Related Experiment Video
Updated: Aug 5, 2026
Cheek Injection Model for Simultaneous Measurement of Pain and Itch-related Behaviors
Published on: September 27, 2019
Machine Learning-Derived Neural Signatures of Itch and Pain That Reliably Distinguish the Two Sensations in Humans: A
Hideki Mochizuki1, Elly Georgas1, Odelia Schwartz2
1Dr. Phillip Frost Department of Dermatology & Cutaneous Surgery, Miami Itch Center, Miller School of Medicine, University of Miami, Miami, Florida, USA.
Background:
Recent advances in human neuroimaging combined with machine learning have enabled identification of neural signatures representing various internal states, providing a promising framework for developing objective biomarkers. However, no study has investigated neural signatures that can reliably identify and distinguish itch and pain. Such neural signatures were explored in the present study using functional MRI (fMRI) and support vector machine (SVM).
Methods:
We measured brain activity in 33 healthy participants under cowhage-induced itch, mustard oil-induced pain, and control conditions using fMRI. We made seed-based functional connectivity images (R-images), where seed brain regions were the posterior cingulate cortex (PCC) and bilateral anterior insular cortex (aIC). We conducted a cross-validated and bootstrapped SVM using R-images to identify key brain regions with weights that were important to identify and distinguish itch and pain (threshold to identify these regions: p < 0.05). These neural signatures of itch and pain were applied to test sets of R-images to examine classification performance (Itch vs. Control or Pain).
Results:
These signatures showed excellent classification capability, in particular when combining multiple signatures (area under the curve of receiver operating characteristic curve: > 0.9, accuracy: > 90%).
Conclusions:
This is the first neuroimaging study to explore neural signatures that can reliably detect and distinguish itch and pain using machine learning. Our approach using seed-based functional connectivity images combined with cross-validated and bootstrapped SVM demonstrated high classification performance. The present study serves as a proof-of-concept demonstrating the feasibility of this approach to develop brain-based biomarkers for assessing itch and pain.
Significance Statement:
This is the first study to identify neural signatures of itch and pain. These signatures reveal distinct brain network patterns representing itch and pain, enabling reliable detection and differentiation of these two sensations based on brain activity. These signatures hold strong potential for the development of objective assessments of itch and pain.
