Back

The Autonomous Cognitive Examination: Machine-Learning Based Cognitive Examination

Howard, C. W.; Johnson, A.; Barotono, S.; Faust, K.; Peedicail, J.; Ng, M.

2024-09-06 neurology
10.1101/2024.09.05.24313114 medRxiv
Show abstract

Structured AbstractO_ST_ABSINTRODUCTIONC_ST_ABSThe rising prevalence of dementia necessitates a scalable solution to cognitive screening and diagnosis. Digital cognitive assessments offer a solution but lack the extensive validation of older paper-based tests. Creating a digital cognitive assessment which recreates a paper-based assessment could have the strengths of both tests. METHODSWe developed the Autonomous Cognitive Examination (ACoE), a fully remote and automated digital cognitive assessment which recreates the assessments of paper-based tests. We assessed its ability to reproduce entire cognitive screens in a comparison cohort (n = 35), and the ability to reproduce overall diagnoses with an additional validation cohort (n = 11). RESULTSThe ACoE reproduced overall cognitive assessments with excellent agreement (intraclass correlation coefficient = 0.89) and reproduced overall diagnoses with excellent fidelity (area under the curve = 0.96). DISCUSSIONThe ACoE may reliably reproduce the evaluations of the ACE-3, which may help in accessible evaluation of patient cognition. Assessment in larger population of patients with specific diseases will be necessary to determine usefulness.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.