[论文解读] COVIDLite: A depth-wise separable deep neural network with white balance and CLAHE for detection of COVID-19

Manu Siddhartha, Avik Santra|arXiv (Cornell University)|Jun 18, 2020

COVID-19 diagnosis using AI被引用 36

一句话总结

COVIDLite 将白平衡与 CLAHE 预处理结合使用，并结合深度可分离卷积神经网络来从胸部 X 光片检测 COVID-19，达到高准确性和轻量级部署适用性。

ABSTRACT

Background and Objective:Currently, the whole world is facing a pandemic disease, novel Coronavirus also known as COVID-19, which spread in more than 200 countries with around 3.3 million active cases and 4.4 lakh deaths approximately. Due to rapid increase in number of cases and limited supply of testing kits, availability of alternative diagnostic method is necessary for containing the spread of COVID-19 cases at an early stage and reducing the death count. For making available an alternative diagnostic method, we proposed a deep neural network based diagnostic method which can be easily integrated with mobile devices for detection of COVID-19 and viral pneumonia using Chest X-rays (CXR) images. Methods:In this study, we have proposed a method named COVIDLite, which is a combination of white balance followed by Contrast Limited Adaptive Histogram Equalization (CLAHE) and depth-wise separable convolutional neural network (DSCNN). In this method, white balance followed by CLAHE is used as an image preprocessing step for enhancing the visibility of CXR images and DSCNN trained using sparse cross entropy is used for image classification with lesser parameters and significantly lighter in size, i.e., 8.4 MB without quantization. Results:The proposed COVIDLite method resulted in improved performance in comparison to vanilla DSCNN with no pre-processing. The proposed method achieved higher accuracy of 99.58% for binary classification, whereas 96.43% for multiclass classification and out-performed various state-of-the-art methods. Conclusion:Our proposed method, COVIDLite achieved exceptional results on various performance metrics. With detailed model interpretations, COVIDLite can assist radiologists in detecting COVID-19 patients from CXR images and can reduce the diagnosis time significantly.

研究动机与目标

• 在 RT-PCR 资源有限时，通过胸部 X 光实现快速、可及的 COVID-19 筛查之动机。
• 开发适用于移动/网页部署的轻量级 CNN 架构。
• 通过有针对性的预处理提升 CXR 图像质量，以改善 CNN 特征提取。
• 提供模型可解释性，利用 Grad-CAM、LIME 和显著性图帮助放射科医师。

提出的方法

• 将白平衡后再进行 CLAHE 作为预处理流水线，以提升 CXR 可见度。
• 采用深度可分离卷积 CNN (DSCNN) 以减少参数与计算，与标准 CNN 相比。
• 使用 3x3 的核对提出的 COVIDLite 网络进行训练，网络共有 18 层（2 个标准卷积 + 12 个 DSC 层 + 稠密块）。
• 在输入网络前将图像归一化并调整为 224x224x3。
• 使用带 softmax 激活的稀疏分类交叉熵损失进行多类别分类。
• 使用 5 折交叉验证进行验证，并报告标准指标（准确率、精确度、召回率、F1、AUC、Cohen 的 kappa）。
• 评估模型可解释性，使用 Grad-CAM、LIME 和显著性图来突出肺炎相关区域。

实验结果

研究问题

• RQ1 白平衡 + CLAHE 预处理如何影响来自 CXR 图像的 COVID-19 检测性能？
• RQ2 轻量级的深度可分离 CNN 是否能在 CXR 数据上实现有竞争力的 COVID-19 分类准确度（2 类和 3 类）？
• RQ3 预处理启用与禁用的流水线之间在准确度、精确度、召回率、AUC、kappa 等性能权衡？
• RQ4 Grad-CAM、LIME 与显著性图在多大程度上能为模型的决策提供临床意义的解释？

主要发现

• 采用预处理的二分类准确率：98.75%（5 折平均）。
• 采用预处理的三分类准确率：99.58%（5 折平均）。
• 类别级别的性能（带预处理）显示：Precision 99.00%（Normal）/ 100.00%（COVID-19）/ 96.00%（Viral Pneumonia）；Sensitivity 99.00% / 99.58% / 96.00%；Specificity 98.25% / 99.34% / 97.79%；F1-Score 99.00% / 99.79% / 95.00%；AUC 1.0 / 1.0 / 0.99。
• 二分类的 AUC：COVID-19 和 Normal 类的平均 AUC = 1.0；多分类下，类别级 AUC 为 1.0（COVID-19），其他为 0.99–1.0。
• 模型大小轻量级（未量化时 8.4 MB），适合移动/网页部署；预处理提升性能并缩窄 95% 置信区间。