Yi Sui MS, Presenter: Nothing to Disclose

Lei Tang MD, Abstract Co-Author: Nothing to Disclose

Frederick C. Damen PhD, Abstract Co-Author: Nothing to Disclose

Shun-Yu Gao MD, Abstract Co-Author: Nothing to Disclose

Kejia Cai PhD, Abstract Co-Author: Nothing to Disclose

Ying-Shi Sun MD, PhD, Abstract Co-Author: Nothing to Disclose

Xiaohong Joe Zhou PhD, Abstract Co-Author: Nothing to Disclose

To evaluate the performance of a non-Gaussian diffusion model in early prediction of treatment response in recurrent gastrointestinal stromal tumor (GIST) under sunitinib therapy.

With IRB approval, 10 patients (4 men, 6 women) with confirmed failure of previous imatinib therapy, underwent sunitinib (50 mg/day, PO) single-drug targeted treatment. MRI scans were conducted on a 3T scanner before treatment, one week, three weeks and one month after treatment to monitor the tumor response. Diffusion MR images were acquired using 11 b-values up to 3000 sec/mm2. A set of diffusion parameters (apparent diffusion coefficient ADC, intravoxel heterogeneity index β, and mean free diffusion length µ) were fitted pixel by pixel using a fractional order calculus (FROC) model. The percentage change (%Δ) of ADC, β and µ after one week treatment were averaged over the whole tumor regions. The MRI parameters obtained after the first week of treatment were used to predict later treatment outcomes. All 36 tumors in 10 patients were divided into good response (n = 22) and poor response (n = 14) groups based on the Choi and EORTC-ISG-AGITG criteria. The parametric values were compared for each individual parameter using Mann-Whitney U test with a statistical significance set at p < 0.05. ROC analysis was performed to evaluate the performance of individual FROC parameters as well as the combination of all parameters (binary logistic regression) in predicting the therapeutic responses. 

Significant differences between good and poor response groups were found in the %Δ ADC (28% vs 8%, p = 0.032), %Δ β (20% vs -6%, p = 0.013), and %Δ µ (8% vs 2%, p = 0.004). The AUCs of %Δ µ (0.782) and %Δ β (0.747) were larger than %Δ ADC (0.714). When combining all parameters of the FROC model, the AUC was further increased to 0.893, suggesting that the FROC model improved the performance of prediction. The accuracy of prediction was increased to 83.3% using the FROC model, compared to using ADC alone (61.1%).

Our results demonstrate that the FROC diffusion model with high b-values can provide valuable information for early response prediction of sunitinib targeted therapy of GIST.

The FROC diffusion model may provide useful parameters for the prediction of GIST response to sunitinib therapy at early period.

Sui, Y, Tang, L, Damen, F, Gao, S, Cai, K, Sun, Y, Zhou, X, Early Prediction of Response of Gastrointestinal Stromal Tumor to Sunitinib Therapy Using Non-Gaussian Diffusion MRI.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14011015.html