Skip to main content
QUICK REVIEW

[论文解读] Viral Pneumonia Screening on Chest X-ray Images Using Confidence-Aware Anomaly Detection

Jianpeng Zhang, Yutong Xie|arXiv (Cornell University)|Mar 27, 2020
COVID-19 diagnosis using AI参考文献 53被引用 132
一句话总结

本论文将胸部X线图中的病毒性肺炎检测框架化为单类异常检测问题,并引入 CAAD,它结合异常评分与置信预测器以提升筛查病毒性肺炎的灵敏度,包括对COVID-19等未见病原体。

ABSTRACT

Cluster of viral pneumonia occurrences during a short period of time may be a harbinger of an outbreak or pandemic, like SARS, MERS, and recent COVID-19. Rapid and accurate detection of viral pneumonia using chest X-ray can be significantly useful in large-scale screening and epidemic prevention, particularly when other chest imaging modalities are less available. Viral pneumonia often have diverse causes and exhibit notably different visual appearances on X-ray images. The evolution of viruses and the emergence of novel mutated viruses further result in substantial dataset shift, which greatly limits the performance of classification approaches. In this paper, we formulate the task of differentiating viral pneumonia from non-viral pneumonia and healthy controls into an one-class classification-based anomaly detection problem, and thus propose the confidence-aware anomaly detection (CAAD) model, which consists of a shared feature extractor, an anomaly detection module, and a confidence prediction module. If the anomaly score produced by the anomaly detection module is large enough or the confidence score estimated by the confidence prediction module is small enough, we accept the input as an anomaly case (i.e., viral pneumonia). The major advantage of our approach over binary classification is that we avoid modeling individual viral pneumonia classes explicitly and treat all known viral pneumonia cases as anomalies to reinforce the one-class model. The proposed model outperforms binary classification models on the clinical X-VIRAL dataset that contains 5,977 viral pneumonia (no COVID-19) cases, 18,619 non-viral pneumonia cases, and 18,774 healthy controls.

研究动机与目标

  • 推动在胸部X光片上进行快速病毒性肺炎筛查,以帮助疫情监测与控制。
  • 通过避免显式的多类别病毒标签来解决数据集偏移和极端类别不平衡问题。
  • 提出一个以置信预测器增强的单类异常检测框架,以提升灵敏度。

提出的方法

  • 共享特征提取器(EfficientNet-B0)输入一个异常检测模块,该模块通过一个 3x100 神经元的多层感知机输出异常分数。
  • 使用参考高斯分数的对比损失以将异常与正常样本分离。
  • 一个独立的置信预测网络估计每张图像的置信度,并通过概率异常度量进行校准。
  • 推断阶段对异常分数 T_ano 和置信度 T_conf 设定阈值以判定阳性/阴性。
  • 端到端联合训练,先在 ImageNet 上进行预训练,然后进行三阶段训练(异常检测、置信预测、联合微调)。
  • 关键方程包括异常分数 phi(x; theta, alpha),对比损失 L_ano,以及用于失败预测的置信公式 prob 和 g。

实验结果

研究问题

  • RQ1是否可以将病毒性肺炎有效地检测为异常,而不是通过二元的病毒性与非病毒性分类?
  • RQ2将显式的置信预测模块耦合是否能提高筛查灵敏度并识别潜在失败?
  • RQ3在没有微调的情况下,CAAD 对未见病毒性肺炎病例(例如 COVID-19)的泛化能力有多强?
  • RQ4在临床筛查中,异常分数阈值与置信阈值之间有哪些权衡?

主要发现

ModeFeature extractorAccuracySensitivitySpecificityAUC
Binary classificationResNet78.5278.2878.5686.24
Anomaly detectionResNet80.0484.4479.3487.18
Binary classificationEfficientNet78.7179.0978.6586.30
Anomaly detectionEfficientNet80.6585.5179.8787.42
  • CAAD 在 X-VIRAL 数据集上实现了最先进的 AUC(87.57% 且带置信预测)并在某些置信阈值下达到很高的灵敏度(最高 93.01%)。
  • 异常检测在 X-VIRAL 数据集的 AUC 和灵敏度方面优于二元分类基线(例如 EfficientNet-B0:AUC 87.42% 对比最佳二元模型 87.18%)。
  • 置信度学习提升了失败预测能力:异常概率比预测概率更有效地区分正确与错误的预测。
  • 在未见的 X-COVID 数据(COVID-19 病例对比正常)上,CAAD 达到 AUC 83.61% 和灵敏度 71.70%,与文献中的放射科医生相当。
  • 模型在初始训练阶段没有显式的 COVID-19 训练数据也能保持稳健性能,表明对疫情筛查具有良好泛化能力。

更好的研究,从现在开始

从论文设计到论文写作,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。