Improving lymph node characterization in staging malignant lymphoma using first‐order ADC texture analysis from whole‐body diffusion‐weighted MRI. Issue 4 (14th April 2018)

• Record Type: Journal Article
• Authors: De Paepe, Katja N.; De Keyzer, Frederik; Wolter, Pascal; Bechter, Oliver; Dierickx, Daan; Janssens, Ann; Verhoef, Gregor; Oyen, Raymond; Vandecaveye, Vincent
• Is Part Of: Journal of magnetic resonance imaging. Volume 48:Issue 4(2018)
• Page Start: 897
• Page End: 906
• Publication Date: 2018-04-14
• Subjects: apparent diffusion coefficient -- diffusion‐weighted -- MR imaging -- lymphoma -- staging; Magnetic resonance imaging -- Periodicals; 616
• DOI: 10.1002/jmri.26034
• Languages: English
• ISSNs: 1053-1807

Abstract

Abstract : Background: Correct staging and treatment initiation in malignant lymphoma depends on accurate lymph node characterization. However, nodal assessment based on conventional and diffusion‐weighted (DWI) MRI remains challenging, particularly in smaller nodes. Purpose: To evaluate first‐order apparent diffusion coefficient (ADC) texture parameters compared to mean ADC for lymph node characterization in non‐Hodgkin lymphoma (NHL) using whole‐body DWI (WB‐DWI). Study Type: Retrospective. Population: Twenty‐eight patients with NHL. Field Strength/Sequence: 3T whole‐body DWI using two b‐values (0–1000 s/mm 2 ). Assessment: Regions of interest were drawn on the three most hyperintense lymph nodes on b1000‐images, irrespective of size, in all nodal body regions. Diagnostic performance of mean ADC (ADCmean ) was compared with first‐order ADC texture parameters: standard deviation (ADCstdev ), kurtosis (ADCkurt ), and skewness (ADCskew ). Additional subanalyses focused on the accuracy of ADCmean and ADC texture parameters in different lymph node volumes and nodal regions. Statistical Tests: Benign and malignant nodes were compared using Mann–Whitney U ‐tests with 18‐Fluoro‐deoxyglucose positron emission tomography computed tomography and bone marrow biopsy as reference standard. Receiver operating characteristic analyses were performed to determine cutoff values and calculate sensitivity, specificity, accuracy, and positive and negative predictive value (PPV, NPV). Results: ADCmean ( P = 0.008), ADCskew and ADCkurt differed significantly between benign and malignant nodes ( P < 0.001), while ADCstdev didn't ( P = 0.21). ADCskew was the best discriminating parameter, with 79% sensitivity, 86% specificity, 83% accuracy, 85% PPV, and 81% NPV. In every volume category, ADCskew yielded the highest accuracy (88% in 0–25 th percentile volume, 75% in 25 th –75 th percentile, 93% in 75–100 th percentile). On a per‐region basis, ADCskew accuracy varied 13.6% between nodal regions, while ADCmean, ADCkurt, and ADCstdev showed interregional variation of 17.4%, 20.3%, and 14.9%, respectively. Data Conclusion: First‐order ADC texture analysis with WB‐DWI improved lymph node characterization compared to ADCmean . ADCskew was the most accurate and robust discriminatory parameter over all lymph node volumes and nodal body regions. Level of Evidence: 3 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2018;48:897–906.