The Relationship Between Computerized Face and Tongue Image Segmentation and Metabolic Parameters in Patients with Type 2 Diabetes Based on Machine Learning
Song Wen,Yanyan Li,Chenglin Xu,Jianlan Jin,Zhimin Xu,Yue Yuan,Lijiao Chen,Yishu Ren,Min Gong,Congcong Wang,Meiyuan Dong,Yingfan Zhou,Xinlu Yuan,Fufeng Li,Ligang Zhou
DOI: https://doi.org/10.2147/DMSO.S491897
2024-10-29
Abstract:Objective: We aim to examine and reestablish the correlational and linear regression relationships, as well as the predictive value, between the significant facial and tongue features and the metabolic parameters in type 2 diabetes mellitus (T2DM). Materials and methods: From March to May 2024, we studied 269 patients with T2DM in the endocrinology department of Shanghai Pudong Hospital. The patients' facial and tongue characteristics were sampling by a tongue imaging device equipped with artificial intelligence (AI) (XiMaLife, Sinology, China) of automated and advanced machine learning algorithms. Then, the imaging features were examined in relation to the blood examination. Results: Multiple facial and tongue features, as well as dimensional facial and tongue color parameters, were significantly correlated with glycated hemoglobin A1c (HbA1c) (r < 0.3, p < 0.05), glycated albumin (GA) (-0.20 < 0.30, p < 0.05), C-peptide (-0.20.20, p < 0.05), plasma insulin (r < 0.30, p < 0.05), fasting plasma glucose (FPG) (r < 0.3, p < 0.05), significant hepatic and renal function indicators (-0.30 < r < 0.20, p<0.05), cardiac injury markers (-0.30 < r < 0.30, p < 0.05), tumor markers (-0.5 < r < 0.5, p < 0.05), thyroid function (-0.15 < r < 0.55, p < 0.05), and blood cell count, including white blood cells (r < 0.2, p < 0.05), and hemoglobin (Hb) (-0.30 < r < 0.3, 0.0001. The correlational results demonstrated that the tongue's characteristics and signs may be linked with the dynamic of the metabolic status of T2DM. In order to examine the causal relationships, we performed linear regression analyses, which revealed that various facial and tongue imaging parameters partially determined the metabolic indicators. The predictive value of imaging features was evaluated by receiver operating characteristic curve (ROC) to assess metabolic status in T2DM. Conclusion: This study demonstrated that metabolic status, renal and hepatic, cardiac, and thyroid function, the proportion of blood cells, and Hb in T2DM were intimately associated with facial and tongue features. The precise analysis of facial and tongue features through AI and advanced machine learning could be used to predict T2DM's conditions and progression.