How to create a lesion mask for FCD type I lesions?

FCD type I lesions can be very subtle if not impossible to see on MRI. However, as they are some of the most challenging lesions to detect, we definitely want to include examples of them to train our classifiers on.

If you cannot see the lesion or the lesion boundaries on MRI, we suggest that you co-register the post-surgical scan to the pre-surgical scan and use the resection cavity as a mask. This will likely be an overestimate of the lesion. However, it will still contain useful data. Please make a note in the demographics spreadsheet in the “Notes” column.

Does the lesion mask need to be done by a radiologist?

If a researcher / neurologist can see the lesion, it is fine for them to mask it, particularly if they have experience with looking at FCDs. However, for the difficult, subtle lesions or if the researcher is less experienced with FCDs, you may need neuroradiology input. We will send a survey round once centres have done the lesion masking to ask the experience level of the person doing the lesion masking.

How many controls do I need?

We separate controls into ones scanned on 3T and 1.5T scanners. The code will check if you have more than 25 controls scanned on the same field strength scanner as your patients and only normalise the data using the controls if you have more than 25. If you have less than 25 controls, the code will use a control template we have created instead.

If your controls have been scanned using a different scanner or sequence, you can still include them. However, we would like to know this information so that we can check if / how it has an impact on the analyses. In due course we will send out a questionnaire to find out how many scanners / sequences have been used.

  • Please note the final matrix that is shared with us also contains the data before it has been normalised by controls. This means that if we need to renormalise the data using controls acquired from all sites (i.e. a much larger control dataset) we will be able to. *

Can subjects with tension headaches / migraine be used as controls?

Yes. However, please mention this in the “Notes” column of the csv file.

Can patients with FCD type III be included?

Yes. However, as the cortical surfaces do not segment the hippocampus, the analyses will only include the cortical dyplasia and not the hippocampus.