Corticospinal tract damage on baseline CT predicts motor recovery and functional outcome in intracerebral haemorrhage

INTRODUCTION: Corticospinal tract (CST) integrity can predict motor outcome after stroke but requires specialist investigations not routinely performed after intracerebral haemorrhage (ICH). We investigated the feasibility of identifying the CST on routine clinical CT scans, and whether classification of CST overlap with haematoma is associated with motor recovery after ICH.

PATIENTS AND METHODS: An expert observer, blinded to outcome, manually segmented the CST at the posterior limb of the internal capsule (PLIC) and corona radiata (CR) on diagnostic CT scans from 98 randomly selected MISTIE-III trial participants and determined whether CST overlapped with the haematoma. Multivariable linear regression tested for associations between haematoma CST overlap and the motor component of the National Institutes of Health Stroke Scale (baseline & Day 180, rate of recovery), patient reported motor impairment (Stroke Impact Scale [SIS] domain 1) and activity limitation (SIS domains 6&7) at Day 180, and modified Rankin Scale (mRS) at day 180. Three further readers analysed the same scans and the interobserver variability was assessed.

RESULTS: Haematoma and CST overlap occurred exclusively in the CR in 6%, the PLIC in 14% and in both in 52% of patients. CR involvement alone was associated with activity limitation on Day 180. Involvement at the PLIC alone or both the PLIC and CR was independently associated with worse motor outcomes (except rate of recovery, where only involvement of both was associated). Although haematoma and CST overlap remained associated with outcome for other readers, the strength of the association decreased with less expertise, and interobserver kappa scores (κ = 0.47 for CR and κ = 0.45 for PLIC) demonstrated only moderate agreement.

DISCUSSION AND CONCLUSION: Haematoma and CST overlap at the level of the PLIC identified on routine CT scans is independently associated with poor motor outcomes, representing a novel prognostic factor. Given moderate interobserver agreement, a more reliable machine-learning classification would be desirable.