Early detection of Alzheimer’s disease (AD) remains challenging due to limited biomarkers. This retrospective longitudinal study evaluated whether alterations in the deep medullary veins (DMVs) predict cognitive decline. We analyzed 170 participants from the Open Access Series of Imaging Studies-3 (OASIS-3) database [95 cognitive normal (CN), 36 mild cognitive impairment (MCI), and 39 AD]. DMVs were scored on susceptibility-weighted imaging (SWI). Cox proportional hazard models assessed the association between DMVs and longitudinal cognitive decline. Mediation analysis tested whether amyloid-β (Aβ) partially mediates the effect of DMVs on cognition. DMVs scores increased across groups {7 [6, 8] in CN, 13 [8.25, 15] in MCI, and 14 [13, 15] in AD; P<0.001}. DMVs score correlated with age (r=0.371, P<0.001), Mini-Mental State Examination (MMSE) (r=−0.494, P<0.001), cortical volume (r=−0.371, P<0.001), and Aβ burden (r=0.498, P<0.001). A higher DMVs score was associated with an increased risk of cognitive progression in AD [hazard ratio (HR) =1.13, 95% confidence interval (CI): 1.02–1.24, P=0.014], demonstrating good classification performance [area under the curve (AUC) =0.800, P<0.001]. Mediation analysis showed Aβ burden partially mediated the DMVs effect on MMSE (indirect effect β=−0.154, 95% CI: −0.337 to −0.017). DMVs alterations increase with AD severity, correlate with Aβ deposition, and predict cognitive decline, suggesting DMVs score as a promising imaging biomarker for early identification of individuals at risk of AD progression.