Manivannan Siyamalan, Cobb Caroline, Burgess Stephen, Trucco Emanuele
IEEE Trans Med Imaging. 2017 May;36(5):1140-1150. doi: 10.1109/TMI.2017.2653623. Epub 2017 Jan 16.
We propose a novel multiple instance learning method to assess the visibility (visible/not visible) of the retinal nerve fiber layer (RNFL) in fundus camera images. Using only image-level labels, our approach learns to classify the images as well as to localize the RNFL visible regions. We transform the original feature space to a discriminative subspace, and learn a region-level classifier in that subspace. We propose a margin-based loss function to jointly learn this subspace and the region-level classifier. Experiments with a RNFL dataset containing 884 images annotated by two ophthalmologists give a system-annotator agreement (kappa values) of 0:73 and 0:72 respectively, with an inter-annotator agreement of 0:73. Our system agrees better with the more experienced annotator. Comparative tests with three public datasets (MESSIDOR and DR for diabetic retinopathy, UCSB for breast cancer) show that our novel MIL approach improves performance over the state-of-the-art. Our Matlab code is publicly available at https://github.com/ManiShiyam/Sub-category-classifiersfor- Multiple-Instance-Learning/wiki.
我们提出了一种新颖的多实例学习方法，用于评估眼底相机图像中视网膜神经纤维层（RNFL）的可见性（可见/不可见）。仅使用图像级标签，我们的方法就能学习对图像进行分类，并定位RNFL可见区域。我们将原始特征空间转换为一个判别性子空间，并在该子空间中学习一个区域级分类器。我们提出了一种基于边缘的损失函数，以联合学习这个子空间和区域级分类器。对一个包含884张由两位眼科医生标注的图像的RNFL数据集进行实验，得到的系统与标注者的一致性（kappa值）分别为0.73和0.72，标注者之间的一致性为0.73。我们的系统与经验更丰富的标注者的一致性更好。与三个公共数据集（用于糖尿病视网膜病变的MESSIDOR和DR数据集、用于乳腺癌的UCSB数据集）进行的对比测试表明，我们新颖的多实例学习方法比现有技术性能更优。我们的Matlab代码可在https://github.com/ManiShiyam/Sub-category-classifiersfor- Multiple-Instance-Learning/wiki上公开获取。
