Our paper "IRMA: Machine learning-based harmonization of 18F-FDG PET brain scans in multi-center studies" 
by Sofie Lövdal et al. has been published in the European Journal of Nuclear Medicine and Molecular Imaging 
and is available online (open access):

In this work we apply interpretable machine learning systems for the analysis of 18-FDG PET
brain scans from different medical centers. We show that the center origin of healthy control brain  
images acquired with different scanners/protocols can be identified with high confidence. Consequently,
machine learning models  trained with data from different scanners may be heavily impacted by this 
bias when applied to a clinically  relevant problem, e.g. the differential diagnosis of neurodegenerative
disorders. We propose IRMA (Iterated Relevance Matrix Analysis)  as a recursive method to learn and 
disregard bias in PET feature vectors. Related code is freely available at  


---------------------------------------------------
Prof. Dr. Michael Biehl
Bernoulli Institute for Mathematics, 
Computer Science & Artificial Intelligence
P.O. Box 407, 9700 AK Groningen, NL