Ophthalmic Medical Image Analysis MICCAI 2020 Workshop Oct 08, 2020 12:00 AM — Virtual Meeting (online)
Predictive Intelligence in Medicine MICCAI 2020 Workshop Oct 08, 2020 12:00 AM — Virtual Meeting (online)
PET is Wonderful Annual Meeting 2020 Oct 27, 2020 02:00 PM - 05:40 PM — Virtual Meeting (online)
NRS Mental Health Network Annual Scientific Meeting 2020 Nov 04, 2020 09:00 AM - 05:30 PM — Virtual Meeting (online)

eLearning

SINAPSE experts from around Scotland have developed ten online modules designed to explain medical imaging. They are freely available and are intended for non-specialists.


Edinburgh Imaging Academy at the University of Edinburgh offers the following online programmes through a virtual learning environment:

Neuroimaging for Research MSc/Dip/Cert

Imaging MSc/Dip/Cert

PET-MR Principles & Applications Cert

Applied Medical Image Analysis Cert

Online Short Courses

Jenny Lines

Position: SINAPSE PhD Student
Institute:
Department:


Description of Phd:

A limitation of current ERP studies is a reliance on analysis of participants’ responses averaged across trials. Because by-item variance is not accounted for in these analyses, experiments must include large numbers of observations (typically around 30-50, but with a standard minimum of ~16 per condition, per participant). In practice, this feature of ERPs represents a limit to their use; if the data recorded from a given participant does not include enough ‘clean’ observations per cell it is discarded wholesale. This can make data collection wasteful, or at worst, it can simply prevent particular paradigms or populations being studied. The current proposal addresses this problem by developing item-based ERP analyses.

The aim of the current project is to develop analysis techniques for ERPs based on mixed effects models, using the freely available R statistical programming language (a user-developed and extremely powerful statistical framework) together with available libraries (e.g., Bates et al., 2008). In practice, many experiments include items that either comprise a bounded set (letters of the alphabet) or a sample from an infinite population (sentences in a language). Using mixed-effects modelling approaches (Bagiella et al., 2000; Baayen et al., in press), items can be included as fixed or random effects in a model, at the same time as participants are included as random effects. These approaches have the additional advantage that they are highly effective when analysing unbalanced data, and cope well with covariance (e.g., due to high-density electrode arrays, traditionally requiring a correction for non-sphericity). The project will result in a set of R scripts and an analysis protocol that can be used to develop and analyse experiments where fewer ERPs can be measured per participant, or where issues in recording may prevent the collection of balanced data sets (as is likely in some clinical populations).
As a test case, we will collect data based on previously published work (Collard et al., 2008; Corley et al., 2007) using paradigms that are known to suffer from trial-number problems. In these studies we will investigate the memory processes underlying the encoding, and recognition, of words in spoken sentences. Because these studies rely on participants’ memory performance they are by their nature subject to low (and unbalanced) numbers of observations, and will therefore provide an ideal dataset for the development of analysis techniques that should be of considerable benefit to the wider SINAPSE community.