Comments & Discussion

Boston University moderates comments to facilitate an informed, substantive, civil conversation. Abusive, profane, self-promotional, misleading, incoherent or off-topic comments will be rejected. Moderators are staffed during regular business hours (EST) and can only accept comments written in English. Statistics or facts must include a citation or a link to the citation.

There are 2 comments on Could a Computer Diagnose Alzheimer’s Disease and Dementia?

  1. This online test is not accurate. I entered the exact same information twice, with the exact same photos for the clock drawing and follow the path diagram. The first time it says I have a 48% chance of getting dementia. The second time is says I have a 78% chance of getting dementia. How can an accurate test have such a big discrepancy for the exact same input?

    Looks like a lot more work is needed to make this online test credible. Artificial intelligence is just that, artificial. We need real human intelligence to provide a proper diagnosis.

    1. Thank you for taking the time to participate to the study. It is a research study, to help us collect more anonymized data to improve the models. It is not a diagnosis or production-level application and, as a result, all possible tests needed to ensure the integrity of the input have not been implemented. We are very confident that the same clock image (or even variations, e.g., small shifts, rotations, etc.) produces very similar outputs. After looking at recent submissions more carefully, what we believe happened is that the Trails test was submitted in the place of the clock image. The model was expecting a clock, and this led to very different outputs in response to slightly different inputs (different cropping and resolution).

Post a comment.

Your email address will not be published. Required fields are marked *