AI flags aspiration risk from a two-second ‘Ah’ in a research model

A sip can take a wrong turn, sending food, liquid or saliva toward the airway instead of the esophagus. At Seoul National University Bundang Hospital, researchers have developed an AI-based test that asks people to say “Ah” for two seconds after eating to identify those at higher risk of aspiration.

The team, led by rehabilitation medicine professor Ryu Ju-seok, is listening for changes in vocal-cord vibrations. Those changes can appear when food remains around the vocal cords or when the larynx—the structure that helps protect the airway during swallowing—has weakened. Older people and people with neurological disorders are particularly prone to the problem, according to the researchers.

The approach is designed as an alternative to earlier screening methods based on X-ray examinations of swallowing. The researchers collected voice samples from October 2021 to February 2023, then used them to develop and refine the model before testing how well it could identify aspiration risk.

The hospital reports 81.78 percent sensitivity—the share of people at risk that the model identified—and 80.02 percent specificity, meaning the share of people without the risk that it correctly recognized. Repeated aspiration can leave food particles or pathogens in the lungs and potentially contribute to complications such as pneumonia, although coughing clears the material in many cases.

So what changes in practice? If later studies confirm the results, a short voice sample after eating could make aspiration-risk monitoring easier to repeat in everyday life, especially for vulnerable people. It is not yet a clinical tool: Ryu’s team says it still needs validation with data from multiple institutions before the model can be developed for clinical use.