Viljami Sairanen
LightNews LightNews
banner
viljamisairanen.bsky.social
Viljami Sairanen
@viljamisairanen.bsky.social
420 followers 450 following 73 posts
Medical Physicist | PhD | Entrepreneur

Researching brain models (dMRI) at Baby Brain Activity Center of the University of Helsinki, https://www.babacenter.fi/

Department chief physicist in radiology at Hämeenlinna Central Hospital, Finland
Posts Replies Media Videos
viljamisairanen.bsky.social
Is this a tractogram of the default mode network or a Romulan warship from Star Trek?
September 18, 2025 at 7:21 AM
viljamisairanen.bsky.social
Silly ChatGPT 🤭
January 26, 2025 at 7:15 AM
viljamisairanen.bsky.social
Tortinon rakenne koostuu lehtevästä pohjataikinasta, jonka päällä on rapeaksi paistunut marenkimainen ja makea kuorrute sekä kirpeä luumumarmeladi. Marmeladin määrällä voi säätää leivonnaisen happamuutta. Liikaa ei sovi laittaa, koska tällöin taikina ei jaksa nousta
December 19, 2024 at 8:59 AM
viljamisairanen.bsky.social
1 kg torttutaikinalevyjä leikataan 2 - 3 cm levyisiksi suikaleiksi.
December 19, 2024 at 8:59 AM
viljamisairanen.bsky.social
Oletko aina tehnyt joulutortut väärin?

Laitetaanpa lusikka tähänkin luumusoppaan. Muotoillaan klassikkoleivonnainen uudelleen, jolloin myös luumumarmeladi on mahdollista levittää tasaisemmin. Näin joka haukkaus maistuu yhtä hyvälle!

Tässä joululahjani suomelle: joulutortini glassati
December 19, 2024 at 8:59 AM
viljamisairanen.bsky.social
Hey to all the new people here: please write _something_ to your bio (topic of your research, interests etc). Makes it easier to decide to follow back or not.

This is not only to avoid bots but also quite interesting to learn what people are doing!
November 25, 2024 at 8:21 AM
viljamisairanen.bsky.social
Here is an example in Python how to combine functional and structural #brain connectivity (fc&sc) matrices:

github.com/vilsaira/FSC

Surprisingly simple to apply:
> from fsc import fsc
> fscm = fsc(V=fc, R=sc).get_I()

biorxiv.org/content/10.1...

#dipymri #Connectivity #fMRI #dMRI
March 8, 2024 at 8:37 AM
viljamisairanen.bsky.social
Three years from idea to preprint but here it is: a macro-scale model combining functional and structural connectivity matrices of the human brain networks.

This is a technical report so not applied to any clinical data yet.

www.biorxiv.org/content/10.1...

#neuroimaging #fMRI #dMRI #connectivity
March 7, 2024 at 7:46 AM
viljamisairanen.bsky.social
Interested in #Diffusion #MRI ? Look no further! A #free talk, this Friday 10:00 UK time, on exactly that! Looking forward to hearing about how to validate diffusion biomarkers! Talk provided by BIC #ISMRM

If you need a link, let me know.
October 31, 2023 at 6:22 AM
viljamisairanen.bsky.social
Am I unaligned?
October 23, 2023 at 10:37 AM
viljamisairanen.bsky.social
Laitetaas tämä myös suomeksi #minätutkin teeman alle!

Tutkin ihmisaivojen matemaattista mallinnusta MRI-kuvien pohjalta. Mitä käy, kun mittaus meneekin pieleen ja voiko virheen korjata? Oleellista esimerkiksi tutkittaessa vastasyntyneiden aivoja.

Lisätietoa: www.sciencedirect.com/science/arti...
The outline of the experiments used in this study. Subject motion during a dMRI acquisition can cause slice-wise dropout artefacts. These artefacts can be remedied by using outlier replacement with different predictions e.g. from Gaussian Processes (GP) or from Spherical Harmonics (SH). Alternatively, an outlier position can be traced throughout the motion correction registrations and used to downweight such unreliable information during model estimation. We set out to investigate these different approaches using whole-brain simulations with the aim to find out which approach would the most beneficial to use in practice.
October 20, 2023 at 11:00 AM
viljamisairanen.bsky.social
Wait.. what? That escalated quickly.
A picture of the abstract and keywords for scientific paper titled "The Insight ToolKit image registration framework". The last keyword "death" is highlighted because how that describes this work is beyond me.
October 20, 2023 at 8:28 AM
viljamisairanen.bsky.social
For #dMRI folks: what to do if your data contains motion outliers? Should you: Ignore them (definitely not!), replace them, or use weighted modelling?

Replacement can lead into inflation of model precision as shown in figure. More details here: www.sciencedirect.com/science/arti...
Difference in the model variance between replacing and downweighting outliers. Lines indicate the average and shaded areas indicate the standard deviation calculated over the ten simulated brains. Due to only one 0% frequency simulation, there is only one data point and no variance to show. With replacement approaches, the estimated precision of the model parameters is inflated which is seen as a downward trend for both GP repol and SH repol. In contrast, with the downweighting approach, the estimated precision decreases as the outlier frequency increases which is seen as an upward trend for both GP weighted and SH weighted. The latter is the expected behaviour, as clearly the precision will go down with the increasing frequency of missing data. Regardless of the approach, when the point of too many outliers is reached (approximately 12% here), problems arising from image registration start to take effect and the linear trends disappear.
October 20, 2023 at 7:59 AM
viljamisairanen.bsky.social
Once upon a time, I went to Oxford and this came out 👇

www.sciencedirect.com/science/arti...

How to handle outliers in diffusion-weighted brain MRI and what kind of consequences your choices might have?Probabilistic tractographers working with clinical data might want to check this!

#neuroimaging
An illustration of different robust processing pipelines for dMRI data.
October 19, 2023 at 10:01 AM