| Objective:Based on clinical conventional upper abdominal or chest CT scans,bone algorithm reconstruction was performed on L1 and L2,and vertebral bone microstructure measurements were performed using Image J.Combined with the results of QCT on vertebral bone density and paravertebral muscle fat content,a fracture risk prediction model for the elderly was established.Methods:A total of 231 patients aged 50 years or older who underwent CT scans of the upper abdomen or chest from endocrine patients in the imaging department emergency building and imaging building of the Affiliated Hospital of Guizhou Medical University during the period from July 2022 to January 2023 were collected according to the Napier criteria,with an age range of 50 to 97 years(64.65 ± 10.21).Among them,92 patients with a history of fractures were included in the fracture group(48 females and 44 males),139 patients without a history of fractures served as a control group(55 females and 84 males).The fracture group and the control group were randomly divided into a training queue and a test queue at a ratio of 7:3,respectively.The training queue was mainly used to establish a fracture risk prediction model,and the test queue was used to test the effectiveness of the fracture prediction model.Import the Dicom image data of CT images that meet the nanoplatoon standard into the QCT Pro post processing workstation to measure the vertebral body v BMD,and use the soft tissue functional component of the QCT analysis software that comes with it to measure the fat content in the posterior vertical spinal muscle of the L2 vertebral body.The included CT image data are all scanned on a CT machine corrected by the QCT Model 4 standard phantom.Import the Dicom data of L1 and L2 vertebral bodies reconstructed by bone window high-resolution reconstruction into an open source image processing software Image J,select the appropriate region of interest(ROI),perform image binarization on the ROI,and then use the Bone J2 plug-in in Image J to measure the bone trabecular microstructure within the ROI region.SPSS25.0 statistical software was used for data analysis.The measurement data were measured using mean ± standard deviation(x ± S),and independent samples were used to compare the differences between groups using Ttest;Chi-square test was used to compare the differences between groups in classified data.The AUC values between models were compared using the Delong test.The difference was considered statistically significant(p<0.05).RStudio software was used to establish a fracture risk prediction model by using decision tree classification model and Logistic regression analysis model to measure variables such as bone microstructure,bone density,and clinical biochemical indicators.The reliability of the fracture risk prediction model was tested using the Hosmer and Lemes how good of fit(GOF)test,and the predictive efficacy of the two models was compared by comparing their ROC curves and AUC measurements Results:In this study,the overall age of the fracture group was greater than that of the control group,and the proportion of women in the fracture group was higher(p<0.001).After adjusting for patient age and gender,it was found that there was no statistical difference between the two groups;The vertebral body v BMD of the fracture group was significantly lower than that of the control group(p<0.001),and the intramuscular fat content of the vertical spine in the fracture group was higher than that of the control group(p=0.004).In the comparison of the differences in trabecular microstructure between the fracture group and the control group,only the differences in bone volume fraction(BV/TV)and trabecular spacing(Th.Sp)were statistically significant(p<0.001),while there were no significant differences in other variables such as DA,Th.Tb,EF,and Conn.D(Table 2).The correlation between bone microstructure and bone density,paravertebral muscle fat content,and bone metabolism biochemical indicators is intuitively reflected in the heat map,which shows a good correlation between each bone trabecular microstructure and the above factors(Figure 4).Establishing a fracture risk prediction model through Logistic regression analysis has good performance in predicting fracture risk in the elderly(X~2= 3.476,df=8,p=0.901),its predictive performance in the test queue is as follows(accuracy 0.74,accuracy 0.81,specificity 0.75),and the AUC value for predicting vertebral osteoporotic fractures in the elderly is 0.749.Conclusions:Through routine chest and abdominal CT scans,the vertebral bone microstructure obtained has a great correlation with the vertebral bone density measured by QCT.The fracture risk prediction model established in this study based on the measurement of bone microstructure in L1 and L2 vertebral body bone window reconstruction images after clinical routine chest and abdominal CT scans has great predictive ability for osteoporosis fractures in middle-aged and elderly people.In the current era of great health in New China,people’s awareness of health examinations is gradually increasing,and with the popularization of CT scans,the use of routine CT images of patients’ chest and abdomen for opportunistic screening of osteoporosis fractures in middle-aged and elderly patients has great potential value. |
PDF Full Text Request