Dr Rekik gave a talk on “Machine Learning Methods for Neuroscience” at MIUA 2017, Edinburgh
Prof Peter Donnelly, Chair of the Royal Society Working Group on Machine Learning, recently stated that “machine learning will have an increasing impact on our lives and lifestyles over the next five to ten years. There is much work to be done so that we take advantage of machine learning’s potential and ensure that the benefits are shared, especially as this could be a key are of opportunity in the UK in coming years […] We have the opportunity now, as a society, to ensure that machine learning can bring the maximum benefit to the greatest number of people. ”
This will also highly impact the field of medical research and neuroscience. With the dataclysm of medical data, big challenges arise:
- What can we learn from medical data (e.g., imaging data, genomic data, clinical scores, etc)?
- How to ‘intelligently’ process large medical datasets to extract meaningful information for disease diagnosis and prognosis?
- How can machine learning methods advance our understanding of brain disorders progression?
Machine learning methods have great potential to solvie these problems; however they have a few limitations. First, they cannot learn what is not in the medical data (e.g., high-quality 7T MRI reconstruction of an image from low-quality 3T MRI with a lesion while learning from images without any lesions). Second, they are guided by ‘what we assume we know about the medical data’. Third, they are somewhat biased by our limited knowledge. Last, they usually require the use of large medical dataset.