Gary Cook MD, FRCR, Presenter: Research support , General Electric Company Consultant, Alliance Medical Limited

Hyon-Mok Sohn, Abstract Co-Author: Nothing to Disclose

Muhammad Siddique, Abstract Co-Author: Nothing to Disclose

Vicky Joo-Lin Goh MBBCh, Abstract Co-Author: Research Grant, Siemens AG

Janine Zylstra, Abstract Co-Author: Nothing to Disclose

Lucinda Tullie, Abstract Co-Author: Nothing to Disclose

Michael Green, Abstract Co-Author: Nothing to Disclose

James Gossage, Abstract Co-Author: Nothing to Disclose

Robert Mason, Abstract Co-Author: Nothing to Disclose

Andrew Davies, Abstract Co-Author: Nothing to Disclose

To investigate whether texture features describing heterogeneity of 18F-FDG PET uptake in oesophageal cancer are associated with NAC response or survival.

124 patients with oesophageal adenocarcinoma (111 adenocarcinoma, 23 with squamous cell carcinoma), mean age 63.2y, 96M were include. 114 received at least three cycles of NAC. 10 second and high order texture features, standardised uptake values (SUVmax, mean and peak), total lesion glycolysis (TLG) and metabolic tumour volume (MTV) were tested. Patients were divided into responders (PR/CR) and non-responders (SD/PD) by CT RECIST and into good response (grade 1,2) and poor response (grades 3-5) by Mandard histological tumour regression grades. The ability of the parameters to predict responders was tested (Mann-Whitney U test) and correlations with MTV were determined (Spearman rank correlation). Cox regression was performed for survival analysis. A minimum volume threshold (14cm3) was calculated and implemented to exclude bias from small tumours.

There was a significant difference in zone percentage between CT RECIST responders and non-responders (p=0.012) independent of tumour size after the exclusion of small tumours. High order texture features including, coarseness (p=0.053), contrast (p=0.012) and complexity (p=0.021), were able to predict NAC response measured by Mandard grades. However, these features exhibited strong associations with MTV, (ρ=0.892,-0.895, -0.951, respectively, all p<0.001). No texture feature was significantly associated with overall survival.

Small volume tumours are subject to bias in texture feature calculation. However, allowing for this, texture features characterising heterogeneity of 18F-FDG uptake in oesophageal cancer may offer an alternative method of predicting NAC response but we have found no association with any texture feature and survival.

There is no reliable imaging predictor for neoadjuvant chemotherapy (NAC) response or prognostic marker for survival in oesophageal carcinoma. Predictive and prognostic imaging biomarkers would be of value in this aggressive disease.

Cook, G, Sohn, H, Siddique, M, Goh, V, Zylstra, J, Tullie, L, Green, M, Gossage, J, Mason, R, Davies, A, Heterogeneity of 18F-FDG PET Images of Oesophageal Cancer as a Predictive and Prognostic Biomarker.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14045556.html