论文标题
使用深卷积神经网络的马X光片分类
Equine radiograph classification using deep convolutional neural networks
论文作者
论文摘要
目的：评估深卷积神经网络的能力，可以从赛马四肢的48个标准视图中对解剖位置和投影进行分类。 材料和方法：9504马前运动X光片用于训练，验证和测试六个深度学习体系结构，作为开源机器学习框架Pytorch的一部分。 结果：Resnet-34的前1位准确性为0.8408，大多数（88％）错误分类是由于横向性错误。类激活图表明关节形态推动了模型决策。 结论：深度卷积神经网络能够将马射线照相仪分类为48个标准视图，包括对侧向标记的中等歧视，独立于侧面标记。
Purpose: To assess the capability of deep convolutional neural networks to classify anatomical location and projection from a series of 48 standard views of racehorse limbs. Materials and Methods: 9504 equine pre-import radiographs were used to train, validate, and test six deep learning architectures available as part of the open source machine learning framework PyTorch. Results: ResNet-34 achieved a top-1 accuracy of 0.8408 and the majority (88%) of misclassification was because of wrong laterality. Class activation maps indicated that joint morphology drove the model decision. Conclusion: Deep convolutional neural networks are capable of classifying equine pre-import radiographs into the 48 standard views including moderate discrimination of laterality independent of side marker presence.
