文献速递眼底病与葡萄膜炎

基于YOLO v3架构的深度学习模型在超广角眼底照相中检测和定位视网膜裂孔

DETECTION AND LOCALIZATION OF RETINAL BREAKS IN ULTRAWIDEFIELD FUNDUS PHOTOGRAPHY USING a YOLO v3 ARCHITECTURE-BASED DEEP LEARNING MODEL.
PubMed 文献 · Retina · 2022-01-01

本研究回顾性纳入未治疗过的视网膜裂孔或孔源性视网膜脱离患者,利用YOLO v3架构和迁移学习开发深度学习模型,在超广角眼底图像中检测和定位视网膜裂孔。共使用940例患者的4505张图像,其中84例患者的306张图像作为测试集。模型在测试集中每图像分类的AUC为0.957,准确率0.9085、敏感度0.8966、特异度0.9158;每目标检测的平均精度为0.840,最佳阈值下精确率0.6800、召回率0.9189、F1分数0.7816。

要点

  • 模型在测试集每图像分类的AUC为0.957,准确率0.9085、敏感度0.8966、特异度0.9158。
  • 每目标检测的平均精度为0.840,最佳阈值下精确率0.6800、召回率0.9189、F1分数0.7816。
  • 结论认为该超广角眼底图像深度学习模型在诊断和定位视网膜裂孔方面表现良好。

核心结论

  • 基于超广角眼底图像的深度学习模型在诊断和定位视网膜裂孔方面表现良好。

研究设计与样本

  • 回顾性纳入未治疗过的视网膜裂孔或孔源性视网膜脱离患者,且患者具有超广角眼底图像。
  • 使用YOLO v3架构骨干网络并通过迁移学习开发模型,采用每图像分类和每目标检测评估模型性能。
  • 共使用940例患者的4505张超广角眼底图像,其中84例患者的306张图像纳入测试集。

主要发现

  • 每目标检测中,考虑所有视网膜裂孔的物体检测模型平均精度为0.840。
  • 最佳阈值下,总体精确率为0.6800、召回率为0.9189、F1分数为0.7816。
  • 每图像分类中,模型在测试集内的受试者工作特征曲线下面积为0.957。
  • 测试数据集的总体准确率为0.9085、敏感度为0.8966、特异度为0.9158。

局限性

  • 原文未报告局限性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

PURPOSE: We aimed to develop a deep learning model for detecting and localizing retinal breaks in ultrawidefield fundus (UWF) images. METHODS: We retrospectively enrolled treatment-naive patients diagnosed with retinal break or rhegmatogenous retinal detachment and who had UWF images. The YOLO v3 architecture backbone was used to develop the model, using transfer learning. The performance of the model was evaluated using per-image classification and per-object detection. RESULTS: Overall, 4,505 UWF images from 940 patients were used in the current study. Among them, 306 UWF images from 84 patients were included in the test set. In per-object detection, the average precision for the object detection model considering every retinal break was 0.840. With the best threshold, the overall precision, recall, and F1 score were 0.6800, 0.9189, and 0.7816, respectively. In the per-image classification, the model showed an area under the receiver operating characteristic curve of 0.957 within the test set. The overall accuracy, sensitivity, and specificity in the test data set were 0.9085, 0.8966, and 0.9158, respectively. CONCLUSION: The UWF image-based deep learning model evaluated in the current study performed well in diagnosing and locating retinal breaks.

来源:Retina,查看原文。版权归原作者所有。

引用

  1. Oh R, Oh BL, Lee EK, et al. DETECTION AND LOCALIZATION OF RETINAL BREAKS IN ULTRAWIDEFIELD FUNDUS PHOTOGRAPHY USING a YOLO v3 ARCHITECTURE-BASED DEEP LEARNING MODEL. Retina. 2022. PMID: 36129265. doi:10.1097/IAE.0000000000003550

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。