Research directions
The work sits where imaging physics, signal processing and clinical neuroscience meet. Seven strands, held together by a single question: what can a measured signal tell us about the brain that produced it?
SIGNALS
The acronym is the map of the lab.
The voice as a clinical signal
Paralinguistic feature extraction, prosody and timing analysis, and the statistical treatment of speech for distinct mental health conditions. Applied in the Speech Health Lab.
MRI, fMRI, diffusion and flow
Sequence physics and acquisition trade-offs, structural and functional analysis pipelines, diffusion modelling, and 4D flow. Grounded in NMR physics.
Brain connectivity
Structural and functional connectomes, graph-theoretic description of network organisation, and the methodological question of how much of a connectome is biology and how much is pipeline a computational framework.
Systems and neurophysiology
The anatomy and physiology that any imaging result has to be reconciled with. The brain atlas came out of this strand. Many more to come in future across Preclinical domain.
Modelling and computational psychiatry
Statistical modelling for psychology and psychiatry, multimodal biomarkers, and the application of machine learning and medical language models to clinical data with attention to software development for long term sustainability.
Teaching and tooling
Open educational resources on MRI and NMR foundations, statistics and coding, and research communication alongside simulation work that makes the underlying physics visible. Still a work in progress, but the brain atlas is a first step. SignalsLab Review, our application for systematic reviews and references, is free software.
SignalsLab Review
We designed and built our own application for systematic reviews and references, and we run our reviews in it. Free software under the GNU AGPL-3.0; the source code goes to GitHub shortly. The application page describes it; a copy is sent on request.
Four gated stages
Title and abstract (Yes / No / Maybe), full text with the PDF beside the decision, extraction with no blank fields, risk of bias one study at a time in each tool's own words. Every exclusion carries one reason from the protocol's ordered list; conflicts are resolved, proposed and approved before the next stage opens.
Independent reviewers, one record
Each reviewer works on their own computer and compares only afterwards: conflicts of decision or reason are listed, resolved and approved before the next stage opens. Every decision carries who made it, when, and under which protocol version, and nothing is ever deleted.
A project folder and a library
A review lives in a folder you choose: the raw exports per database, the deduplicated library, the EndNote 21 library, every PDF, the exports and the protocol. The library inside the application reads like a reference manager (collections, table, item pane), and its RIS export carries the PDF links Zotero understands.
Study ids, Meta Lab, PRISMA 2020
Included studies receive first-author-and-year ids that follow them into extraction, the bibliography and the exports. PRISMA counts are assembled from the records; the checklist is written into the project folder with what the review already knows. The statistics that reach a manuscript are computed in R; the application's Meta Lab is exploratory.
Built on open work: Zotero, ASReview, SQLite, pdf.js and citation.js. Windows today, macOS to follow.
Publications
A current list of publications is maintained on ORCID and ResearchGate.