Statistical shape analysis for the classification of renal tumors affecting children

In this research project, we describe an application of statistical shape analysis. In
order to differentiate the various kidney tumours appearing in childhood we use shape
analysis on two-dimensional magnetic resonance images (MRI). We show that this
mathematical procedure can be an interesting tool to assist the radiologist who is required
to make a decision based on their intuition and their experience in lack of specific tumour
characteristics. This study is the first one using MR images in oncology for statistical
shape analysis. Our method is innovative in the way to find suitable landmarks and to test
the differences, even if the sample size is small. In order to test the mean shape, the
statistical test of Ziezold is used.