A predictive corticospinal model for pain perception

Pain perception arises from integrated corticospinal circuits, yet most neuroimaging biomarkers are brain centric. We develop the Corticospinal Pain Intensity Pattern, a multivariate model trained and validated on 330 simultaneous corticospinal fMRI data. Across independent datasets, the model predicts pain intensity more accurately than cortical signatures and generalizes to electrical pain, while remaining insensitive to itch and observed pain. The model further tracks analgesia induced by transcutaneous electrical nerve stimulation in healthy participants. In a chronic pain cohort, corticospinal model expression derived from low-frequency spontaneous activity predicts baseline pain and longitudinal changes closely mirror treatment-induced pain relief. A corticospinal hidden Markov model reveals that dynamic transitions between pro- and anti-nociceptive states underpin static spectral abnormalities. Together, these findings establish a corticospinal biomarker that bridges experimental and clinical pain by linking task-evoked and spontaneous neural activity.