Correlation of Pre –Treatment Volumetric Apparent Diffusion Coefficient Histogram Analysis of Diffusion Weighted MRI in Rectal Cancer with Post ChemoRadiation Treatment Response and Clinical Outcomes.
Kartik Jhaveri1, Vijay Chidambaram2, James Brierley1, Bernard Cummings1,3, Rajesh Bhayana1, Ravi Menezes1, Erin Kennedy4, and Richard Kirsch4

1UHN,university of Toronto, Toronto, ON, Canada, 2Royal Liverpool Hospital, Liverpool, United Kingdom, 3Toronto, Canada, 4Mt.Sinai Hospital,University of Toronto, Toronto, Canada


Rectal cancer patients treated with preoperative chemoradiation therapy with complete treatment response can be considered for individualised therapies such as ‘wait and watch’ avoiding surgery. However this currently can only be definitely identified by postoperative histopathology. Diffusion weighted-MRI has been well correlated to tumor biology and treatment response. Volumetric ADC histogram analysis eliminates errors resulting from tumor heterogeneity and variable diffusivities by including the tumor volume and could provide promise for prediction of treatment response and clinical outcomes,thereby providing means for individualised therapy.


Preoperative ChemoRadiation Therapy (CRT) of rectal cancer has been shown to reduce pelvic tumor recurrence but can have significant adverse side effects. Patients with complete response can be considered for individualised therapies such as ‘wait and watch’ (1). However, it is not possible to identify complete responders following CRT besides post-operative histopathological evaluation. DW-MRI has been well correlated to tumor biology and treatment response. Mean ADC calculated by operator dependent region of interest (ROI) can result in inaccuracies from inter-observer variability and tumor heterogeneity (2, 3). Volumetric ADC histogram analysis eliminates these limitations via interrogation of tumor heterogeneity and diffusivities in the entire tumor volume.


To investigate the relationship between pre-treatment volumetric ADC histogram parameters in rectal cancer with postoperative histopathology and clinical outcomes following pre-operative CRT.


78 patients with rectal cancer with pre-operative CRT and rectal MRI with diffusion weighted imaging (DWI) sequences were included in this retrospective study. Standard rectal MRI was performed on 1.5T or 3T MRI(Siemens Avanto or Verio, Erlangen,Germany) including high resolution T2 weighted sequence, DWI with three diffusion sensitising gradients and ADC maps in addition to routine pelvic MR sequences. DWI analysis was performed on a stand-alone workstation with the use of software tool “MR OncoTREAT” (Siemens AG, Germany) for registration, segmentation and analysis of the MR data. The volumetric ADC data was displayed as a histogram with data export function to a spreadsheet. The software distributed the total range of ADC values into multiple color coded parts (Fig. 1). Color maps of ADC are overlaid on the T2 weighted MR images to visualise the distribution of ADC values. Multiple variables were calculated including tumor volume, mean, standard deviation, histogram kurtosis and skewness, 25th, 50th and 75th percentiles were calculated. ADC % distribution for ADC values of 0.8, 0.8-1.0 and > 1.0 x 10-3 mm2/s were determined. Correlation was made to post-operative pathological complete response, perineural invasion, clinical or radiological evidence of disease progression using the Mann-Whitney test. All tests were two-sided, and a p-value <0.05 was significant.


The mean rectal tumor volume was 24cc (range: 1cc -134cc). 8 patients had metastatic disease at time of initial diagnosis. Post CRT, 13 of 78 patients showed distant disease progression. 8 patients showed pathologic complete response. Pre-treatment MRI mean ADC was 1.2 x 10-3 mm2/s (range: 0.3 to 1.99 x 10-3 mm2/s). Mean kurtosis measured was 0.56 (range: -1 to 6; standard deviation: 1.36). Mean skewness was 0.3 (range: -1 to 2; standard deviation: 0.69). Skewness had significant correlation (p-value = 0.006) with disease progression (Fig.2,3). Pre-treatment MRI tumor volume showed significant correlation (p value = 0.013) with pathologic complete response. Mean ADC and percentage voxels for various ADC ranges did not reveal any significant correlation with treatment response or progression. There was no significant correlation between the ADC histogram parameters or tumor volume with perineural invasion.


Good correlation was seen between the skewness of the ADC histogram with clinical disease progression (local recurrence and metastasis).The patients with disease progression had a skewness value of -1 or 0. A negative skew indicated that most voxels contain an ADC value greater than the mean, with relatively few voxels containing an ADC value less than the mean. Significant correlation was seen between pre-treatment MRI volume and pathologic complete response following CRT. There was no significant correlation between kurtosis, mean ADC, 25th, 50th and 75th ADC percentiles these variables and the binary outcome variables (perineural invasion, disease progression or pathological complete response).


Volumetric ADC histogram analysis of pre CRT MRI provides some promising quantitative measures for prediction of post-CRT pathological complete response and disease progression.


Robert Grimm, Siemens Healthcare GmbH, Application Predevelopment, Erlangen, Germany


1. K. N. Fischkoff et al .“Nonoperative approach to locally advanced rectal cancer after neoadjuvant combined modality therapy: challenges and opportunities from a surgical perspective.,” Clin. Colorectal Cancer, vol. 10, no. 4, pp. 291–7, Dec. 2011.

2. Y.-S. Sun et al.“Locally advanced rectal carcinoma treated with preoperative chemotherapy and radiation therapy: preliminary analysis of diffusion-weighted MR imaging for early detection of tumor histopathologic downstaging.,” Radiology, vol. 254, no. 1, pp. 170–8, Jan. 2010.

3. D. M. J. Lambregts et al .“Tumour ADC measurements in rectal cancer: effect of ROI methods on ADC values and interobserver variability.,” Eur. Radiol., vol. 21, no. 12, pp. 2567–74, Dec. 2011.cancer.,” Eur. Radiol., vol. 21, no. 5, pp. 987–95, May 2011.


The ADC histogram bar graph(right) shows frequency of voxels at respective ADC. The blue bars represent those with ADC less than 1.0 x 10-3 mm2/sec.AxialT2 MRI section(left upper) and Axial ADC map(left lower)with superimposed ADC distributions per voxel

51-year-old woman with rectal cancer with Axial T2 image(left upper inset) and ADC histogram bar graph.Tumor volume 11cc, mean ADC 1.3, kurtosis 3 and skewness -1.

Box and whisker plot of ADC histogram skewness for patients with complete response, partial response and disease progression.

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