In today's era of 'big data', it is increasingly becoming important to find a reliable way of combining the vast amount of rich data that is acquired across a variety of MRI scanners. For one, this will greatly improve the statistical power of clinical trials, thereby increasing the likelihood of detecting subtle changes at the earliest stages of brain disorders, such as dementia. To achieve such statistical power, it is often necessary to conduct multi-center trials which consist of a large number of subjects, especially those suffering from rare diseases, from as many centers as possible. However, aggregating data that is obtained from different centers and scanners poses a challenge due to the inherent variabilities that exist in the acquired data. One way of overcoming this challenge is by performing data harmonisation -- a technique which is used to make datasets obtained from different scanners as similar as possible.

Here we attempt to develop Machine learning based approaches for performing data haromonisation.