MRI in Assessing Response to Neoadjuvant Chemo-radiation in Locally Advanced Rectal Cancer Using DCE-MR and DWI Data Sets: Before, During and After the Treatment
Ke Nie1, Liming Shi2, Ning Yue1, Jabbour Salma1, Xi Hu2, Liwen Qian2, Tingyu Mao2, Qin Chen2, Xiaonan Sun2, and Tianye Niu2,3,4

1Radiation Oncology, Rutgers-Cancer Institute of New Jersey, Rutgers-Robert Wood Johnson Medical School, New Brunswick, NJ, United States, 2Radiation Oncology, Sir Run Run Shaw Hospital, Zhejiang University of Medicine, Hangzhou, China, People's Republic of, 3Institute of Translational Medicine, Hangzhou, China, People's Republic of, 4Radiology, Sir Run Run Shaw Hospital, Zhejiang University of Medicine, Hangzhou, China, People's Republic of


We are one of the first to investigate the predictive value of combined anatomical, DCE-MRI and DWI for good pathological response at different time points during the pre-operative chemo-radiation treatment (CRT) in patients with locally advanced rectal cancer (LARC). The pre-treatment ADC and internal heterogeneity enhancement measured by texture features from DCE-MRI and the relative change of the ADC values during the treatment showed good prognostic value with pathological response. Overall, this study provides new information of the optimal use of MRI in predicting response to the pre-operative CRT, which may further help tailor the treatment into the era of personalized medicine.


The recent trend toward patient-tailored treatment for locally advanced rectal cancer (LARC) has highlighted the need for reliable methods in early prediction of treatment outcome. The inclusion of imaging tools that mirror the behavior of rectal cancer may potentially improve such predictive accuracy. In this study, we investigate the prognostic value of DCE-MRI and DWI at different time points during the course of preoperative chemo-radiotherapy (CRT) for good pathological response prediction.


Twenty-one patients (60.1±6.2 years, between 01/2012 – 9/2013) with LARC were enrolled. All received 5-week concurrent CRT followed by surgery 8-10 weeks afterward. MRI scans were acquired at three time points: prior to the treatment (pre-CRT), early follow-up (F/U) at 4 weeks after the start of CRT, and 4 weeks after the completion of CRT (post-CRT). Image sequences included anatomical T1, T2, DCE-MRI and DWI (b=0, and 800 s/mm2) all in transverse planes. A manual segmentation of the tumor volume was performed on third-phase post-contrast DCE images by an experienced radiologist. The contours were further mapped back to other image sequences. The tumor volume (TV) from anatomical MRI, shape feature as wash-in (WI)/ wash-out (WO) slope and maximum enhancement (MEn) from DCE-MRI time-intensity-curve (TIC), the mean-ADC value from DWI were obtained for each MR scan. The internal image signal variations in both DCE-MRI and DWI were further evaluated using Gray Level Co-occurrence Matrix (GLCM) texture features. The absolute values of these parameters at different time points, and relative changes to the pre-treatment values were recorded. For analyses, the pathological good responders (GRs, defined as Mandard tumor regression grade, TRG 1-2) were grouped and compared with the pathological moderate/poor responders (non-GRs, defined as TRG 3-5). Due to limited number of pathological complete response (pCR), no group comparison was performed between pCR vs. non-pCR. Individual variables were analyzed for significant differences using the Mann-Whitney test with a 2-sided p-value <0.05 as statistical significant.


Of all the patients, 9(43%) were classified as GRs and 3 had pathological complete response (pCR). The GRs and non-GRs groups were homogeneous in types of age, sex, distance from anal verge, pre-treatment CEA levels and TNM stage. The tumor volume (TV) or %TV changes during the treatment did not significantly differ between two groups. On the contrary, the pre-treatment ADC values were lower in the GRs compared to non-GRs with mean of 0.83 × 10-3 mm2/s vs. 0.97×10-3 mm2/s, p=0.03. The relative change, ∆ADC, was also predictive for GRs. The increase in ADC value from pre-CRT until 4 weeks F/U for GRs (54.7±41.2%) was significantly higher compared to the corresponding value in non-GRs (21.9±24.2%) with p=0.02. The increase in ADC value from pre-CRT until 4 weeks post-CRT for GRs was 77.2±54.6% versus 36.0±29.4% for non-GR with p=0.01. The pre-CRT time-intensity-curve (TIC) shape features as WI/WO slope or maximum enhancement (MEn) from DCE-MRI did not differ between two response groups. While the pre-CRT heterogeneity distribution measured by GLCM textures showed GRs were more likely to have homogeneous enhancement compared to non-GRs with p<0.01. However, the relative changes in neither TIC shape features nor the internal textures at all time points (during or post-CRT) compared to pre-CRT values showed good correlation with the response. Figure 1 depicts an example of MR images of two patients. These two patients were both at 60 years old (y/o) with mid-rectum cancer at stage of cT3N+M0. Pre-treatment CEA levels were comparable at 9.36 and 10.57 respectively. The pre-CRT tumor volume (TV), the associated changes or the enhancement maps from DCE-MRI did not reveal differences. The pre-CRT ADC value increased from 0.70 x10-3mm2/s (pre-CRT) to 1.14x10-3 mm2/s (F/U) and further to 1.24x10-3 mm2/s post-CRT, with a percentage of ADC increase of 77% for GR, while the non-GR case the ADC value increased from 1.24x10-3 pre-CRT to 1.51x10-3 post-CRT, with a relatively less percentage change of 21%.


This study showed that selection of good pathological responders during pre-operative CRT for locally advanced rectal cancer is feasible with diffusion and DCE-MRI. This is of important clinical value that may enable to individualize and adapt the therapy for each patient during treatment. Patients may be selected for organ-sparing local excision or intensified treatment with higher accuracy such as a radiation boost. This could be applied in patients with a moderate response to increase the number of good treatment responders, leading to a higher percentage of patients eligible for organ-sparing treatments






Examples of two patients MR images, good responder (GR) vs. non-good responders (non-GR) at different image sequences as (a) T2 weighted MRI, (b) third phase post-contrast MRI, and (c) DWI (b=800 s/mm2), from pre-CRT (first row), early follow-up (F/U), and 4 weeks after the completion of CRT (post-CRT).

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)