肝胆胰人机会诊平台双盲验证数据
收藏资源简介：
"基于肝胆胰疑难病的多模态数据集:本数据集全部源于301医院医疗信息系统，病历数据涵盖影像和病历文本两种数据类型，其中影像数据有CT和MRI，病历文本数据包含患者基本信息、入院记录等多方面内容，数据来源可靠。通过病历系统筛查符合要求的患者病历，涵盖肝癌、肝血管瘤、肝囊肿、胆囊结石、胆囊息肉、胆管囊肿、胰腺癌、胰腺炎等多种肝胆胰疑难疾病，充分保障样本多样性；同时，对数据进行去隐私处理，遵循相关法律法规和伦理原则，保护患者的隐私信息。 人工评价指标MOS分问卷:数据来源专业可靠:本数据集的问卷填写者为具备肝胆胰临床经验的医生，这些医生凭借其专业知识和临床经验，能够对人机融合会诊系统进行较为准确和专业的主观评价。问卷数据以线上方式进行，通过问卷星编辑问卷，并以二维码形式收集填写结果，确保了数据来源的明确性和可追溯性。参与调查的医生数量为23名，保证了一定的样本量，使得数据具有一定的代表性。 采集过程规范透明：使用问卷星编辑问卷，该平台具有成熟的问卷设计和数据收集功能，能够保证问卷的规范性和数据收集的准确性。以二维码形式收集问卷填写结果，操作简便且易于统计，同时这种方式也方便了参与对象填写问卷，提高了数据收集的效率。在数据收集过程中，严格遵循既定流程，确保数据的公正有效，并且对数据的采集过程进行记录，保证数据获取流程透明化。 双盲实例会诊结果:本数据集的测试病历来自于实际的医疗场景，由来自全国公立三级医院且具备3年及以上肝胆胰主治医师经验的医生团队通过腾讯线上会议的形式进行会诊，输出疾病诊断结果和治疗方案，数据来源清晰明确，真实反映了实际的会诊情况。并且由具备3年及以上肝胆胰专科主任/副主任医师经验的301医院专家对匿名处理后的结果进行标注，整个过程严格遵循数据采集的规范流程，保证数据获取流程透明化。专家的专业水平和临床经验具有权威性，也保证了数据标注的准确性和可靠性。 测试报告：本数据集来源于第三方测试机构出具的委托测试报告。第三方测试机构具有独立性和专业性，能够客观地对面向肝胆胰学科复杂病情的辩论型多方人机联合会诊平台软件以及人机联合团队和全医生团队在双盲实例会诊考核中的情况进行评估。委托指标明确，涵盖了软件本身以及双盲实例会诊考核中两者得分比较等关键内容，数据来源清晰且可靠，保证了数据的公正性和客观性。在数据提供过程中，遵循严格的流程和规范，确保数据的完整性和准确性。第三方测试公司在接收数据后，组织专家评审，整个采集和评审过程都有明确的记录和监督，保证数据获取流程透明化，使得数据能够真实反映考核的实际情况。"
Multimodal Dataset for Complex Hepatobiliary and Pancreatic Diseases: This entire dataset is sourced from the medical information system of the Chinese PLA General Hospital (301 Hospital). The medical record data covers two types: imaging data and medical record text data. The imaging data includes CT and MRI scans, while the medical record text data contains comprehensive information such as patients' basic personal details and admission records, with reliable data sources. Eligible patient medical records were screened via the hospital's medical record system, covering a variety of complex hepatobiliary and pancreatic diseases including liver cancer, hepatic hemangioma, hepatic cyst, cholelithiasis, gallbladder polyp, choledochal cyst, pancreatic cancer, and pancreatitis, which fully ensures the diversity of the samples. Meanwhile, the data has been de-identified to comply with relevant laws, regulations and ethical principles, protecting patients' private information. Mean Opinion Score (MOS) Questionnaire for Manual Evaluation Metrics: The data source is professional and reliable: the questionnaire respondents are physicians with clinical experience in hepatobiliary and pancreatic departments. Relying on their professional knowledge and clinical experience, these physicians can conduct relatively accurate and professional subjective evaluations on the human-AI collaborative consultation system. The questionnaire data was collected online: questionnaires were developed via Questionnaire Star, and responses were collected via QR codes, which ensures the clarity and traceability of the data source. A total of 23 physicians participated in the survey, ensuring a certain sample size and making the data representative to some extent. Standardized and Transparent Data Collection Process: Questionnaires were developed using Questionnaire Star, a platform with mature questionnaire design and data collection functions, which can guarantee the standardization of the questionnaire and the accuracy of data collection. Collecting responses via QR codes is easy to operate and count, and also facilitates respondents to fill in the questionnaire, improving the efficiency of data collection. During the data collection process, strict adherence to established procedures ensures the fairness and validity of the data, and the entire data collection process is recorded to ensure transparency of the data acquisition workflow. Double-blind Case Consultation Results: The test cases in this dataset are derived from actual medical scenarios. A team of physicians from national public tertiary hospitals with at least 3 years of experience as attending physicians in hepatobiliary and pancreatic departments conducted the consultation via Tencent Online Meeting, and output disease diagnosis results and treatment plans. The data source is clear and directly reflects the actual consultation situation. Additionally, experts from the Chinese PLA General Hospital (301 Hospital) with at least 3 years of experience as chief/associate chief physicians in hepatobiliary and pancreatic specialties annotated the anonymized results. The entire process strictly follows the standardized data collection procedures, ensuring transparency of the data acquisition workflow. The professional level and clinical experience of the experts are authoritative, which also guarantees the accuracy and reliability of the data annotation. Test Reports: This dataset is sourced from entrusted test reports issued by third-party testing institutions. Third-party testing institutions are independent and professional, and can objectively evaluate the debate-based multi-party human-AI collaborative consultation platform software for complex hepatobiliary and pancreatic diseases, as well as the performance of human-AI collaborative teams and full-physician teams in the double-blind case consultation assessment. The entrusted evaluation indicators are clear, covering key contents such as the software itself and the comparison of scores between the two teams in the double-blind case consultation assessment. The data source is clear and reliable, ensuring the fairness and objectivity of the data. During the data provision process, strict procedures and specifications are followed to ensure the integrity and accuracy of the data. After receiving the data, the third-party testing company organized expert reviews, and the entire collection and review process has clear records and supervision, ensuring transparency of the data acquisition workflow, so that the data can truly reflect the actual situation of the assessment.
