Diffusion-weighted magnetic resonance imaging of rectal cancer: tumour volume and perfusion fraction predict chemoradiotherapy response and survival
MetadataShow full item record
Original versionActa Oncologica. 2017, 56 (6), 813-818. 10.1080/0284186X.2017.1287951
Background: In locally advanced rectal cancer (LARC), responses to preoperative treatment are highly heterogeneous and more accurate diagnostics are likely to enable more individualised treatment approaches with improved responses. We investigated the potential of diffusion-weighted magnetic resonance imaging (DW MRI), with quantification of the apparent diffusion coefficient (ADC) and perfusion fraction (F), as well as volumetry from T2-weighted (T2W) MRI, for prediction of therapeutic outcome. Material and methods: In 27 LARC patients receiving neoadjuvant chemotherapy (NACT) before chemoradiotherapy (CRT), T2W- and DW MRI were obtained before and after NACT. Tumour volumes were delineated in T2W MRI and ADCs and Fs were estimated from DW MRI using a simplified approach to the intravoxel incoherent motion (IVIM) model. Mean tumour values and histogram analysis of whole-tumour heterogeneity were correlated with histopathologic tumour regression grade (TRG) and 5-year progression-free survival (PFS). Results: At baseline, high tumour F predicted good tumour response (TRG1-2) (AUC = 0.79, p = 0.01), with a sensitivity of 69% and a specificity of 100%. The combination of F and tumour volume (Fpre/Vpre) gave the highest prediction of poor tumour response (AUC = 0.93, p < 0.001) with a sensitivity of 88% and a specificity of 91%, and also predicted PFS (p < 0.01). Baseline tumour ADC was not significantly related to therapeutic outcome, whereas a positive change in ADC from baseline to after NACT, ΔADC, significantly predicted good tumour response (AUC = 0.83, p < 0.01, 83% sensitivity, 73% specificity), but not PFS. Conclusions: The MRI parameter F/V at baseline was a remarkably strong predictor of both histopathologic tumour response and 5-year PFS in patients with LARC.