Optic Nerve Head (ONH)-based prediction enhancement for the segmentation of atrophy and fovea localization. Finally, we are the first to approach fovea localization using segmentation instead of detection or regression models. Evaluation metrics include area under the receiver operating characteristic curve (AUC) for PM detection, Euclidean distance for fovea localization, and Dice and F1 metrics for the semantic segmentation tasks (optic disc, retinal atrophy and retinal detachment).
Models trained with 400 available training images achieved an AUC of 0.9867 for PM detection, and a Euclidean distance of 58.27 pixels on the fovea localization task, evaluated on a test set of 400 images. Dice and F1 metrics for semantic segmentation of lesions scored 0.9303 and 0.9869 on optic disc, 0.8001 and 0.9135 on retinal atrophy, and 0.8073 and 0.7059 on retinal detachment, respectively.
We report a successful approach for a simultaneous classification of pathological myopia and segmentation of associated lesions. Our work was acknowledged with an award in the context of the "Pathological Myopia detection from retinal images" challenge held during the IEEE International Symposium on Biomedical Imaging (April 2019). Considering that (pathological) myopia cases are often identified as false positives and negatives in glaucoma deep learning models, we envisage that the current work could aid in future research to discriminate between glaucomatous and highly-myopic eyes, complemented by the localization and segmentation of landmarks such as fovea, optic disc and atrophy.
病理性近视（PM）是导致失明的第七大主要原因，据报道全球患病率高达3%。在近视患病率不断上升的世界人口中，从眼底图像中早期自动检测病理性近视有助于预防失明。我们旨在评估卷积神经网络（CNN）在最近引入的一个参考数据集上用于检测病理性近视以及对近视引起的病变进行语义分割的效用。
本研究报告了针对最近引入的病理性近视（PALM）数据集开发的卷积神经网络的结果，该数据集由1200张图像组成。我们的卷积神经网络将病变分割和病理性近视分类捆绑在一起，因为这两项任务紧密交织。还通过引入一种基于视盘（ONH）的新预测增强方法来插入领域知识，以用于萎缩分割和黄斑定位。最后，我们是首个使用分割而非检测或回归模型来进行黄斑定位的。评估指标包括用于病理性近视检测的受试者工作特征曲线下面积（AUC）、用于黄斑定位的欧几里得距离，以及用于语义分割任务（视盘、视网膜萎缩和视网膜脱离）的Dice和F1指标。
在400张可用训练图像上训练的模型，在400张图像的测试集上评估时，病理性近视检测的AUC为0.9867，黄斑定位任务的欧几里得距离为58.27像素。病变语义分割的Dice和F1指标在视盘上分别为0.9303和0.9869，在视网膜萎缩上分别为0.8001和0.9135，在视网膜脱离上分别为0.8073和0.7059。
我们报告了一种成功的方法，可同时对病理性近视进行分类并对相关病变进行分割。在2019年4月举行的IEEE国际生物医学成像研讨会期间举办的“从视网膜图像中检测病理性近视”挑战赛中，我们的工作获得了奖项。鉴于在青光眼深度学习模型中，（病理性）近视病例常被误判为假阳性和假阴性，我们设想当前的工作有助于未来的研究区分青光眼性眼和高度近视性眼，并通过黄斑、视盘和萎缩等标志物的定位和分割来辅助。
