医学人工智能——“汇知”医学知识图谱数据

“汇知”医学知识图谱主要包含疾病、药品、检验检查和手术操作等几大类医学实体间的相关关系，关系类型涉及病因、诊断、治疗、预防、操作规范和患者教育等各方面，可以作为医学人工智能行业的底层知识库，为医学领域的语义检索（如文献检索与推荐），智能问答（如智能导诊与问诊）和辅助决策（如医疗决策预警和提示）等智能化场景提供底层的知识数据支撑，提升系统的智能化程度和可解释性。将医学知识源利用OCR识别、爬虫等技术整理形成半结构化文本，利用机器学习结合规则词典的方式，对文本中的疾病、检验检查、手术操作、药品等进行命名实体识别，并抽取这些实体类型之间的关系，然后利用词典、正则表达式和自定义规则等方法对数据进行清洗、标准化和去冗余，最后得到医学知识图谱数据。

The "Huizhi" Medical Knowledge Graph mainly covers the relational correlations among multiple categories of medical entities, including diseases, pharmaceuticals, laboratory tests and diagnostic examinations, and surgical procedures. Its relationship types involve various aspects such as etiology, diagnosis, treatment, prevention, operational specifications, and patient education. It can serve as the underlying knowledge base for the medical artificial intelligence industry, providing underlying knowledge and data support for intelligent scenarios in the medical field, including semantic retrieval (e.g., literature retrieval and recommendation), intelligent question answering (e.g., intelligent guided diagnosis and patient consultation), and auxiliary decision-making (e.g., medical decision warning and prompt), thereby enhancing the intelligence level and interpretability of relevant systems. The medical knowledge sources are first organized into semi-structured texts through technologies such as OCR recognition and web crawling. Subsequently, named entity recognition is conducted on entities like diseases, laboratory tests and diagnostic examinations, surgical procedures, and pharmaceuticals in the texts using machine learning combined with rule dictionaries, and the relational correlations between these entity types are extracted. Finally, data cleaning, standardization and redundancy removal are performed via methods including dictionaries, regular expressions and custom rules, resulting in the finalized medical knowledge graph data.