Niloufar Zarinabad1,2, Christopher Bennett1,2, Simrandip Gill1,2, Martin P Wilson1, Nigel P Davies1,2,3, and Andrew Peet1,2
1Institute of Cancer and Geonomic Sciences, University of Birmingham, Birmingham, United Kingdom, 2Birmingham Children’s Hospital NHS foundation trust, Birmingham, United Kingdom, 3Department of Medical Physics,University Hospitals Birmingham NHS Foundation Trust, Birmingham, United Kingdom
Classification of paediatric brain tumours from Magnetic-Resonance-Spectroscopy has
many desirable characteristics. However the imbalanced nature of the data
introduces difficulties in uncovering regularities within the small rare tumour
type group and attempts to train learning algorithms without correcting the
skewed distribution may be premature. By fusing oversampling and
classification techniques together, an improved classification performance across
different classes with a good discrimination for minority class can be
achieved. The choice of learning algorithm, use of oversampling-technique and
classifier input (complete spectra versus metabolite-concentration) depends on
the data distribution, required accuracy in discriminating specific groups
and degree of post-processing complexity.
