中医舌脉诊标注数据

采用满足国家医疗器械标准的设备采集舌体图像数据、脉象波形数据。由医生对数据进行标注和分析，获取舌体特征信息、脉型判断，进而生成规范化的中医舌脉数据与特征的结构化报告，涉及以下步骤和算法规则：1、数据预处理，由医生对原始采集到的舌图像、脉诊数据进行清洗，按照纳排标准去掉不符合标准的数据，例如图片不清晰、伸舌动作不标准、脉象数据不稳定等数据需要删除。2、基于神经网络模型的特征提取：利用多标签分类网络处理舌图数据与脉象数据，获取初步的舌图健康特征与脉型信息，例如：红舌、裂纹舌、点刺舌、滑脉、涩脉等信息。3、健康特征校验：由三名取得中医执业医师资格证的医生对数据进行校准，校准规则为三名医师至少有两名对分析特征认可后，数据才可纳入数据集。4、生成结构化的数据报告：按照json的文件格式，将数据内容、数据标签存储起来，其中舌图图像数据以jpg格式文件的形式存储。文件内不涉及任何人员信息。5、质量控制：对生成的结构化报告进行质量控制，确保信息的准确性和完整性。8、持续优化和扩充数据集：根据数据集的应用反馈，持续改进数据集的数据量，单例数据包含舌脉特征数量，舌脉特征分析的准确度等信息。

Tongue body image data and pulse waveform data were collected via equipment compliant with national medical device standards. Doctors annotated and analyzed the collected data to extract tongue body feature information and determine pulse types, then generated standardized structured reports of traditional Chinese medicine (TCM) tongue-pulse data and features, involving the following steps and algorithmic rules:
1. Data Preprocessing: Doctors clean the originally acquired tongue images and pulse diagnosis data, and eliminate non-compliant data per inclusion and exclusion criteria. For instance, data with unclear images, non-standard tongue protrusion movements, or unstable pulse waveform data shall be deleted.
2. Feature Extraction via Neural Network Models: A multi-label classification network is employed to process the tongue image data and pulse waveform data, to obtain preliminary health features of tongue images and pulse type information, such as red tongue, cracked tongue, petechial tongue, slippery pulse, hesitant pulse, and other related features.
3. Health Feature Verification: Three physicians holding the practicing Chinese medicine physician qualification certificate calibrate the data. The calibration rule stipulates that data can only be included in the dataset when at least two of the three physicians approve the analyzed features.
4. Structured Data Report Generation: Store the data content and labels in JSON file format. Tongue image data is stored as JPG format files. No personal identifiable information is included in the files.
5. Quality Control: Perform quality control on the generated structured reports to ensure the accuracy and integrity of the information.
8. Continuous Optimization and Expansion of the Dataset: Continuously enhance the dataset scale, the number of tongue-pulse features per single case, the accuracy of tongue-pulse feature analysis, and other relevant information based on the application feedback of the dataset.

搜集汇总

数据集介绍

特点

中医舌脉诊标注数据包含1008条结构化数据，涵盖舌图图像、脉象数据及特征信息，适用于中医辅助诊疗、教学和科研。数据通过标准化采集和标注流程生成，经过质量控制，旨在提升中医诊断的准确性和一致性。