Query on use of FreeSurfer with Clinical Epilepsy MRI Protocols (Anisotropic Voxel Size)
External Email - Use Caution
Dear FreeSurfer Support Team,
Season's Greetings! I hope this email finds you well.
May I take this opportunity to seek your support regarding the use of FreeSurfer for structural MRI data acquired using clinical epilepsy imaging protocols with non-isotropic voxel sizes in our hospital setting.
To date, I have processed MRI data from over 90 epilepsy patients using *FreeSurfer v 6.0* (3T scanner 64-channel head coil, TA: 3:49 min; Voxel Size: 1.0 × 1.0 × 0.5 mm³; Acceleration factor: 2; Relative SNR: 1.00). The complete recon-all pipeline ran successfully for all subjects, generating cortical and subcortical segmentations as well as quantitative statistics files without any processing errors.
However, I learnt that 1 mm³ isotropic resolution is recommended for optimal performance, although voxel sizes up to *1.3 × 1 × 1.3 mm³* are considered acceptable to reduce scan time or improve signal-to-noise ratio, as per FreeSurfer Wiki. Therefore recently, I have acquired one additional patient data using an isotropic protocol (1.0 × 1.0 × 1.0 mm³) and compared to epilepsy protocol mentioned above.
To conclude , I got two different quantitative values. For example, in Left Hippocampus - Anisotropic : 3640.2 mm³ vs Isotropic : 3876.2 mm³ , Total intracranial volume- Anisotropic : 1504943.411530 mm³ vs Isotropic : 1529703.225834 mm3 , Mean cortical thickness – Anisotropic : 2.18512 mm vs Isotropic: 2.28844 mm.
In this context, I would greatly appreciate your guidance on the following:
1. Would you recommend resampling 1 × 1 × 0.5 mm³ images to 1 mm³ isotropic resolution prior to running recon-all, or is it fine to process the native anisotropic resolution data?
2. Are there any appropriate FreeSurfer steps or commands to convert my already acquired 1 × 1 × 0.5 mm³ MRI data to isotropic resolution before running the FreeSurfer recon-all pipeline?
3. Are there any known limitations, biases, or segmentation inaccuracies, particularly affecting hippocampal volumetry or cortical thickness estimates, when FreeSurfer is applied to anisotropic voxel data in epilepsy research like I did?
I am attaching the recon-all log files of both the protocols here.
Thank you very much for your time and guidance. I would greatly appreciate your recommendations. -- Warm regards, *Uttara Bakshi* *Doctoral Fellow* *Institute Of Neurosciences, Kolkata*
External Email - Use Caution
*Uttara Bakshi* *Doctoral Fellow* *Institute Of Neurosciences Kolkata*
---------- Forwarded message --------- From: Uttara Bakshi uttarabakshi4@gmail.com Date: Thu, 15 Jan, 2026, 1:14 pm Subject: Query on use of FreeSurfer with Clinical Epilepsy MRI Protocols (Anisotropic Voxel Size) To: freesurfer@nmr.mgh.harvard.edu
Dear FreeSurfer Support Team,
Season's Greetings! I hope this email finds you well.
May I take this opportunity to seek your support regarding the use of FreeSurfer for structural MRI data acquired using clinical epilepsy imaging protocols with non-isotropic voxel sizes in our hospital setting.
To date, I have processed MRI data from over 90 epilepsy patients using *FreeSurfer v 6.0* (3T scanner 64-channel head coil, TA: 3:49 min; Voxel Size: 1.0 × 1.0 × 0.5 mm³; Acceleration factor: 2; Relative SNR: 1.00). The complete recon-all pipeline ran successfully for all subjects, generating cortical and subcortical segmentations as well as quantitative statistics files without any processing errors.
However, I learnt that 1 mm³ isotropic resolution is recommended for optimal performance, although voxel sizes up to *1.3 × 1 × 1.3 mm³* are considered acceptable to reduce scan time or improve signal-to-noise ratio, as per FreeSurfer Wiki. Therefore recently, I have acquired one additional patient data using an isotropic protocol (1.0 × 1.0 × 1.0 mm³) and compared to epilepsy protocol mentioned above.
To conclude , I got two different quantitative values. For example, in Left Hippocampus - Anisotropic : 3640.2 mm³ vs Isotropic : 3876.2 mm³ , Total intracranial volume- Anisotropic : 1504943.411530 mm³ vs Isotropic : 1529703.225834 mm3 , Mean cortical thickness – Anisotropic : 2.18512 mm vs Isotropic: 2.28844 mm.
In this context, I would greatly appreciate your guidance on the following:
1. Would you recommend resampling 1 × 1 × 0.5 mm³ images to 1 mm³ isotropic resolution prior to running recon-all, or is it fine to process the native anisotropic resolution data?
2. Are there any appropriate FreeSurfer steps or commands to convert my already acquired 1 × 1 × 0.5 mm³ MRI data to isotropic resolution before running the FreeSurfer recon-all pipeline?
3. Are there any known limitations, biases, or segmentation inaccuracies, particularly affecting hippocampal volumetry or cortical thickness estimates, when FreeSurfer is applied to anisotropic voxel data in epilepsy research like I did?
I am attaching the recon-all log files of both the protocols here.
Thank you very much for your time and guidance. I would greatly appreciate your recommendations. -- Warm regards, *Uttara Bakshi* *Doctoral Fellow* *Institute Of Neurosciences, Kolkata*
freesurfer@nmr.mgh.harvard.edu
-
Uttara Bakshi