优化简化损伤量表严重度以提升损伤严重度评分预测性能的数据集

数据集获取（Figshare）

• 源站标题：Optimizing abbreviated injury scale severity using neural networks to enhance the predictive performance of injury severity scores
• DOI：10.6084/m9.figshare.29464496.v1
• 发布日期：2025-07-02
• 文件数：4 个
• 文件总体积：113.57 KB
• 源站页面：https://tandf.figshare.com/articles/dataset/Optimizing_abbreviated_injury_scale_severity_using_neural_networks_to_enhance_the_predictive_performance_of_injury_severity_scores/29464496

命令行下载

curl -L -o "gcpi_a_2518224_sm8829.xlsx" "https://ndownloader.figshare.com/files/55938863"
curl -L -o "gcpi_a_2518224_sm8828.pdf" "https://ndownloader.figshare.com/files/55938866"
curl -L -o "gcpi_a_2518224_sm8827.xlsx" "https://ndownloader.figshare.com/files/55938869"
curl -L -o "gcpi_a_2518224_sm8826.xlsx" "https://ndownloader.figshare.com/files/55938872"

文件清单

• gcpi_a_2518224_sm8829.xlsx（14.75 KB）
• gcpi_a_2518224_sm8828.pdf（75.84 KB）
• gcpi_a_2518224_sm8827.xlsx（12.93 KB）
• gcpi_a_2518224_sm8826.xlsx（10.06 KB）

文件总体积 113.57 KB（源站 API 实测）。

数据加载示例（图像类）

from PIL import Image
import glob, os

files = (glob.glob(os.path.join(path, "**", "*.png"), recursive=True)
       + glob.glob(os.path.join(path, "**", "*.jpg"), recursive=True)
       + glob.glob(os.path.join(path, "**", "*.tif"), recursive=True))
print("图像文件数:", len(files))
img = Image.open(files[0]); print("尺寸/模式:", img.size, img.mode)

# torchvision Dataset 方式:
# from torchvision import datasets
# ds = datasets.ImageFolder(path)  # 要求 子目录=类别

目录组织与标注格式以源站说明和下载后实际文件为准。

数据说明

To optimize Abbreviated Injury Scale (AIS) severity using neural networks to improve the predictive performance of the Injury Severity Score (ISS) for patient mortality.
Data were obtained from the Japan Trauma Data Bank (2019–2022). Cases involving cardiac arrest upon arrival and pediatric patients younger than 16 years were excluded. A single-layer perceptron neural network was implemented, with AIS 2008 codes as inputs, AIS severities as weights, the logistic function as the activation function, and survival or death as the dependent variable. Data from 2019 to 2021 were used as a derivation set, and data from 2022 were used as a test set. The neural network optimized the weights (AIS severities) using the derivation set, with updating them iteratively through backpropagation. Two sets of AIS severity were created: one in which severity levels were restricted between one and six (restricted AIS severity set) and another without such restrictions (non-restricted AIS severity set). After optimization, the updated severities were determined by rounding the absolute values of the weights to the nearest integer. Using the test data, ISSs were calculated based on the original and the two optimized severity sets, and their predictive performance was evaluated using the area under the receiver operating characteristic curve (AUROC).
Data from 107,349 cases (310,232 AIS codes) were analyzed. Among 1,998 AIS 2008 codes, 1,732 (86.7%) were represented in the dataset. ISSs based on the restricted and non-restricted AIS severity sets demonstrated significantly higher predictive performance than the ISS based on the original severity set (AUROC: 0.82 [95% CI, 0.81–0.83] for the restricted AIS severity set, 0.83 [95% CI, 0.82–0.84] for the non-restricted AIS severity set, vs. 0.78 [95% CI, 0.77–0.80] for the original set). Compared to the original AIS severity set, 177 and 231 codes had lower severity levels in the restricted and non-restricted sets, respectively, while 10 and 30 codes had higher severity levels.
AIS severities were successfully optimized using a neural network. This approach may support future AIS revisions by improving the accuracy of severity grading

许可

CC BY 4.0

引用

学术使用请引用 DOI 10.6084/m9.figshare.29464496.v1（source: https://tandf.figshare.com/articles/dataset/Optimizing_abbreviated_injury_scale_severity_using_neural_networks_to_enhance_the_predictive_performance_of_injury_severity_scores/29464496）。