Meeting Banner
Abstract #1850

3D Texture Analysis of Heterogeneous MRI Data for the Characterisation of Childhood Brain Tumours

Ahmed E Fetit 1,2 , Jan Novak 2,3 , Daniel Rodriguez 4 , Dorothee P Auer 4,5 , Chris A Clark 6 , Richard G Grundy 4,5 , Tim Jaspan 5 , Andrew C Peet 2,3 , and Theodoros N Arvanitis 1,2

1 Institute of Digital Healthcare, WMG, University of Warwick, Coventry, United Kingdom, 2 Birmingham Children's Hospital, Birmingham, West Midlands, United Kingdom, 3 University of Birmingham, Birmingham, West Midlands, United Kingdom, 4 University of Nottingham, Nottingham, United Kingdom, 5 University Hospital Nottingham, Nottingham, United Kingdom, 6 University College London, London, United Kingdom

There is an increasing interest in developing quantitative MR image analysis tools that can capture information below human visual perception and hence assist the diagnosis of childhood brain tumours. In this work, we compare the performance of 3D and 2D texture analysis on multi-modal, heterogeneous MR data sets of children diagnosed with Medulloblastoma, Pilocytic Astrocytoma and Ependymoma. Additionally, we address the problem of class imbalance by creating synthetic tumour samples using Synthetic Minority Over-Sampling Technique (SMOTE) through operating in feature space. Our results support the use of texture analysis as an automated, quantitative technique to assist with the diagnosis of paediatric brain tumours.

This abstract and the presentation materials are available to members only; a login is required.

Join Here