Lung Masks for Shenzhen Hospital Chest X-ray Set 深圳医院胸腔X光组肺口罩

由于价格相对便宜，便于获得胸部X光（CXR）成像被广泛用于许多肺部疾病（肺炎、肺结核、癌症等）的健康监测和诊断。CXR对这些疾病的标记进行人工分析和检测，由放射科专家进行，这是一个漫长而复杂的过程。然而，通用图形处理卡 （GPU） 硬件 （1）和用于医学图像分析的软件（2） 的现代演变，特别是深度学习技术（3），使科学家能够自动检测 CXR 图像中的许多肺部疾病，其水平超过经过认证的放射科医生（4）。尽管取得了这些成功，但专家们坚信，深度学习技术对于非常大的数据集（+10，000 张图像）来说会变得高效，因为对于较小的数据集（+lt;1000 图像），它们会产生非常低的适应率的不良预测（如果有的话）。

Chest X-ray (CXR) imaging, which is relatively inexpensive and readily accessible, is widely utilized for health monitoring and diagnosis of numerous pulmonary diseases such as pneumonia, tuberculosis, lung cancer, and others. Manual analysis and detection of disease-related markers in CXR images, performed by radiologists, constitutes a lengthy and complex workflow. However, modern advancements in general-purpose graphics processing unit (GPU) hardware (1) and medical image analysis software (2), particularly deep learning technologies (3), have enabled researchers to automatically identify a wide range of pulmonary abnormalities from CXR images at performance levels exceeding those of certified radiologists (4). Despite these notable successes, experts maintain that deep learning models only achieve optimal efficiency when trained on extremely large datasets (containing more than 10,000 images); for smaller datasets (with fewer than 1,000 images), they typically produce poor predictions with very low adaptation rates, if any valid predictions are generated at all.

搜集汇总

背景与挑战

背景概述

该数据集为深圳医院胸腔X光组肺口罩数据集，包含手动分割的肺口罩，用于支持深度学习技术在胸部X光图像分析中的应用，特别是在小数据集情况下提高肺部疾病（如肺结核）的预测准确性。数据集还结合了数据增强技术，以提升统计可靠性。