| Objective:Osteoarthritis is the most common joint disease,is the leading cause of long-term disability,to individuals and society enormous economic burden.Due to the occurrence of osteoarthritis associated with many factors of aging,obesity,gender,smoking,genetic,inflammatory,bio-mechanics,leading to the pathogenesis of osteoarthritis remains unclear,there is no clear biomarkers of early osteoarthritis diagnosis.Early studies suggest that osteoarthritis is mainly based on the degeneration of articular cartilage as the main pathological changes.In recent years,a series of studies have suggested that osteoarthritis affects the entire joint involving cartilage,ligament,meniscus,subchondral bone.The structural changes in bone are usually preceded by changes in articular cartilage has been widely accepted.Texture analysis is a method of statistical analysis based on the distribution of gray(pixel)values and the spatial layout of an image or region of interest,texture analysis has the potential to identify subtle changes early in the disease.The application of texture analysis method to automatically analyze medical imaging images such as X-ray,CT,MRI to diagnose diseases has become a hot spot in the field of artificial intelligence.There are a few retrospective studies on texture analysis of subchondral bone to predict the occurrence of osteoarthritis,but few studies have systematically analyzed the changes of texture and the progression of arthritis.By establishing a spontaneous osteoarthritis model of guinea pig and analyzing the correlation between subchondral bone texture parameters and osteoarthritis,we screened the best combination of angle and step in the analysis of X-ray gray level co-occurrence matrix associating local binary pattern(LBP-GLCM)in the medial subchondral bone of guinea pig tibia,and use the LBP-GLCM texture eigenvalue fitting artificial neural network(ANN)pattern recognition model extracted from the subchondral bone X-ray image to differentiate osteoarthritis of different degrees.Meanwhile,analyze the trabecular parameters of subchondral,predicting the occurrence of knee osteoarthritis.Methods: First,the guinea pig knee osteoarthritis texture analysis 1.Based on the LBP-GLCM theory,45 guinea pigs of different age were examined by X-ray examination in vitro;2.Secial staining were conducted on knee joint of guinea pigs,and according to the pathological results,45 guinea pig of knee osteoarthritis will be divided into grade 0,1,2,3;3.The medial region of interest take from the tibial plateau of guinea pig were constructed under different angles θ and steps d of LBP-GLCM,and 6 texture parameters of subchondral bone were extracted.One-way ANOVA was used to analyze texture parameters.The ANN model was used to verify the different of osteoarthritis,and the effect of classification of osteoarthritis of guinea pigs by different LBP-GLCMs was tested by the “leave-one-case-out”.Second,the texture analysis to predict the occurrence of knee osteoarthritis 1.Subjects from the Osteoarthritis Initiative with no sign of radiographic OA at baseline were included.Cases that developed a global radiographic OA defined by the Kellgren-Lawrence scale,were compared with the controls with no changes after 36 months of follow-up;2.Baseline right knee subchondral were analyzed using LBP-GLCM to characterize the subchondral bone of tibial;3.The prediction was explored using logistic regression models evaluated by the area under the receiver operating characteristic curves.Results: First,the guinea pig knee osteoarthritis texture analysis 1.The pathological grading of 45 guinea pigs was: grade 0=20,grade 1=18,grade 2=7,grade 3=0;2.When the angle is 0 degree and the step is 2 pixels.Except for entropy,the variance analysis of other texture parameters had statistical significance(p<0.05);3.The results of interactive test showed that the ANN recognition model based on quantitative indicators of radiological texture got the accuracy of classification was 100.00%,100.00%,100.00% on the non-osteoarthritis group(grade 0),mild osteoarthritis group(grade 1),moderate osteoarthritis group(grade 2)respectively,and 100.00% of guinea pigs could be accurately graded.Second,the texture analysis to predict the occurrence of knee osteoarthritis 1.According to the conditions of Osteoarthritis Initiative database baseline and follow-up of 36 months of data screening,34 were included in the baseline imaging progression group,30 were included in the control group.There was no significant difference in the clinical characteristics of the two groups at baseline(gender,age,BMI,PASE);after 36 months of follow-up,there was no statistical difference in the clinical features between the two groups;2.Multivariate regression analysis,PASE is a risk factor for osteoarthritis,clinical covariates alone were not able to predict the occurrence of radiographic knee;3.The area under the curve of each texture parameter is 0.5-0.67;the area under the curve of combination of texture parameters is 0.67-0.7;the diagnosis odds ratio was in the range of 4-7.1;the area under the curve of texture parameters in combination with clinical covariates was 0.68-0.76,and the diagnostic odds ratio was 5.6-10.45.Conclusions: 1.The classification effect of the X-ray texture of the subchondral bone of the guinea pig obtained by LBP-GLCM was the best when the angle was 0 degree and the step was 2 pixels;2.The texture features based on LBP-GLCM combined with ANN model have very high accuracy;3.Texture analysis parameters assessed when radiographic signs are not yet apparent on radiographs may be useful in predicting the onset of radiological knee OA as well as identifying at-risk patients for future clinical trials. |
PDF Full Text Request