Detectability of brain structure abnormalities related to autism through MRI-derived measures from multiple scanners

Multi-site studies utilizing MRI-derived measures from multiple scanners present an opportunity to increase the power of statistical models by pooling data. This represents the potential of detecting finer and more subtle effects related to pathology and is thus very appealing. It remains however unclear whether or not the potential confound introduced by scanner-related variations will devalue the integrity of any result. Specifically, it is crucial to test whether or not the error introduced into brain morphometry descriptors by image properties variations across scanners can exceed the effect of the disease itself. Moreover, evaluating the influence of the scanner on age-related effects in measurements is essential in the context of developmental disorders such as Autism Spectrum Disorders. In this study, we investigated a data-set of anatomical MRI pooled form three different scanners, representing a total of 186 subjects, 97 patients suffering from ASD and 89 controls. We quantitatively assessed the effects of age, pathology and scanner factors on cortical thickness measurements.