Print Close

ML Detects Brain Volume Normalization Following Surgical Pain Relief for Trigeminal Neuralgia

Poster No:

1144

Submission Type:

Abstract Submission

Authors:

Jerry Li¹^,2, Jessica Sun³^,4, Timur Latypov⁵, Daniel Jörgens⁶, Patcharaporn Srisaikaew⁷, Min Wu⁸, Mojgan Hodaie⁹

Institutions:

¹Institute of Medical Science, University of Toronto, Toronto, Ontario, Canada, ²Division of Brain, Imaging & Behaviour, Krembil Brain Institute, University Health Network, Toronto, Ontario, Canada, ³Department of Psychology, Faculty of Arts & Science, University of Toronto, Toronto, Ontario, ⁴Division of Brain, Imaging & Behavior, Krembil Brain Institute, University Health Network, Toronto, Ontario, Canada, ⁵Hospital for Sick Children, Toronto, Ontario, ⁶Division of Brain, Imaging & Behaviour, Krembil Brain Institute, University Health Network, Toronto, Ontario, ⁷University Health Network, Toronto, Ontario, ⁸Department of Neurosurgery, 1st Affiliated Hospital of USTC, Division of Life Sciences and Medicine, Hefei, Anhui, ⁹University Health Network, Toronto, ON

First Author:

Jerry Li
Institute of Medical Science, University of Toronto|Division of Brain, Imaging & Behaviour, Krembil Brain Institute, University Health Network
Toronto, Ontario, Canada|Toronto, Ontario, Canada

Co-Author(s):

Jessica Sun
Department of Psychology, Faculty of Arts & Science, University of Toronto|Division of Brain, Imaging & Behavior, Krembil Brain Institute, University Health Network
Toronto, Ontario|Toronto, Ontario, Canada

Timur Latypov, MD, PhD
Hospital for Sick Children
Toronto, Ontario

Daniel Jörgens, PhD
Division of Brain, Imaging & Behaviour, Krembil Brain Institute, University Health Network
Toronto, Ontario

Patcharaporn Srisaikaew, PhD
University Health Network
Toronto, Ontario

Min Wu, MD
Department of Neurosurgery, 1st Affiliated Hospital of USTC, Division of Life Sciences and Medicine
Hefei, Anhui

Mojgan Hodaie
University Health Network
Toronto, ON

Introduction:

Trigeminal neuralgia (TN) is a neuropathic pain condition that causes debilitating facial pain. Conventional magnetic resonance imaging (MRI) of the brain is unable to identify biomarkers that predict surgical response in patients, given the absence of MR features that distinguish pain from non-pain states (Fig. 1A). Nonetheless, the search and definition of these biomarkers is important and clinically relevant to find more objective assessments of pain and associated treatment protocols. Machine learning (ML) can be a pivotal tool for biomarker identification, due to its efficiency in analyzing large, heterogenous clinical and imaging datasets. We have previously used ML to describe brain signatures associated with TN (Latypov et al., 2024). In this longitudinal study, we use ML to explore primarily subcortical structures, and how they may change in response to surgical pain relief. We hypothesized that subcortical volumes could distinguish TN brains from healthy brains, and then normalize following surgical pain relief, thereby adopting a profile similar to healthy brains (Fig. 1B).

Methods:

MRI scans of 117 local TN patients who underwent pain-relieving surgery and 1068 healthy subjects from 4 external, geographically distinct databases were collected. TN scans were collected within 4 months pre- and 6-10 months post-surgery. All patients had at least 75% pain reduction following microvascular decompression (n = 60) or Gamma Knife radiosurgery (57) and were considered responders to surgery. Automated reconstruction of the brain and subcortical and hippocampal segmentation were performed with FreeSurfer 7.0 to extract subcortical volumes. All volumes were checked for quality and normalized using total intracranial volume before analyses. An ML pipeline optimized 1000 unique support vector classifiers (SVCs) using nested cross-validation to differentiate between healthy brains and TN brains pre-surgery. The accuracy of these SVCs was plotted to visualize the distribution of performance. An ensemble of 10 randomly chosen SVCs at the median of accuracy was validated on 100 local healthy subjects and tested on pre- and post-surgery data, with two holdout sets of age- and sex-matched healthy controls (HC).

Results:

The ensemble classified 99/100 local HC correctly for validation. Before surgery, the ensemble correctly classified all 117 TN patients and 115/117 HC (99% accuracy, ROC-AUC = 0.99; Fig. 1C). 6-10 months after surgery, a significant number of patients (n = 28; χ2 = 15.4, q < 0.001) switched from being classified as TN before surgery to now being classified as HC (ROC-AUC = 0.98). Prediction probabilities showed a significant shift towards being classified as HC, in both these 28 patients (W = 0, q < 0.001; Fig. 2A) and the remaining TN patients that were not reclassified (W = 804, q < 0.001). Further investigation revealed that the left accumbens area, involved in reward, learning, and motivation, was the most impactful feature for ensemble predictions (Fig. 2B). Amongst the next 14 most impactful features were the putamen, thalamus, hippocampus, and amygdala; all major regions involved in cognition and executive functioning. Thus, after pain relief, a substantial proportion of surgical responders had brain volume motifs significantly similar to healthy brains. These results highlight normalization of subcortical volumes in TN after surgical pain relief, and the discriminating features between TN and healthy brain profiles (Fig. 2C).

Conclusions:

Our longitudinal study demonstrates that brain structure alone is sufficient for ML to distinguish between having TN and being pain-free. Notably, recovery from neuropathic pain may be detectable between 6-10 months after pain-relieving intervention. These findings provide a foundation for mechanistic studies of subcortical regions in TN, many of which are related to cognition. This study highlights the potential for ML to augment clinicians in biomarker identification and expedition of clinical timelines.

Modeling and Analysis Methods:

Classification and Predictive Modeling ¹

Segmentation and Parcellation

Neuroanatomy, Physiology, Metabolism and Neurotransmission:

Subcortical Structures

Novel Imaging Acquisition Methods:

Anatomical MRI

Perception, Attention and Motor Behavior:

Perception: Pain and Visceral ²

Keywords:

Computational Neuroscience

Machine Learning

Modeling

Neurological

NORMAL HUMAN

Pain

Peripheral Nerve

STRUCTURAL MRI

Sub-Cortical

Treatment

^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.

I accept

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:

I do not want to participate in the reproducibility challenge.

Please indicate below if your study was a "resting state" or "task-activation” study.

Other

Healthy subjects only or patients (note that patient studies may also involve healthy subjects):

Patients

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.

Yes

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.

Not applicable

Please indicate which methods were used in your research:

Structural MRI

Computational modeling

For human MRI, what field strength scanner do you use?

3.0T

Which processing packages did you use for your study?

Free Surfer

Provide references using APA citation style.

1. Latypov, T. H., Wolfensohn, A., Yakubov, R., Li, J., Srisaikaew, P., Jörgens, D., Jones, A., Colak, E., Mikulis, D., Rudzicz, F., Oh, J., & Hodaie, M. (2024). Signatures of chronic pain in multiple sclerosis: A machine learning approach to investigate trigeminal neuralgia. PAIN, 10.1097/j.pain.0000000000003497.

UNESCO Institute of Statistics and World Bank Waiver Form

I attest that I currently live, work, or study in a country on the UNESCO Institute of Statistics and World Bank List of Low and Middle Income Countries list provided.