A Study On Correlation Analysis Of Quantitative Indicators And Prediction Of TNF-α Inhibitors Treatment Efficacy For Axial Spondyloarthritis Based On Multi-sequence MRI

Background Magnetic resonance imaging(MRI)provides objective indicators for evaluating the disease activity and efficacy of early axial spondyloarthritis(axSpA),which is marked by sacroiliitis.Previous studies have shown that both dynamic contrastenhanced(DCE)MRI and intravoxel incoherent motion diffusion-weighted imaging(IVIM-DWI)can quantitatively evaluate disease activity and guide the treatment of axSpA,but the correlations between the quantitative parameters derived from the two imaging methods have not been confirmed.Tumor necrosis factor-α(TNF-α)inhibitor is one of the common treatment options for axSpA,but the curative effect of some patients was poor after treatment and the efficacy of other biologics may be reduced.Therefore,it is important to identify patients who can benefit from TNF-α inhibitor prior to treatment.Objective 1.To explore the correlation between the quantitative parameters derived from IVIMDWI and DCE-MRI of sacroiliac joints in patients with axSpA,and explore the factors influencing the correlations.2.Using machine learning algorithm to construct models to predict the efficacy of TNF-α inhibitors on axSpA based on multimodal MRI-based Radiomics features and clinical features,and to compare the clinical application value of the prediction models.Methods 1.234 patients with axSpA were prospectively enrolled.IVIM-DWI and DCE-MRI images of all patients were collected and corresponding quantitative parameters were measured.Univariable and multivariable linear regression analyses were performed to evaluate independent correlations between the quantitative parameters derived from the two imaging methods,and to observe confounding factors affecting their correlation.Pearson’s correlation coefficients(r)were used to indicate the degree of correlations.2.Prospectively enrolled 183 axSpA patients treated with TNF-α inhibitors and had a reexamination within one year.Contrast-enhanced MRI of the sacroiliac joints,laboratory tests,and efficacy assessments based on four criteria were collected for all patients before treatment and at review.The efficacy prediction models based on machine learning were constructed by combining the radiomics features extracted from three MRI sequences with clinical features and semi-quantitative MRI scores.Area under the curve(AUC)and decision curve analysis(DCA)were used to compare the prediction performance of the models.Results 1.With controlling confounders,pure diffusion coefficient(D)was independently related to Ktrans(regression coefficient[b]=27.593,P<0.001),Kep(b=-6.707,P=0.021),and Ve(b=131.074,P=0.003),whereas perfusion fraction(f)and perfusion-related incoherent microcirculation(D*)had no independent correlation with the parameters derived from DCE-MRI.The correlations above were exhibited with Pearson’s correlation coefficients(r)(r=0.662,-0.408 and 0.396,respectively,all P<0.001).2.Among the four groups of models based on different efficacy evaluation criteria,the prediction performance of all clinical-radiomics combined models were higher than that of other types of models,among which the combined model based on multimodal MRI had the best prediction performance(AUC=0.810,0.697,0.807,0.762,respectively),and the corresponding DCA showed the highest clinical application value.Conclusion 1.There were independent correlations between D derived from IVIM-DWI and several quantitative parameters(Ktrans,Kep,and Ve)derived from DCE-MRI of sacroiliac joints in axSpA patients.The factors which affect the above correlations mainly included the presence of bone marrow edema(BME),gender,and scopes of lesions.2.The model combining the multimodal MRI-based Radiomics features and clinical features could help predict the efficacy of TNF-α inhibitors on axSpA more accurately,and has greater guiding significance in realizing personalized treatment for patients.