Comparing automated tractography with expert manual tractography in tumour and epilepsy patients
Poster No:
1287
Submission Type:
Abstract Submission
Authors:
Steven Greenstein1, Sila Genc1,2, Alison Wray3, Wirginia Maixner3, Joseph Yang2,4,5
Institutions:
1Developmental Imaging, Murdoch Children's Research Institute, Melbourne, Victoria, Australia, 2Neuroscience Advanced Clinical Imaging Service,Department of Neurosurgery, Royal Children's Hospital, Melbourne, Victoria, Australia, 3Department of Neurosurgery, Royal Children's Hospital, Melbourne, Victoria, Australia, 4Neuroscience Research, Murdoch Children's Research Institute, Melbourne, Victoria, Australia, 5Department of Paediatrics, University of Melbourne, Melbourne, Victoria, Australia
First Author:
Steven Greenstein
Developmental Imaging, Murdoch Children's Research Institute
Melbourne, Victoria, Australia
Developmental Imaging, Murdoch Children's Research Institute
Melbourne, Victoria, Australia
Co-Author(s):
Sila Genc
Developmental Imaging, Murdoch Children's Research Institute|Neuroscience Advanced Clinical Imaging Service,Department of Neurosurgery, Royal Children's Hospital
Melbourne, Victoria, Australia|Melbourne, Victoria, Australia
Developmental Imaging, Murdoch Children's Research Institute|Neuroscience Advanced Clinical Imaging Service,Department of Neurosurgery, Royal Children's Hospital
Melbourne, Victoria, Australia|Melbourne, Victoria, Australia
Wirginia Maixner
Department of Neurosurgery, Royal Children's Hospital
Melbourne, Victoria, Australia
Department of Neurosurgery, Royal Children's Hospital
Melbourne, Victoria, Australia
Joseph Yang
Neuroscience Advanced Clinical Imaging Service,Department of Neurosurgery, Royal Children's Hospital|Neuroscience Research, Murdoch Children's Research Institute|Department of Paediatrics, University of Melbourne
Melbourne, Victoria, Australia|Melbourne, Victoria, Australia|Melbourne, Victoria, Australia
Neuroscience Advanced Clinical Imaging Service,Department of Neurosurgery, Royal Children's Hospital|Neuroscience Research, Murdoch Children's Research Institute|Department of Paediatrics, University of Melbourne
Melbourne, Victoria, Australia|Melbourne, Victoria, Australia|Melbourne, Victoria, Australia
Introduction:
Diffusion MRI tractography is a valuable imaging adjunct for pre-surgical planning in neurosurgery (Essayed, 2017). Expert-based manual tractography is considered the clinical silver-standard, but the processing is time-intensive and requires neuroanatomical expertise. Automated tractography methods are being developed to address these challenges, but their performance remains uncertain in cases with pathology. This study aims to evaluate performance of four current state-of-the-art automated tractography methods against expert-based manual tractography in paediatric brain tumour and epilepsy surgery patients with challenging lesion imaging characteristics
Methods:
Fourteen brain tumour or epilepsy surgery patients at the Royal Children's Hospital were selected based on having peri-lesional oedema or diffuse infiltrative tumours and received expert-based manual tractography for pre-surgical planning (8 males, median age=11.46 years [interquartile range, IQR 8.41-13.62], median lesion volume=81.89 cm³ [IQR 57.87-131.77])
Pre-surgical 3T MR scans used included 3D T1-weighted data (voxel-sizes=0.8-1.0 mm) and multi-shell diffusion MRI (dMRI) data with multi-band accelerated EPI (2.3 mm³ isotropic voxels, TE/TR=77/3500 ms,11 interleaved b0s, b=1000 s/mm², 30 directions, and b=3000 s/mm², 60 directions). dMRI data were pre-processed with MRtrix3, FSL, ANTs to correct for imaging noise, Gibbs ringing artefacts, motion and susceptibility distortion, and bias field inhomogeneity.
Tractography was performed using multi-tissue constrained spherical deconvolution and iFOD-2 probabilistic algorithm in MRtrix3. Tracking ROIs were manually delineated based on expert anatomical priors. We evaluated the arcuate fasciculus, corticospinal tract, and optic radiation (n=84) bilaterally for all cases due to their common surgical implications. Manual tracts were used as references to evaluate automated tractography outputs.
We assessed four automated methods; TractSeg (Wasserthal, 2018); BundleSeg (St-Onge, 2023); Classifyber (Berto, 2021); White Matter Analysis (WMA; Zhang, 2018). Each method was performed with default settings to generate automated tract outputs for each patient.
Each automated method was compared to manual tracts by computing DSC, percentages of false-positive volumes (FP) for automated-only segments, and false-negative volumes (FN) for manual-only segments. FP and FN voxel distances from the lesion were calculated. Lesion and non-lesion side DSC, FP and FN were compared using Wilcoxon rank sum test, with significance at p<0.05
Pre-surgical 3T MR scans used included 3D T1-weighted data (voxel-sizes=0.8-1.0 mm) and multi-shell diffusion MRI (dMRI) data with multi-band accelerated EPI (2.3 mm³ isotropic voxels, TE/TR=77/3500 ms,11 interleaved b0s, b=1000 s/mm², 30 directions, and b=3000 s/mm², 60 directions). dMRI data were pre-processed with MRtrix3, FSL, ANTs to correct for imaging noise, Gibbs ringing artefacts, motion and susceptibility distortion, and bias field inhomogeneity.
Tractography was performed using multi-tissue constrained spherical deconvolution and iFOD-2 probabilistic algorithm in MRtrix3. Tracking ROIs were manually delineated based on expert anatomical priors. We evaluated the arcuate fasciculus, corticospinal tract, and optic radiation (n=84) bilaterally for all cases due to their common surgical implications. Manual tracts were used as references to evaluate automated tractography outputs.
We assessed four automated methods; TractSeg (Wasserthal, 2018); BundleSeg (St-Onge, 2023); Classifyber (Berto, 2021); White Matter Analysis (WMA; Zhang, 2018). Each method was performed with default settings to generate automated tract outputs for each patient.
Each automated method was compared to manual tracts by computing DSC, percentages of false-positive volumes (FP) for automated-only segments, and false-negative volumes (FN) for manual-only segments. FP and FN voxel distances from the lesion were calculated. Lesion and non-lesion side DSC, FP and FN were compared using Wilcoxon rank sum test, with significance at p<0.05
Results:
All automated methods successfully reconstructed non-lesion side tracts. On the lesion side, 83.3% TractSeg, 71.4% BundleSeg, 92.9% Classifyber, and 90.5% WMA tracts were reconstructed. However, all tracts varied greatly in their appearance (Figure 1). All automated methods yielded low DSC (median 0.44, IQR 0.34-0.50) and high FP (median 30.06%, IQR 16.02-43.68) and high FN (median 37.04% IQR 21.27-57.12). Lesion side tracts had significantly lower DSC and greater FN (both p<0.01), and marginally higher FP (p=0.079) compared with non-lesion side tracts. Most FP and FN voxels were situated close to lesions, suggesting that pathology-related anatomical distortion and peri-lesional oedema negatively impact automated tractography performance (Figure 2).
Conclusions:
Discrepancies between automated and manual tracts largely stemmed from differing tract anatomy definitions, highlighting a need for standardising tract definitions to improve automated tractography development and performance. Proximity of false tracking near the lesion caries significant surgical implications, as they could lead to adverse functional outcomes if used for surgical decision making. Our findings advise caution against relying on current automated tractography for complex paediatric tumour and epilepsy surgery cases, emphasising the need for improved training of automated model using similar surgical cases.
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural)
Diffusion MRI Modeling and Analysis 1
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
White Matter Anatomy, Fiber Pathways and Connectivity 2
Keywords:
Epilepsy
Neurological
PEDIATRIC
Tractography
Treatment
WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC
Other - Neurosurgery, brain tumour, automation
1|2Indicates the priority used for review
Abstract Information
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute your work in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels, the OHBM website, or other electronic publications and media.
The Open Science Special Interest Group (OSSIG) is introducing a reproducibility challenge for OHBM 2025. This new initiative aims to enhance the reproducibility of scientific results and foster collaborations between labs. Teams will consist of a “source” party and a “reproducing” party, and will be evaluated on the success of their replication, the openness of the source work, and additional deliverables. Click here for more information. Propose your OHBM abstract(s) as source work for future OHBM meetings by selecting one of the following options:
Please indicate below if your study was a "resting state" or "task-activation” study.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Was this research conducted in the United States?
Were any human subjects research approved by the relevant Institutional Review Board or ethics panel? NOTE: Any human subjects studies without IRB approval will be automatically rejected.
Were any animal research approved by the relevant IACUC or other animal research panel? NOTE: Any animal studies without IACUC approval will be automatically rejected.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references using APA citation style.
Essayed, W. I. et al (2017). White matter tractography for neurosurgical planning: A topography-based review of the current state of the art. NeuroImage Clin. 15, 659–672
St-Onge, E. et al (2023): A versatile, reliable and reproducible approach to white matter bundle segmentation. 1–11
Wasserthal, J. et el (2018):TractSeg - Fast and accurate white matter tract segmentation. Neuroimage 183, 239–253
Zhang, F. et al (2018): An anatomically curated fiber clustering white matter atlas for consistent white matter tract parcellation across the lifespan. Neuroimage 179, 429–447