npj Dementia paper reports a Kernel Flow brain-imaging marker of mild cognitive impairment

Kernel/Kernel Flow

In 50 people with mild cognitive impairment and 51 controls, machine-learning models on Kernel Flow brain recordings during cognitive tasks separated the groups with an AUC of 0.92, better than self-report and behaviour alone (0.79). This suggests short functional neuroimaging scans could help screen for mild cognitive impairment.

• Machine-learning models on Kernel Flow brain recordings achieved AUC 0.92 distinguishing MCI from controls.
• Better performance than self-report and behaviour screening alone (AUC 0.79).
• Suggests rapid functional neuroimaging could become practical screening tool for mild cognitive impairment.