University of Maryland Intelligent Imaging Center, Department of Radiology, University of Maryland School of Medicine, Baltimore, MD, USA.
The Russell H. Morgan Department of Radiology and Radiological Science, Johns Hopkins University School of Medicine, Baltimore, MD, USA.
Emerg Radiol. 2021 Oct;28(5):949-954. doi: 10.1007/s10140-021-01953-y. Epub 2021 Jun 5.
To develop and test the performance of deep convolutional neural networks (DCNNs) for automated classification of age and sex on chest radiographs (CXR).
We obtained 112,120 frontal CXRs from the NIH ChestX-ray14 database performed in 48,780 females (44%) and 63,340 males (56%) ranging from 1 to 95 years old. The dataset was split into training (70%), validation (10%), and test (20%) datasets, and used to fine-tune ResNet-18 DCNNs pretrained on ImageNet for (1) determination of sex (using entire dataset and only pediatric CXRs); (2) determination of age < 18 years old or ≥ 18 years old (using entire dataset); and (3) determination of age < 11 years old or 11-18 years old (using only pediatric CXRs). External testing was performed on 662 CXRs from China. Area under the receiver operating characteristic curve (AUC) was used to evaluate DCNN test performance.
DCNNs trained to determine sex on the entire dataset and pediatric CXRs only had AUCs of 1.0 and 0.91, respectively (p < 0.0001). DCNNs trained to determine age < or ≥ 18 years old and < 11 vs. 11-18 years old had AUCs of 0.99 and 0.96 (p < 0.0001), respectively. External testing showed AUC of 0.98 for sex (p = 0.01) and 0.91 for determining age < or ≥ 18 years old (p < 0.001).
DCNNs can accurately predict sex from CXRs and distinguish between adult and pediatric patients in both American and Chinese populations. The ability to glean demographic information from CXRs may aid forensic investigations, as well as help identify novel anatomic landmarks for sex and age.
开发和测试深度卷积神经网络（DCNN）在胸部 X 光片（CXR）上自动分类年龄和性别的性能。
我们从 NIH ChestX-ray14 数据库中获得了 112120 张正面 CXR，其中 48780 名女性（44%）和 63340 名男性（56%）年龄在 1 至 95 岁之间。数据集分为训练集（70%）、验证集（10%）和测试集（20%），用于微调在 ImageNet 上预训练的 ResNet-18 DCNN，以（1）确定性别（使用整个数据集和仅儿科 CXR）；（2）确定年龄<18 岁或≥18 岁（使用整个数据集）；和（3）确定年龄<11 岁或 11-18 岁（仅使用儿科 CXR）。在中国对 662 张 CXR 进行了外部测试。接收器工作特征曲线下的面积（AUC）用于评估 DCNN 的测试性能。
在整个数据集和儿科 CXR 上训练用于确定性别的 DCNN 的 AUC 分别为 1.0 和 0.91（p<0.0001）。在整个数据集和儿科 CXR 上训练用于确定年龄<18 岁或≥18 岁和<11 岁与 11-18 岁的 DCNN 的 AUC 分别为 0.99 和 0.96（p<0.0001）。外部测试显示，性别为 0.98（p=0.01），年龄<18 岁或≥18 岁为 0.91（p<0.001）。
DCNN 可以从 CXR 中准确预测性别，并区分美国和中国人群中的成年和儿科患者。从 CXR 中提取人口统计信息的能力可能有助于法医调查，以及帮助确定性别和年龄的新解剖标志。
