The value of MRI combined with minimum apparent diffusion coefficient value in diagnosis and differentiation of primary central nervous system lymphoma

Abstract:

Objective：To investigate the value of conventional MRI signs combined with minimum apparent diffusion coefficient value（ADCmin） in the diagnosis and differentiation of primary central nervous system lymphoma（PCNSL） under logistic regression model. Methods：The conventional MRI characteristics and ADCmin of 34 cases of PCNSL confirmed by clinical and pathology were observed and analyzed. Another 36 cases of high-grade glioma（HGG） were selected as the control group. The differences in MRI characteristics and ADCmin between the two groups were statistically analyzed and compared，and logistic regression analysis was performed. The receiver operating characteristics（ROC） curve was used to analyze and compare the efficacy of single diagnosis and combined diagnosis for the two groups of tumors. Results：The proportions of cystic degeneration，necrosis and hemorrhage in PCNSL group were significantly lower than those in HGG group（?字2=25.2，P<0.001）. The proportion of PCNSL group in clear boundary，involving midline structure or（and） ventricle，uniform enhancement，existence of “notch sign” and（or） “sharp angle sign” and high signal signs on diffusion-weighted imaging（DWI） was higher than that of HGG group（all P<0.05）. There was no significant difference between the two groups in the degree of peritumoral edema and the degree of mass occupation（P >0.05）. The ADCmin of PCNSL group was significantly lower than that of HGG group（t=-7.962，P<0.001），while there was no statistical significance in the ADC value of the control side between the two groups（t=1.208，P=0.231）. When ADCmin was diagnosed separately，the optimal diagnostic cut-off point for the two groups was ADCmin=0.727×10-3 mm2/s，and the sensitivity and specificity of the two groups for differential diagnosis were 97.1% and 80.6%，respectively. The area under the curve was the largest（AUC=0.922） and the diagnostic efficiency was the highest. Logistic regression equation model was：Logistic（P）=15.269+3.963×uniform enhancement -24.695×ADCmin. Both ADCmin and homogeneous enhancement were influential factors for the diagnosis of PCNSL. When the optimal threshold of combined diagnosis was 0.46，the AUC of PCNSL was 0.977. The sensitivity and specificity of the two groups for differential diagnosis were 97.1% and 94.4%（P<0.001），with the highest efficacy of combined diagnosis. Conclusion：Routine MRI feature analysis combined with ADCmin can provide morphological and molecular imaging basis for non-invasive differentiation of PCNSL and HGG. Especially for the two tumors with overlapping manifestations of conventional MRI，the combined diagnosis under Logistic regression model can effectively improve the diagnostic efficiency，thus providing early diagnosis for patients with different tumors in the two groups，so as to provide reliable basis for early intervention，timely adjustment of treatment plan and judgment of prognosis value.