see below

On 29 Aug 2017, at 7:00 pm, Matteo Respino <[log in to unmask]> wrote:

Hi Ludovica,
here's what you indicate as a necessary first step in the Userguide and I have a few questions about it
Before running BIANCA you need to:
Choose your base image. This will be your reference space for your input and output images.
Perform brain extraction on (at least) one modality to be able to derive a brain mask *
Register all the other modalities to your base image for each subject.
1. Does my "base image" need to be a structural image or an MNI is also ok? do you have any specific suggestion?

The base image is an image in subject space that you pick as reference modality. For each subject, you register the other modalities to the base image (of that subject) and you will get your WMH to be segmented in the same space. I suggest to register everything to a T2 or FLAIR image, as it has the best contrast for WMH (in this way you would avoid resampling, possibly loosing information or blurring the edges of the WMH). A fair alternative is to use a T1 image, which is usually the one with the best resolution. In our paper using either T1 or T2-FLAIR as base image gave comparable results.

You don’t need to register everything in MNI space, as BIANCA works at single subject level. If you need to do further analyses in MNI space I would suggest to run BIANCA in single subject space and then register to MNI.

2. I performed fsl_anat and the command to get a bianca_mask: I have no clarity of idea in relation to whether I have to premask or postmask or both

The masking is not mandatory, but improves the results. You need to do it either before or after BIANCA, not both.

If you do it before, you should mask one of the input image with the mask created by BIANCA and give that modality to the option brainmaskfeaturenum. In this way you limit the area where BIANCA will look for WMH (possibly reducing the heterogeneity of the training points selected as non-lesion). If you do it afterwards you can have a look at the results in the full brain and decide whether to mask or not. I never quantitatively compared the use of the mask pre or post, but I doubt the difference would be massive.

3. Does a linear registration work fine enough?

If you mean registration within subject to register all the modality to the base image, you actually need to use linear registration with 6 DOFs, as it’s the same subjects, so you don’t expect deformations from one modality to the other.

For the use of the spatial features you need a transformation matrix from the single subject base image to MNI. In this case, currently BIANCA only supports linear registration, so you will need a 12 DOFs registration from single subject to MNI space. We are working to include the possibility to use a non linear registration warp field instead of the registration matrix to improve the spatial feature extraction, but it will be included in a future version of BIANCA.

Thanks,
Matteo

Hope it helps,

Ludovica

—

Ludovica Griffanti, PhD
Analysis Postdoctoral Research Assistant
Wellcome Centre for Integrative Neuroimaging (WIN)

Oxford Centre for Functional MRI of the Brain (FMRIB)
Nuffield Department of Clinical Neurosciences, University of Oxford
John Radcliffe Hospital
Oxford, OX3 9DU, UK
email: [log in to unmask]

2017-07-28 4:57 GMT-04:00 Ludovica Griffanti <[log in to unmask]>:
Dear Matteo,
yes, BIANCA requires some images where the WMH are labelled, to learn how to classify them in the rest of the subjects.
You will need to prepare some binary images of WMH (1 inside the lesions, 0 outside) from your data (usually FLAIR is used as the WMH are better identifiable)
I would suggest 10 subjects to start with (in our experience it works better if you use the ones with higher WMH load) and then see if you are happy with the results or you need to add more.
Hope it helps,
Ludovica

—
Ludovica Griffanti, PhD
Analysis Postdoctoral Research Assistant
Wellcome Centre for Integrative Neuroimaging (WIN)
Oxford Centre for Functional MRI of the Brain (FMRIB)
Nuffield Department of Clinical Neurosciences, University of Oxford
John Radcliffe Hospital
Oxford, OX3 9DU, UK
email: ludovica.griffanti@ndcn.ox.ac.uk

On 27 Jul 2017, at 7:53 pm, Matteo Respino <[log in to unmask]> wrote:

Hi everyone,

it's the first time I try Bianca and I've been reading the provided user guide.
I do not get the "training dataset preparation" step: am I supposed to manually mask the WMH within some of my FLAIR images?
If yes, how many sbj are required for having a training set?

Thanks in advance
Best
Matteo

--
Matteo Respino, MD
Resident in Psychiatry
Visiting Research Fellow in Psychiatry

Clinica Psichiatrica Universitaria
Facolta' di Medicina e Chirurgia di Genova
Largo Rosanna Benzi, Padiglione 32A
Genova, 16100, Italia
Weill Cornell Medicine
21 Bloomingdale Rd,
White Plains, NY 10605