Joseph T. Colonel
Machine learning researcher (PhD) at the Icahn School of Medicine at Mount Sinai whose JAMA Neurology study uses acoustic analysis of primary-care conversations to screen for cognitive impairment.
1
APPEARANCES
1
PODCASTS
Jun 15, 2026
15:27
If this occurs, it's we're on the normal scale, nothing to worry about for all intents and purposes.
Joseph T. ColonelGUEST
Unfortunately, the answers were not entirely sure, primarily because our best performing model was using acoustic features from a neural network.
Joseph T. ColonelGUEST
And so because it's part of this new regime of ways to summarize this audio, we don't know specifically what is driving the performance of that model.
Joseph T. ColonelGUEST
We did take a look at our best performing model of these expert-defined interpretable features.
Joseph T. ColonelGUEST
So those relate to how someone talks, how does the pitch of their voice change, how does the volume of their voice change, how quickly are they speaking.
19:45
If you were to pitch this to our listeners, anyone skeptical, what would you say about AI and aiding us in revolutionizing pre-screening for our patients, especially with difficult diagnoses?