|
New neuroinformatics approaches to predicting and understanding mental health| title | New neuroinformatics approaches to predicting and understanding mental health |
|---|
| start_date | 2024/02/15 |
|---|
| schedule | 14h |
|---|
| online | no |
|---|
| location_info | Amphithéâtre Serge Kampf |
|---|
| summary | Recent technological advances, and the availability of increasingly large, multimodal datasets, have created high hopes for AI to shed fresh light on our understanding of mental health. My group develops neuroinformatics approaches towards two core scientific goals: 1) to build a better understanding of the brain across the lifespan, in both health and disease; and 2) to predict development and mental health, at the individual level. To that end, we use a wide range of data modalities, including brain MRI and speech.
In this talk, I will give an example of work in the first direction, in which we propose a new computational method for robust estimation of structural similarity between brain regions from MRI data, as a marker of structural brain connectivity (Sebenius et al, Nature Neuroscience 2023). Our new Morphometric INverse Divergence (MIND) method showed greater correspondence with gold standard macaque tract-tracing data, greater consistency across subjects and higher heritability than previous approaches (Seidlitz et al., Neuron 2018). We also observed a remarkably strong association (r=0.73) between MIND structural brain similarity and transcriptomic similarity from the Allen Human Brain Atlas (Hawrylycz et al., Nature Neuroscience 2015); providing compelling fresh evidence that the cortex shares a unified structural and transcriptomic organisation. Finally, MIND outperformed prior methods, including DTI networks, in an age prediction task, suggesting MIND has potential to provide a new perspective on the principles of cortical organisation that govern development and ageing.
I will also give a brief overview of our work in the second direction, using speech and language data to predict mental health outcomes. One recent focus has been developing Netts- a new tool that maps the semantic content of speech as a network (Nettekoven et al., Schizophrenia Bulletin 2023). Speech networks generated by Netts showed greater fragmentation in patients with psychotic disorders, suggesting they are sensitive to alterations in patients’ speech. Ultimately, automated approaches to studying behavioural data at scale could revolutionise our understanding and treatment of mental health conditions. |
|---|
| responsibles | Sadoul |
|---|
Workflow history| from state (1) | to state | comment | date |
|---|
| submitted | published | | 2024/02/08 14:02 UTC |
| |
|