Construction and validation of a risk prediction model for postoperative delirium in elderly patients undergoing general anesthesia surgery based on cranial MRI radiomics and clinical features
YANG Li
ZHANG Jin
LI Beiping
AN Xiaolei
Abstract：Objective To construct and validate a risk prediction model for postoperative delirium(POD)in elderly patients undergoing general anesthesia surgery based on cranial magnetic resonance imaging(MRI)radiomics and various clinical characteristics.Methods Totally 300 elderly patients(aged 65 years and above)who underwent elective surger-ies requiring general anesthesia at Xuzhou First People's Hospital from January 2022 to December 2023 were enrolled.Fol-lowing the principle of random allocation,70%of the cases(210 cases)were designated as the training set,while the re-maining 30%(90 cases)constituted the validation set.Cranial scans were performed using a 3.0 T Siemens MR scanner to collect MRI features,including hippocampal volume,intracranial volume,and diffuse white matter hyperintensity(WMH)volume.Additionally,the clinical data were collected,and the occurrence of POD was assessed.MRI features and clinical data were compared between the training and validation sets.A preliminary screening of factors associated with POD was conducted using univariate analysis,and independent risk factors were further identified through multivari-ate Logistic regression analysis.Statistical analysis was conducted using R 4.2.2 software to construct and validate the pre-diction model.Results Among the 300 enrolled patients,120(42.00%)were diagnosed with POD,including 85(40.43%)in the training set and 35(39.32%)in the validation set.There were no statistically significant differences in the incidence of POD,MRI imaging characteristics,or clinical data between the training and validation sets.Univariate analysis indicated that the American Society of Anesthesiologists(ASA)classification of grade Ⅲ or higher,intraoperative hypotension,operation time,age,history of diabetes,postoperative Visual Analogue Scale(VAS)score,hippocampal volume,intracranial volume,and WMH volume were associated with POD(all P<0.05).Multivariate analysis revealed that advanced age,prolonged operative time,elevated postoperative VAS score,reduced hippocampal volume,increased WMH volume,and enlarged vascular lacunae were independent risk factors for POD(all P<0.05).The POD risk predic-tion model constructed based on these factors demonstrated robust predictive performance in both the training and valida-tion sets,with area under the ROC curve of 0.973 and 0.966,sensitivities of 0.959 and 0.971,and specificities of 0.913 and 0.884,respectively.Conclusions The POD nomogram prediction model for elderly patients undergoing sur-gery with general anesthesia,constructed using cranial MRI radiomics and clinical features,demonstrates statistically sig-nificant predictive and validation efficacy.This model can serve as a clinical decision support tool,assisting physicians in identifying high-risk patients,implementing preventive measures,and ultimately reducing the incidence of POD..
Keywords：postoperative deliriumMRI radiomicsrisk prediction modelnomogram
Publication Date：2025-07-25
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：8( 46-53 )