Construction of a prediction model for prognosis in patients with craniocerebral injuries based on CT imaging charac-teristics and clinical data

LIU Yunchao; LI Bin; ZHU Liangliang

doi:10.20258/j.cnki.1006-9011.2025.09.004

Abstract：Objective To explore the construction of a predictive model for the prognosis of patients with craniocerebral in-jury based on CT imaging characteristics and clinical data,in order to provide a reference for early clinical prediction of the prog-nosis of patients and targeted intervention plans.Methods A total of 225 patients with craniocerebral injury were selected.All patients were given surgical treatment.According to the disease outcome at 30 days after surgery,patients were assigned to a good prognosis group(n=168)and a poor prognosis group(n=57),the CT imaging characteristics and clinical data,in terms ofage,gender,body mass index,cause of injury,time from injury to treatment,injury type,injury severity,underlying dis-ease,Glasgow Coma Scale(GCS)scale,pupillary reflex,location of cerebral hematoma,white blood cell count(WBC),blood lactate(Lac),and hemoglobin(Hb)of the two groups,were compared.Logistic regression analysis and Lasso cross-validation were used to screen the factors influencing the prognosis of patients undergoing surgery for craniocerebral injuries.The relation-ship between CT imaging characteristics and poor prognosis in patients undergoing craniocerebral injury surgery was analyzed us-ing the column correlation coefficient.The R software rms program was used to construct a nomogram prognostic prediction model for patients undergoing craniocerebral injury surgery(referred to as the prognostic prediction model).The receiver operating characteristic(ROC)curve analysis was used to evaluate the prognostic prediction performance of the nomogram model for pa-tients undergoing craniocerebral injury surgery.Results There were significant differences in the time from injury to treatment,injury type,injury severity,proportion of diabetes,GCS score,pupillary reflex,and WBC of patients between the two groups(all P<0.05).There were significant differences in the basal cistern status,proportion of subarachnoid hemorrhage,midline shift distance,ventricular pressure,and Rotterdam CT score(all P<0.05).Logistic regression analysis showed that the time from in-jury to treatment,open injury,diabetes,severe brain injury,bilateral abnormal pupillary reflex,and high WBC count were in-dependent risk factor for poor prognosis in patients with craniocerebral injury,while,low GCS score was an independent protec-tive factor for the risk of poor surgical prognosis of patients with craniocerebral injury(OR=1.440,5.262,3.339,6.187,3.454,1.442,0.852,all P<0.05).Lasso cross-validation method(Lambda=0.019)was used to screen the variables,including five clinical indicators,i.e.,time from injury to treatment,type of injury,severity of injury,GCS score,and WBC,which were key factors affecting poor prognosis in patients with craniocerebral injury.The R software rms program was used to screen the final variables of five clinical factors,including the time from injury to treatment,injury type,injury severity,GCS score,and WBC based on the Lasso cross-validation method,which were used to construct a nomogram prognostic prediction model(M1 model)for patients undergoing surgery for craniocerebral injuries.Based on five CT imaging characteristics,including basal cistern sta-tus,subarachnoid hemorrhage,midline shift,ventricular compression,and Rotterdam CT score,combined with the aforemen-tioned five clinical indicators,a nomogram prognostic prediction model(M2 model)for patients undergoing craniocerebral injury surgery was constructed.The AUC of the M2 model for predicting the risk of poor prognosis in patients undergoing craniocerebral injury surgery was higher than that of the M1 model(Z=2.851,P=0.017).Conclusion Based on CT imaging characteristics and clinical data related to craniocerebral injuries,a prognostic prediction model for patients undergoing craniocerebral injury surgery has been successfully constructed,and this prediction model demonstrates good predictive performance for the prognosis of patients undergoing craniocerebral injury surgery.