OBJECTIVE: Our purpose was to evaluate the diagnostic performance of diffusion-weighted imaging, the relative minimum apparent diffusion coefficient (rADCmin) in differentiating primary central nervous system lymphomas (PCNSLs) from glioblastomas (GBMs) and inflammatory demyelinating pseudotumors (IDPs). MATERIALS AND METHODS: Magnetic resonance images were reviewed retrospectively in 82 patients including 39 PCNSLs, 35 GBMs, and 8 IDPs. Regions of interest were drawn around the tumor on contrast-enhanced axial images; these images were transferred onto coregistered ADC maps to obtain the ADCmin, and the normalized ADCmin ratios (rADCmin) were calculated using the formula rADCmin = ADCmin of the lesion / ADCmin of the normal white matter. The rADCmin values were compared between PCNSLs, GBMs, and IDPs using the analysis of variance test. Receiver operating characteristic curves were constructed to evaluate the diagnostic performance of rADCmin values and to determine the optimum thresholds. Simple logistic regression was analyzed to evaluate the relationship between ADCs and tumor cellularity. RESULTS: The rADCmin value was significantly lower in PCNSLs (0.675 ± 0.113) than GBMs (0.765 ± 0.059) and IDPs (0.834 ± 0.067) (PCNSL vs GBM, P < 0.001; PCNSL vs IDP, P < 0.001). Relative ADCmin was a significant assessor for differentiating PCNSLs from non-PCNCLs (P < 0.001). The optimal cutoff value was 0.722 (sensitivity, 74.5%; specificity, 74.1%; area under the curve, 0.803) on receiver operating characteristic analysis. A stronger negative correlation (r = -0.755, P = 0.000) was obtained between the cytoplasm and rADCmin. CONCLUSIONS: Relative ADCmin value is helpful in differentiating PCNSL from GBM and IDP. Thus, ADC values may provide a useful supplement to the information obtained from conventional contrast-enhanced magnetic resonance imaging and assist in future treatment planning.
OBJECTIVE: Our purpose was to evaluate the diagnostic performance of diffusion-weighted imaging, the relative minimum apparent diffusion coefficient (rADCmin) in differentiating primary central nervous system lymphomas (PCNSLs) from glioblastomas (GBMs) and inflammatory demyelinating pseudotumors (IDPs). MATERIALS AND METHODS: Magnetic resonance images were reviewed retrospectively in 82 patients including 39 PCNSLs, 35 GBMs, and 8 IDPs. Regions of interest were drawn around the tumor on contrast-enhanced axial images; these images were transferred onto coregistered ADC maps to obtain the ADCmin, and the normalized ADCmin ratios (rADCmin) were calculated using the formula rADCmin = ADCmin of the lesion / ADCmin of the normal white matter. The rADCmin values were compared between PCNSLs, GBMs, and IDPs using the analysis of variance test. Receiver operating characteristic curves were constructed to evaluate the diagnostic performance of rADCmin values and to determine the optimum thresholds. Simple logistic regression was analyzed to evaluate the relationship between ADCs and tumor cellularity. RESULTS: The rADCmin value was significantly lower in PCNSLs (0.675 ± 0.113) than GBMs (0.765 ± 0.059) and IDPs (0.834 ± 0.067) (PCNSL vs GBM, P < 0.001; PCNSL vs IDP, P < 0.001). Relative ADCmin was a significant assessor for differentiating PCNSLs from non-PCNCLs (P < 0.001). The optimal cutoff value was 0.722 (sensitivity, 74.5%; specificity, 74.1%; area under the curve, 0.803) on receiver operating characteristic analysis. A stronger negative correlation (r = -0.755, P = 0.000) was obtained between the cytoplasm and rADCmin. CONCLUSIONS: Relative ADCmin value is helpful in differentiating PCNSL from GBM and IDP. Thus, ADC values may provide a useful supplement to the information obtained from conventional contrast-enhanced magnetic resonance imaging and assist in future treatment planning.