User-centered design of a Brain-Computer Interface headset using modern anthropometric methods

In this doctorate, the possibilities of 3D anthropometry for product development were explored by applying a statistical shape model of the human scalp to the design of more ergonomic brain-computer interface (BCI) headsets. First, a statistical shape model of the human scalp was created from a set of 100 MRI scans. This model was parametrized by intuitive anthropometric measurements and evaluated in terms of its ability to predict complete scalp surfaces given a set of anthropometric measurements. Using eight anthropometric measurements resulted in an average prediction error of only 1.60 +/- 0.36 mm, indicating the model accurately represents the underlying population. The choice of parametrization measurements should be based on their combined prediction errors, their sensitivity to variation in input measurements and a minimum population percentage that remains below a predetermined prediction error threshold. Next, the use of the statistical shape model for comparing the morphological differences between subpopulations and the application to the design of BCI headsets were briefly discussed. After this, the shape model of the scalp was used for the design of one-size-fits-all BCI headset with 14 electrode channels. Electrode placement, stability and reliability of the prototype headset were evaluated and compared to current EEG practices, as well as to a commercial BCI headset. The prototype met all design standards and performed well within EEG practices. It also offered 10\% improvement in electrode placement according to the international 10-20 system and a 15% increase in reliability. A functional headset of this type would therefore be more consistent in longitudinal BCI studies and between studies of different research groups. The results prove that 3D anthropometry is a feasible design method for a one-size-fits-all BCI headset. Following this observation, the application of 3D anthropometry for product sizing was considered. Whereas sizing systems are usually based on statistical clustering of one-dimensional head measurements, a new method for 3D head shape clustering was proposed, taking into account the need for intuitive sizing and simple sizing tables. The method was labeled "constrained clustering" and was compared to clustering of traditional anthropometric features as well as unconstrained k-medoids clustering of the 3D shapes. Intra- and inter-cluster scalp shape variability and within-cluster point-to-point distances were used as criteria. The results of constrained clustering were similar to those of unconstrained k-medoids clustering of head shapes and offered a 20.69% improvement in cluster validity index and a decrease of size-weighted variances by 6.6% compared to traditional feature-based clustering. This research resulted in three journal publications that form the main part of this thesis. This doctorate proves that head-based products that require accurate shape and size fit would benefit from a design process in which 3D shape models are included, and that 3D anthropometry has a place in the product design process. Compared to traditional anthropometry, the use of 3D anthropometry will result in devices that are better fitting, more comfortable and potentially even more functional.