Tract-specific lesion mapping with deep learning predicts balance deficits in degenerative cervical myelopathy

In 73 DCM patients, deep-learning lesion mapping showed T2 lesions peaked rostral to maximal compression and independently predicted balance deficits (dorsal column P=0.032, reticulospinal P=0.011), shifting focus from global stenosis to tract-specific vulnerability.

• 45% had T2-hyperintense lesions, defining a more disabled phenotype; lesion density peaked at C4, rostral to C5 compression, implicating venous congestion/ascending degeneration.
• Balance impairment independently predicted by dorsal column and lateral reticulospinal tract damage; dexterity not attributable to any single tract, implying reliance on global cord integrity.
• Residual tissue bridges conferred protective benefit for balance (P=0.044) only when reticulospinal tract integrity was accounted for, unlike in traumatic spinal cord injury.

Study limitations (e.g., single-center, external validation) not extractable from abstract; primary article not reviewed.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.