A framework for hybrid machine and human computation for the accurate and scalable analysis of human clinical EEG recordings

点击翻译按钮获取中文摘要

Electroencephalography (EEG the measurement of human brain activity using electrodes on the scalp) is a key tool in the diagnosis of epilepsy. It is also a key step in the diagnosis of sleep disorders. Right now, interpretation of EEGs is dependent on specialized neurologists. This limits access to EEG services in smaller communities. Even in large communities, the laborious nature of EEG interpretation and the heavy reliance on specialists imposes substantial costs on the health care system, and can delay diagnosis and treatment. Full automation of EEG interpretation could potentially address these issues. However, attempts to do so have had limited success, in part because some aspects of EEG interpretation such as visual pattern recognition, while relatively easy for trained human experts, are very difficult to automate. Human computing is a field of computer science that aims to combine automated algorithms with judicious application of human input into one framework to leverage the strengths of each while avoiding the limitations of either. Human computing has proven to be a powerful tool to solve many key problems in science that had proven intractable to standard artificial intelligence approaches. We propose to build a system combining state of the art machine learning algorithms and human computing approaches to enable rapid, scalable, accurate interpretation of human EEGs, while minimizing the dependence on experts. We will adapt this system for use in Canadian hospitals and also integrate it with a unique smartphone-based EEG recording device to enable EEG diagnosis in communities without local specialized neurologists. Benefits will include improved timeliness and accuracy of epilepsy and sleep disorder diagnosis, accompanied by reduced costs. For more remote regions in Canada and elsewhere, our smartphone-integrated device will provide local access to accurate and cost-efficient EEG diagnosis avoiding the need to travel to larger centers.