Department of Ophthalmology, Otorhinolaryngology, and Dermatology of Korean Medicine, Kyung Hee University Hospital at Gangdong, Seoul, 05278, Republic of Korea.
Acupuncture and Meridian Science Research Center, College of Korean Medicine, Kyung Hee University, Seoul, 130-701, Republic of Korea; Department of Science in Korean Medicine, Graduate School, Kyung Hee University Korean Medicine Hospital, Seoul, 130-701, Republic of Korea.
Comput Biol Med. 2017 Aug 1;87:70-76. doi: 10.1016/j.compbiomed.2017.05.023. Epub 2017 May 22.
Syndrome differentiation (SD) results in a diagnostic conclusion based on a cluster of concurrent symptoms and signs, including pulse form and tongue color. In Korea, there is a strong interest in the standardization of Traditional Medicine (TM). In order to standardize TM treatment, standardization of SD should be given priority. The aim of this study was to explore the SD, or symptom clusters, of patients with atopic dermatitis (AD) using non-negative factorization methods and k-means clustering analysis. We screened 80 patients and enrolled 73 eligible patients. One TM dermatologist evaluated the symptoms/signs using an existing clinical dataset from patients with AD. This dataset was designed to collect 15 dermatologic and 18 systemic symptoms/signs associated with AD. Non-negative matrix factorization was used to decompose the original data into a matrix with three features and a weight matrix. The point of intersection of the three coordinates from each patient was placed in three-dimensional space. With five clusters, the silhouette score reached 0.484, and this was the best silhouette score obtained from two to nine clusters. Patients were clustered according to the varying severity of concurrent symptoms/signs. Through the distribution of the null hypothesis generated by 10,000 permutation tests, we found significant cluster-specific symptoms/signs from the confidence intervals in the upper and lower 2.5% of the distribution. Patients in each cluster showed differences in symptoms/signs and severity. In a clinical situation, SD and treatment are based on the practitioners' observations and clinical experience. SD, identified through informatics, can contribute to development of standardized, objective, and consistent SD for each disease.
辨证（SD）是根据同时出现的症状和体征（包括脉象和舌象）得出诊断结论。在韩国，人们对传统医学（TM）的标准化有着浓厚的兴趣。为了规范 TM 治疗，应优先规范 SD。本研究旨在使用非负矩阵分解方法和 k-均值聚类分析探讨特应性皮炎（AD）患者的 SD 或症状群。我们筛选了 80 名患者，纳入了 73 名符合条件的患者。一名 TM 皮肤科医生使用现有的 AD 患者临床数据集评估症状/体征。该数据集旨在收集与 AD 相关的 15 种皮肤病学和 18 种系统性症状/体征。使用非负矩阵分解将原始数据分解为具有三个特征和权重矩阵的矩阵。将每个患者的三个坐标的交点置于三维空间中。使用五个聚类，轮廓得分达到 0.484，这是从两个到九个聚类中获得的最佳轮廓得分。根据同时出现的症状/体征的严重程度对患者进行聚类。通过对 10000 次置换检验生成的零假设的分布，我们从分布的上 2.5%和下 2.5%的置信区间中找到了具有显著聚类特异性的症状/体征。每个聚类中的患者在症状/体征和严重程度上存在差异。在临床情况下，SD 和治疗是基于从业者的观察和临床经验。通过信息学确定的 SD 可以为每个疾病制定标准化、客观和一致的 SD 做出贡献。
