Disconnectome studio - Web application

1

http://disconnectomestudio.bcblab.com

In this channel, you can post your questions on the disconnectome studio web application and the Disconnectome Symptom Discoverer (DSD) tool functionality to predict the 1-year aspected neuropsychological scores from acute stroke disconnectome images.

2

Hi, I am attempting to predict motor scores from lesion masks using the DSD. Do I need to first calculate disconnectomes and then upload those to http://165.232.73.88/motor_page_calc/? If so, I believe the link to calculate the disconnectomes (“Calculate disconnectome first”) is broken. Thank you!

3

Hi, yes, you need to calculate the Disconnectome maps. They must be registered to the MNI152 2mm space registered, and .nii.gz format. If those requirements are fulfilled, the web-application runs smoothly. For this reason, you thought the link was broken. To overcome this issue, press the “Download patient Disconnectome example” on the home page, and you will download a Disconnectome template. Try it on the app, and you will see that the web-application is working. Thanks. Lia

4

Hi, thank you for the response. I am not trying to upload a disconnectome map at this point.
When I click “Calculate disconnectomes first” from the homepage (http://165.232.73.88/), I am taken to this page with this message:

Screen Shot 2023-02-03 at 1.40.49 PM
Screen Shot 2023-02-03 at 1.40.49 PM1726×556 82.9 KB

Thanks,
Emily

5

Hi Emily, apologies about that. We moved the website, and some links still need to be updated.
Essentially this page is supposed to lead you to bcbtoolkit so that you can compute your disconnectome map. For maximal compatibility with the DSD you might want to use 2x2x2mm lesions and boost your disconnectome with [DISCONNECTOME Package X (180 additional subjects HCP7T 2mm) – Thiebaut de Schotten et al. Nat Comm 2020 available at Open_data cheers mich

1 Like
6

The link has just been updated!

7

Hello,

I have just used the DSD to predict some motor scores for a sample of chronic stroke subjects. I am noticing that the variance of the predicted scores is quite low for some of the measures. Can you confirm that that is the expected behavior of the model?

Thank you,
Emily

8

Hi Emily,
Yes, we also noticed that when an impairment is expected, the variance of the predicted scores is reduced compared to measured scores. This is due to the UMAP coordinate smoothing and consequent averaging of the neuropsychological score in the UMAP space. However, in our analyses, the predicted score variations agreed with the measured scores. The obtained model R2s are reported in our preprint and upcoming publication. Moreover, in the supplementary materials, the mean absolute error % is reported to estimate the difference between the predicted and measured scores. On average, it was below 20% among the scores.
I hope this was helpful.
Cheers
Lia

10

Hello! Thank you for creating the disconnectome web application.

I was wondering if there is any way to upload multiple .nii.gz disconnectomes and then download the corresponding .csv files, as opposed to uploading one .nii.gz disconnectome file at a time?

Thank you

Best regards

11

Hi Beatrice, unfortunately, the web application does not have multiple image processing. However, we can provide standalone Python codes that you could run on your server at a population level. Here my email for more information on this (liatalozzi@gmail.com). Thanks. Lia

12

Hello, thank you very much for your reply. I sent you an email. Best, Beatrice

13

Continuing the discussion from Disconnectome studio - Web application:

Hi, is the ‘Calculate disconnectome’ function still unable to be used online directly? When I click “Calculate disconnectomes first” from the homepage (http://165.232.73.88/), the page return to the bcblabweb (Page 1)).

14

Is there simple method to calculate the Disconnectome maps. :face_with_raised_eyebrow:

15

@TongjianBai @iamolivasmith yes you just download the bcbtoolkit on our webpage and the package x from the open data section of our website and then run the disconnectome from bcbtoolkit. This will convert your data in disconnectomes like in Talozzi et al 2023 were the methods indicate all steps:

“Stroke lesions were manually delineated in MRI scans and subsequently normalized to the MNI152 space (2 mm resolution) using the enantiomorphic normalization tool in the BCBtoolkit (http://toolkit.bcblab.com). Accordingly, before registering the patient’s T1-weighted image to the MNI152, the lesioned areas were replaced with the contralateral healthy tissues to calculate the normalization transformation (54) Then, disconnectome profiles were processed with the BCBtoolkit (55) HCP tractography was filtered considering only streamlines passing through each stroke lesion. To obtain a normative population group statistic, every filtered tractography was binarized. Thus, if at least one streamline passed in a voxel, the voxel value would be one. This step allowed the creation of a summarising percentage (%) map. Namely, for each stroke patient, a map ranging from 0 to 1 was obtained according to the number of HCP participants who would have reported a streamline disconnection in that voxel.”

happy to help if you are struggling at any point.
cheers

mich