Balaji Ganeshan PhD, Presenter: Scientific Director, TexRAD Limited

Asim Afaq FRCR, Abstract Co-Author: Nothing to Disclose

Shonit Punwani MBBS, Abstract Co-Author: Nothing to Disclose

Alec Engledow, Abstract Co-Author: Nothing to Disclose

Daren Francis, Abstract Co-Author: Nothing to Disclose

Nicholas Rhys-Jones, Abstract Co-Author: Nothing to Disclose

Tan Arulampalam, Abstract Co-Author: Nothing to Disclose

Sanjay Dindyal, Abstract Co-Author: Nothing to Disclose

Omer Jalil, Abstract Co-Author: Nothing to Disclose

Anna Barnes, Abstract Co-Author: Nothing to Disclose

Celia O'Meara, Abstract Co-Author: Nothing to Disclose

Manuel Rodriguez-Justo, Abstract Co-Author: Nothing to Disclose

Peter Josef Ell MD, Abstract Co-Author: Consultant, Ion Beam Applications, SA

Kenneth Miles, Abstract Co-Author: Director, TexRAD Limited Shareholder, TexRAD Limited

Ashley McAllister Groves MBBS, Abstract Co-Author: Investigator, GlaxoSmithKline plc Investigator, General Electric Company Investigator, Siemens AG Advisory Board, Merck & Co, Inc

To describe the technique and initial results obtained from texture analysis of MR Dixon images derived from PET-MRI in primary colorectal cancer.

10 consecutive prospectively recruited primary colorectal cancer patients (6 male and 4 female; Mean age 61.3±10.02) underwent PET-MRI including acquisition of Dixon images for attenuation correction, measurement of tumor fluorodeoxyglucose uptake (SUVmean) and MRI apparent diffusion coefficient (ADCmean). A parametric image of fractional water content was produced from the Dixon images from the ratio of the water-weighted image and the summed water- and fat-weighted images. Fractional water images underwent texture analysis using a filtration-histogram method. Filtration highlighted image features ranging between approximately 2mm and 7mm diameter. Histograms of filtered images quantified as standard-deviation (SD) and proportion of positive pixels (PPP) were correlated against SUVmean and ADCmean using Spearman Rank Correlation.

The mean tumor fractional water content was 0.88 (range: 0.74 – 0.95). Fractional water content did not correlate significantly with ADCmean (rs = 0.358, p=0.310) and SUVmean (rs = -0.030, p=0.934). Fractional water texture expressed as SD correlated negatively with ADCmean (rs = -0.75, p=0.013) with PPP values correlated positively with SUVmean ( rs = 0.75, p=0.013). 

Texture analysis of Dixon images can potentially assess tumor water distribution. Tumor ADCmean and SUVmean measurements may be related to tumor water distribution in colorectal cancer.

Texture analysis of Dixon images in colorectal cancer can potentially provide information about tumor biology with possible applications in personalized medicine.

Ganeshan, B, Afaq, A, Punwani, S, Engledow, A, Francis, D, Rhys-Jones, N, Arulampalam, T, Dindyal, S, Jalil, O, Barnes, A, O'Meara, C, Rodriguez-Justo, M, Ell, P, Miles, K, Groves, A, Texture Analysis of MR Dixon Images in Primary Colorectal Cancer: Initial Experience Using PET-MRI.  Radiological Society of North America 2013 Scientific Assembly and Annual Meeting, December 1 - December 6, 2013 ,Chicago IL. http://archive.rsna.org/2013/13014214.html