Time and cost factors are of major importance in the context of analyzing human clinical EEG data because experienced experts like neurologists or EEG technicians are both rare and expensive. One of our ongoing studies in the field of active machine learning, therefore, tries to answer the question how we can automate the process of accurately classifying a given EEG record into sleep stages while at the same time minimizing the number of times the machine classifier has to query a human expert for information about the correct sleep stage.
The CrowdEEG team is currently conducting a study to investigate the effect of implicit contextual information in multi-channel biosignal time series data on the agreement rates among multiple clinical experts in a sleep staging task. The aim of the study is to identify different types of information that trained experts evaluate implicitly to inform their classification decision. In a subsequent step, we aim to make these supportive insights explicitly available also to non-experts to facilitate fast and reliable sleep staging in a large-scale crowdsourcing setting.