[论文解读] Self-supervised Learning from 100 Million Medical Images

Florin C. Ghesu, Bogdan Georgescu+9 more|arXiv (Cornell University)|Jan 4, 2022

Radiomics and Machine Learning in Medical Imaging被引用 26

一句话总结

本文提出一种自监督学习方法，结合对比学习和在线特征聚类，在超过1亿张医疗影像上进行预训练，从而提升下游异常检测能力并加速收敛。

ABSTRACT

Building accurate and robust artificial intelligence systems for medical image assessment requires not only the research and design of advanced deep learning models but also the creation of large and curated sets of annotated training examples. Constructing such datasets, however, is often very costly -- due to the complex nature of annotation tasks and the high level of expertise required for the interpretation of medical images (e.g., expert radiologists). To counter this limitation, we propose a method for self-supervised learning of rich image features based on contrastive learning and online feature clustering. For this purpose we leverage large training datasets of over 100,000,000 medical images of various modalities, including radiography, computed tomography (CT), magnetic resonance (MR) imaging and ultrasonography. We propose to use these features to guide model training in supervised and hybrid self-supervised/supervised regime on various downstream tasks. We highlight a number of advantages of this strategy on challenging image assessment problems in radiography, CT and MR: 1) Significant increase in accuracy compared to the state-of-the-art (e.g., AUC boost of 3-7% for detection of abnormalities from chest radiography scans and hemorrhage detection on brain CT); 2) Acceleration of model convergence during training by up to 85% compared to using no pretraining (e.g., 83% when training a model for detection of brain metastases in MR scans); 3) Increase in robustness to various image augmentations, such as intensity variations, rotations or scaling reflective of data variation seen in the field.

研究动机与目标

• 通过使多模态大规模数据集的自监督特征学习，降低医学影像的标注成本。
• 开发一个带有在线聚类的对比学习框架，以生成可转移的表示供下游任务使用。
• 在胸部X线胸片异常检测、MRI脑转移灶检测以及CT脑出血检测上验证该方法。
• 展示对图像增强的鲁棒性并在有监督微调阶段提升收敛速度。

提出的方法

• 提出一种基于对比学习与在线特征聚类、并带有 swapped prediction 损失的自监督学习方法。
• 使用一个长度为 K 的可训练原型集合来将特征分配到簇中，并通过代码与原型相似度之间的交叉熵来优化。
• 引入单模态和多模态聚类策略，具有熵等分约束和 Sinkhorn-Knopp 归一化以防止退化解。
• 在混合预训练/微调目标中，将自监督损失与有监督损失结合，并通过平衡超参数来调控。
• 设计面向医学影像的特定增强，包括多尺度裁剪、基于能量的强度调整、Gamma及线性强度变换，以及随机裁剪。

实验结果

研究问题

• RQ1 基于100M+张医学影像的大规模自监督预训练，是否能在下游诊断任务中优于有监督预训练或无预训练？
• RQ2 多模态预训练是否能在放射影像、CT、MRI和超声之间产生更鲁棒、可迁移的特征？
• RQ3 所提出的基于聚类的对比预训练对收敛速度和对增强的鲁棒性在医学影像任务中的影响如何？

主要发现

方法	100%	50%	25%	10%
无预训练	0.77	0.73	0.65	0.53
SimCLR	0.90	0.88	0.82	0.79
SwAV	0.90	0.89	0.85	0.80
有监督 NI	0.91	0.89	0.82	0.80
我们的方案	0.94	0.91	0.85	0.85

• 在医学影像上的自监督预训练可显著提升AUC（平均提升6-8%）用于各任务。
• 下游训练收敛速度提升至最多85%，相较于无预训练，并且相对于有监督预训练有显著加速。
• 在自监督预训练后，对强度变化、旋转和缩放等图像增强具有更高的鲁棒性。
• 胸部X线异常检测、MRI脑转移灶检测以及CT脑出血检测均从该方法获益。