Pre-treatment magnetic resonance-based texture features as potential
imaging biomarkers for predicting event free survival in anal cancer
treated by chemoradiotherapy
Baudouin Denis De Senneville
Received: 26 July 2017 /Revised: 8 December 2017 / Accepted: 22 December 2017 / Published online: 5 February 2018
European Society of Radiology 2018
Aim To assess regular MRI findings and tumour texture features on pre-CRT imaging as potential predictive factors of event-free
survival (disease progression or death) after chemoradiotherapy (CRT) for anal squamous cell carcinoma (ASCC) without
Materials and methods We retrospectively included 28 patients treated by CRT for pathologically proven ASCC with a pre-CRT
MRI. Texture analysis was carried out with axial T2W images by delineating a 3D region of interest around the entire tumour
volume. First-order analysis by quantification of the histogram was carried out. Second-order statistical texture features were
derived from the calculation of the grey-level co-occurrence matrix using a distance of 1 (d1), 2 (d2) and 5 (d5) pixels. Prognostic
factors were assessed by Cox regression and performance of the model by the Harrell C-index.
Results Eight tumour progressions led to six tumour-specific deaths. After adjusting for age, gender and tumour grade, skewness
(HR = 0.131, 95% CI = 0-0.447, p = 0.005) and cluster shade_d1 (HR = 0.601, 95% CI = 0-0.861, p = 0.027) were associated
with event occurrence. The corresponding Harrell C-indices were 0.846, 95% CI = 0.697-0.993, and 0.851, 95% CI = 0.708-
Conclusion ASCC MR texture analysis provides prognostic factors of event occurrence and requires additional studies to assess
its potential in an Bindividual dose^ strategy for ASCC chemoradiation therapy.
• MR texture features help to identify tumours with high progression risk.
• Texture feature maps help to identify intra-tumoral heterogeneity.
• Texture features are a better prognostic factor than regular MR findings.
Keywords Anal squamous cell carcinoma
Magnetic resonance imaging
Electronic supplementary material The online version of this article
(https://doi.org/10.1007/s00330-017-5284-z) contains supplementary
material, which is available to authorized users.
* Arnaud Hocquelet
Department of Radiodiagnostic and Interventional Radiology, Centre
Hospitalier Universitaire Vaudois (CHUV) and University of
Lausanne, Lausanne, Switzerland
Department of Diagnostic and Interventional Radiology, Hopital
Haut Lévêque, Centre Hospitalier Universitaire de Bordeaux,
33600 Pessac, France
EA IMOTION (Imagerie Moléculaire et Thérapies Innovantes en
Oncologie) Université de Bordeaux, 146 rue Leo Saignat, Case 127,
33076 Bordeaux, France
Institut de Mathématiques de Bordeaux (IMB), UMR 5251
CNRS/Univ, Bordeaux, 351 cours de la Libération,
33405 Talence, France
Department of Nuclear Medicine, CHU de Bordeaux,
33000 Bordeaux, France
Departement of Radiotherapy, Hopital Haut Lévêque, CHU de
Bordeaux, 33600 Pessac, France
European Radiology (2018) 28:2801–2811