Machine Learning for Objective Diagnosis of Dystonia
Classification of spasmodic dysphonia from the normal state as well as discrimination between different phenotypes and genotypes of the disorder based on altered functional connectivity of cortical regions (Battistella et al., 2016, Eur J Neurol).
The diagnostic accuracy of dystonia is unreliable, and the consensus between physicians is generally hard to achieve. The development of an objective machine learning-based objective diagnosis of dystonia would be imperative in significantly reducing the burden of the physician’s rate of mis- and under-diagnosis, improving the patient’s quality of life, and greatly decreasing the overall costs associated with delayed treatment of this disorder. Funded by the National Institute on Deafness and other Communication Disorders, National Institutes of Health (NIDCD/NIH R01DC011805), our current studies combine large-scale multimodal neuroimaging (structural and functional MRI) with advanced machine-learning algorithms to develop objective and accurate diagnostic markers of spasmodic dysphonia. Supported by Amazon Web Services, we use convolutional neural networks for the development of automated tools of dystonia diagnosis.