CT image of pneumonia

The dataset includes three categories: CAP, COVID and NonCOVID, NonCOVID represents CT images of healthy lungs, with clear lung textures, no abnormal lines, and symmetrical distribution of the thoracic cage on both sides. CAP represents a CT image of common pneumonia, which may have alveolar consolidation, hyperdense opacities, and unclear lesion margins. COVID stands for CT image of COVID-19, and the lungs may have patchy or segmental ground-glass opacities, which may be gridded in texture, with a few interlobar pleural thickenings. The initial data ratio was not balanced, so the images were augmented with data, including random cropping, horizontal flipping and other methods, and finally about 6000 images per category, and the number of samples was basically balanced.本数据集涵盖三个类别：CAP、COVID与NonCOVID。其中NonCOVID指代健康肺部的CT影像，其肺部纹理清晰可见，无异常条索影，双侧胸廓呈对称分布。CAP代表普通肺炎的CT影像，可出现肺泡实变、高密度阴影及病变边界模糊等征象。COVID指代新型冠状病毒肺炎（COVID-19）的CT影像，肺部可出现斑片状或节段性磨玻璃影，纹理可呈网格状，伴少量叶间胸膜增厚。由于初始数据比例失衡，故采用随机裁剪、水平翻转等方法开展数据增强，最终每个类别约含6000张图像，样本数量基本均衡。