A Modified Tri-exponential Model For Nulti-b-value Diffusion-weighted Imaging And Its Application In Grading And Differential Its Application In Grading And Differential Diagnosis Of Gliomas

Part Ⅰ. A modified tri-exponential model for multi-b-value DWI and its comparison with bi-exponential model and conventional tri-exponential modelObjectivethis study was to present a modified tri-exponential model to detect the strictly diffusion-limited compartment, and compare it with the bi-exponential model and the conventional tri-exponential model.Materials and methodsmulti-b-value DWI images with 17 b-values ranging from 0 to 8000 s/mm2 were achieved from 6 healthy volunteers. The regions of interest (ROIs) were drawn in multiple areas of white matter and gray matter. The DWI images with the first 16 b-values were used for curve fitting. The residual sums of squares (RSSs), the corrected Akaike information criterions (AICcs) were used for evaluating the goodness of fitting. The squared prediction errors (SPEs) at b= 8000 s/mm2 were used for assessing the predictability of the models.ResultsIn all ROIs of white matter:the RSSs of the modified tri-exponential model were significantly smaller than those of the other two models (p< 0.05); the AICcs of the conventional tri-exponential model were significantly larger than those of the other two models (p< 0.05); the SPEs of the bi-exponential model were significantly larger than those of the other two models (p< 0.05). The median values of fo were 0.8% in gray matter and ranging 11.9-19.6% in white matter. The median value of ADCvery-slow (487×10-6 mm2/s) was not extremely small in gray matter.ConclusionThe new model is better than the other two models in fitting the DWI images. The new model can detect the strictly diffusion-limited compartment in tissues, while the conventional tri-exponential model cannot.Part Ⅱ. A modified tri-exponential model for multi-b-value diffusion-weighted imaging in grading and differential diagnosis of gliomasObjectiveThis study was to explore the clinical value of a modified tri-exponential model in grading gliomas and distinguishing gliomas from primary central nervous system lymphomas (PCNSL).Materials and methodsThere were 18 low-grade gliomas (LGG),45 high-grade gliomas (HGG) and 5 PCNSL enrolled in this study, and diffusion-weighted imaging (DWI) maps were achieved with 9 b-values. Maps of parameters (fo, fsiow, ADCsiow, ffast, ADCfast) were obtained by the modified tri-exponential model. All values of parameters were compared among LGG, HGG and PCNSL. Receiver operating characteristic and Pearson rank correlation were used for statistical analysis.ResultsThe fo values were significantly different among PCNSL (13.98%), HGG (6.98%) and LGG (3.14%) (p< 0.05). Significant differences were also found in values of fslow, ffast and ADCsiow between HGG and LGG (p< 0.05). The area under curve, sensitivity and specificity of fo were 0.901,82.2% and 83.3% respectively in distinguishing HGG from LGG, and were 0.981,100% and 96.8% respectively in distinguishing PCNSL from gliomas. Significant correlations were found between all parameters and Ki-67 index (p< 0.05).ConclusionThe fo is a significant compartment in high malignant tumors. The modified tri-exponential model has a potential in grading and differential diagnosis of gliomas.