Premium
Prediction of 3‐month treatment outcome of IgG4‐DS based on BP artificial neural network
Author(s) -
Shao Yanxiong,
Wang Zhijun,
Cao Ningning,
Shi Huan,
Xie Lisong,
Fu Jiayao,
Zheng Lingyan,
Yu Chuangqi
Publication year - 2021
Publication title -
oral diseases
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.953
H-Index - 87
eISSN - 1601-0825
pISSN - 1354-523X
DOI - 10.1111/odi.13601
Subject(s) - spearman's rank correlation coefficient , medicine , artificial neural network , reduction (mathematics) , rank correlation , correlation , correlation coefficient , matlab , mathematics , statistics , machine learning , computer science , geometry , operating system
Objective The study aimed to establish an effective back‐Propagation artificial neural network (BP‐ANN) model for automatic prediction of 3‐month treatment outcome of IgG4‐DS. Methods A total of 26 IgG4‐DS patients at Shanghai Ninth People's Hospital from January 2018 to December 2019 were involved in the study. They were all followed for >3 months. The primary outcome was reduction of serum IgG4 (sIgG4) after 3‐month treatment. The association between risk factors and reduction of sIgG4 was analyzed by Spearman's rank correlation test. According to the R values, we built a BP‐ANN model by MATLAB R2019b. Results The average reduction of sIgG4 was 5.55 ± 5.03. After Spearman's rank correlation test, ESR, sIgG4, and sIgG were independently associated with reduction of sIgG4 ( p < .05) and were selected as input variables. Take into account these parameters, BP‐ANN model was developed and the coefficient of determination (R 2 ) model was 0.95512. Conclusion The BP‐ANN model based on ESR, sIgG4, and sIgG could predict the 3‐month reduction of sIgG4 for IgG4‐DS patients. It showed potential clinical application value.