In order to understand how the human brain works, a common step is to measure signals originating from neuronal activity, or to interfere with brain functioning via external stimulation. These techniques do not answer to the question how, per se, but they give vital information about where and when brain is activated due to the stimulation or during the task performance such as speech. This information can help us understand the investigated brain mechanisms and behavior. In this Thesis, I developed experimental and analysis techniques for navigated transcranial magnetic stimulation (nTMS) and for electro-/magnetoencephalography (EEG, MEG), with a particular focus on providing means to answer where in the brain the interesting processes take place.
Navigated TMS allows us to interfere with brain activity non-invasively. As the navigation allows anatomically accurate targeting of the stimulation to a speciﬁc area in the brain, the stimulated target is causally linked to the evoked outcome measure, for example, word-production errors or hand movement. EEG and MEG are used to measure the electromagnetic signals generated by the brain with a set of sensors placed over the head. They provide useful information about the evoked brain activity, measured directly from the brain.
We introduced a method for locating cortical language areas non-invasively with nTMS. When short TMS bursts are delivered to different brain areas time-locked with a picture-naming task, language areas are revealed by stimulation-evoked naming errors. This "nTMS language mapping" provides a spatial distribution of cortical areas related to language. This information is useful in language studies and also when the neurosurgeon needs to know which areas of the brain must be protected during operation. We applied nTMS during different naming tasks and in bilingual speakers of Finnish and Swedish, showing that task parameters affect the outcome of nTMS.
We also developed novel source localization techniques to analyze EEG and MEG. These methods belong to the family of multiple signal classiﬁers (MUSIC). Two different method types were introduced for different applications; they were compared with other source localization methods in simulations and with measured MEG data. Conventional recursive MUSIC has been shown to be efﬁcient in estimating temporally uncorrelated or weakly/moderately correlated sources, but the method can be unreliable when estimating the number of sources; our ﬁrst method called TRAP-MUSIC solves this problem. On the other hand, if the data are also due to highly correlated and synchronous sources, neither conventional methods nor TRAP-MUSIC are sufﬁcient. Our second method, RDS-MUSIC, provides a solution to this issue.
The presented methods provide new possibilities to neuroscientists for investigating where the brain processes governing human behavior and actions, in particular, but not limited to, language, take place.
|Publication status||Published - 2018|
|MoE publication type||G5 Doctoral dissertation (article)|
- language, source localization, inverse methods