Analysis on the efficacy of random forest model based on retinal blood flow density indicators in diagnosis of chronic kidney disease

LAN Lin; LEI Daizai; YANG Yingxia; HONG Yiyi; HU Xiangyu; YE Kun; XU Fan

doi:10.3969/j.issn.1674-3806.2024.12.13

Abstract：Objective To construct a random forest model for the diagnosis of chronic kidney disease(CKD)based on retinal blood flow density indicators to promote the early diagnosis and personalized prevention and treatment of CKD.Methods Twenty-seven patients with CKD(44 eyes)who were admitted to the People's Hospital of Guangxi Zhuang Autonomous Region from February 2022 to December 2023 were recruited as the case group,and 47 healthy people(71 eyes)were recruited as the control group during the same period.Retinal blood flow density indicators were obtained by using optical coherence tomography angiography(OCTA).A random forest model used for diagnosing CKD was constructed based on retinal blood flow density indicators and general clinical indicators,and the interpretability of the model was provided by Shapley importance analysis and Shapley method.Results The whole superficial capillary plexus(SCP)blood flow density and the lower half SCP blood flow density in the case group were higher than those in the control group,and the differences were statistically significant(P＜0.05),while there were no statistically significant differences in the other indicators of retinal blood flow density between the two groups(P＞0.05).The accuracy of the model was 86.96％,and the sensitivity of the model was 100.00％,and the specificity of the model was 72.73％.The upper half SCP blood flow density,the whole SCP blood flow density,Early Treatment Diabetic Retinopathy Study(ETDRS)region of the SCP blood flow density,the lower half SCP blood flow density,the whole deep capillary plexus(DCP)blood flow density,and the upper half DCP blood flow density were the main contributory indicators in the predictive model.Conclusion The random forest model based on retinal blood flow density indicators can effectively assist the diagnosis of CKD.