四川省HER2型乳腺癌辅助诊断模型训练数据

通过对样本的数据处理和数据加工，提供给辅助诊断人工智能模型进行训练，帮助人工智能模型更好地理解四川省样本场景下HER2型乳腺癌的情况，提取特征，发现规律，最终提高诊断人工智能模型的准确性、鲁棒性和泛化能力。1数据采集：通过正式合作协议，从医疗机构取得匿名化的样本临床数据，包括是否有术后病理结果，术后Her2情况，术后fish情况；2数据处理：对数据进行检查核对，确保所有数据去标识化，处于完全匿名化状态且不可还原的状态，将没有病理结果的数据去除，对异常数据进行清洗去除，对部分缺失数据进行生成式补充；3数据加工：基于原始数据以及算法规则HER2型乳腺癌的术后状态，生成阴性阳性分型标记，具体规则为：如果Her2满足3+或Her2满足2+且同时术后Fish为扩增则标记为阳性，其余标记为阴性。

This dataset is developed for training AI-assisted diagnostic models through sample data processing and curation, aiming to help the models better comprehend the scenario of HER2-positive breast cancer in the cohort from Sichuan Province, extract features and discover patterns, thereby improving the accuracy, robustness and generalization capability of the diagnostic AI models.
1. Data Collection: Anonymized clinical sample data is obtained from medical institutions via formal cooperation agreements, including the availability of postoperative pathological results, postoperative HER2 status, and postoperative FISH results.
2. Data Processing: The data is inspected and verified to ensure all records are de-identified, fully anonymized and irreversibly untraceable; data without postoperative pathological results is removed, outlier data is cleaned, and partially missing data is generatively imputed.
3. Data Curation: Positive/negative classification labels for the postoperative status of HER2-positive breast cancer are generated based on the raw data and algorithmic rules. The specific labeling rules are: Label a sample as positive if its HER2 status is 3+, or its HER2 status is 2+ with amplified postoperative FISH results; otherwise, label it as negative.