BACKGROUND: Detecting cortical demyelination using magnetic resonance imaging (MRI) in multiple sclerosis (MS) remains a challenge. Magnetization transfer ratio (MTR), T1-weighted/T2-weighted ratio (T1T2R), and T2-weighted (T2w) signal are sensitive to cortical demyelination, but their accuracy is unknown. OBJECTIVES: To quantify the sensitivity, specificity, and accuracy of postmortem T1T2R, MTR, and T2w in detecting cortical demyelination. METHODS: postmortem MRIs from 9 patients were used to measure T1T2R, MTR, and T2w along the midline of cortical gray matter and classified as normal or abnormal. MRIs were co-registered and compared to hemispheric myelin staining. The sensitivity, specificity, and accuracy of T1T2R, MTR, and T2w in detecting cortical demyelination were measured. RESULTS: The mean age (standard deviation) at death was 64.7 (+/-13.7) years with a disease duration of 23.8 (+/-10.5) years. The sensitivity was 78% for MTR, 75% for T1T2R, and 63% for T2w. The specificity was 46% (T2w), 13% (T1T2R), and 29% (MTR). The accuracy was 71% (T2w), 39% (MTR), and 42% (T1T2R). There were no significant differences between different MRI measures in cortical demyelination or intracortical/subpial lesion detection. CONCLUSIONS: Although somewhat sensitive, the modest specificity of conventional MRI modalities for cortical demyelination indicates that they are influenced by cortical changes other than demyelination. Improved acquisition and post-processing are needed to reliably measure cortical lesion load.