IVIM-DKI with parametric reconstruction method for lymph node evaluation and characterization in lymphoma: A preliminary study comparison with FDG-PET/CT

Abstract

To investigate IVIM-DKI in diagnosing benign and malignant lymph nodes in lymphoma and its subtypes in comparison to FDG-PET/CT. A total of twenty-one (n = 21) patients diagnosed with biopsy-proven Hodgkin lymphoma(HL: n = 13) and non-Hodgkin lymphoma (NHL: n = 8) were prospectively evaluated. All patients underwent MRI(T1-weighted, Short-Tau-Inversion-Recovery (STIR)), and IVIM-DKI was acquired with 9b-values (0–2000s/mm2) at 1.5T and whole-body FDG-PET/CT. The maximum and average standard uptake values (SUVmax and SUVmean) were calculated using PET images. IVIM-DKI parameters (diffusion coefficient(D), pseudo-diffusion coefficient (D*), perfusion fraction (f), and kurtosis (k)) were estimated using IVIM-DKI model with total variation (TV) method (IDTV model). Area-under-curve from receiver-operating-curve (ROC) analysis was used to examine diagnostic value of IVIM-DKI parameters in differentiation between benign and malignant lymph nodes and HL from NHL. Machine learning-based classification of histogram features were performed using linear classifier model. For malignant vs. benign lymph nodes, apparent diffusion coefficient (ADC), f, and k were significantly (p < 0.05) lower in malignant vs. benign lymph nodes. f (AUC:0.88) and k (AUC:0.83) showed high AUC and histogram features combination of f (variance + skewness + kurtosis) showed highest accuracy of 97.2% and AUC of 1. ADC, D, D*, and f were significantly (p < 0.05) lower in NHL than HL. D* showed highest AUC of 0.85 than D (AUC:0.84), ADC (AUC:0.84), and f (AUC:0.74) in NHL vs. HL. SUVmax (spearman-rho = 0.85), ADC (spearman-rho = 0.50), D (spearman-rho = 0.48), and D* (spearman-rho = 0.49) showed significant (p < 0.05) positive correlation with SUVmean. Multi-b-values IVIM-DKI with histogram analysis helps characterize benign and malignant lymph nodes in lymphomas. IVIM-DKI parameters can differentiate malignant lymph nodes in HL and NHL superior to PET parameters.