School of Biomedical Informatics, The University of Texas Health Science Center at Houston, Houston, TX, USA.
Division of Dermatology, Washington University School of Medicine, St. Louis, MO, USA.
BMC Med Inform Decis Mak. 2023 Aug 18;23(Suppl 1):162. doi: 10.1186/s12911-023-02251-y.
The utilization of dermoscopic analysis is becoming increasingly critical for diagnosing skin diseases by physicians and even artificial intelligence. With the expansion of dermoscopy, its vocabulary has proliferated, but the rapid evolution of the vocabulary of dermoscopy without standardized control is counterproductive. We aimed to develop a domain-specific ontology to formally represent knowledge for certain dermoscopic features.
The first phase involved creating a fundamental-level ontology that covers the fundamental aspects and elements in describing visualizations, such as shapes and colors. The second phase involved creating a domain ontology that harnesses the fundamental-level ontology to formalize the definitions of dermoscopic metaphorical terms.
The Dermoscopy Elements of Visuals Ontology (DEVO) contains 1047 classes, 47 object properties, and 16 data properties. It has a better semiotic score compared to similar ontologies of the same domain. Three human annotators also examined the consistency, complexity, and future application of the ontology.
The proposed ontology was able to harness the definitions of metaphoric terms by decomposing them into their visual elements. Future applications include providing education for trainees and diagnostic support for dermatologists, with the goal of generating responses to queries about dermoscopic features and integrating these features to diagnose skin diseases.
皮肤科医生甚至人工智能在诊断皮肤疾病时越来越依赖于皮肤镜分析的应用。随着皮肤镜分析的扩展，其词汇量也在不断增加，但如果不对词汇进行标准化控制，这种快速发展的词汇反而会适得其反。我们旨在开发一个特定领域的本体，以正式表示特定皮肤镜特征的知识。
第一阶段涉及创建一个基本层次本体，涵盖描述可视化效果的基本方面和元素，如形状和颜色。第二阶段涉及利用基本层次本体来正式定义皮肤镜隐喻术语。
皮肤镜视觉元素本体（DEVO）包含 1047 个类、47 个对象属性和 16 个数据属性。与同领域的类似本体相比，它具有更好的符号学得分。三位人类注释员还检查了本体的一致性、复杂性和未来应用。
所提出的本体能够通过将隐喻术语分解为其视觉元素来利用它们的定义。未来的应用包括为学员提供教育和为皮肤科医生提供诊断支持，目标是生成有关皮肤镜特征的查询响应，并将这些特征集成到皮肤疾病的诊断中。
