Google tests AMIE clinical video consultations with promising results
Google's video-based medical AI system AMIE matches primary care physicians in clinical evaluations, utilizing a multi-agent architecture to reduce conversational latency.
Stock photo for illustration only, not from the actual event
- Google tested its medical AI system AMIE in synchronous video consultations with trained actor patients.
- Independent clinical evaluators rated AMIE on par with primary care physicians across core measures.
- The system uses a multi-agent architecture separating dialogue, clinical reasoning, and perception.
- Patient actors preferred the synchronous video interface over text chat for health consultations.
Google's research medical AI system, AMIE, has undergone synchronous video consultations with professional patient actors, receiving clinical evaluator ratings on par with primary care physicians across several core measures. The study marks an expansion of Google's healthcare AI testing into more interactive diagnostic environments.
Fifteen trained actors portrayed conditions across cardiopulmonary, abdominal, HEENT, neurological or psychiatric, and musculoskeletal presentations. Google emphasized that studies involving real patients and their own health conditions must follow before any definitive conclusions about clinical deployment can be drawn.
Stock photo for illustration only, not from the actual event
AMIE sets itself apart by utilizing an asynchronous multi-agent architecture rather than assigning dialogue, clinical reasoning, and perception to a single model process. Google explained that a single agent cannot currently maintain natural conversational response times while simultaneously conducting detailed reasoning and processing audio-visual input continuously. The system delegates specific roles:
- Talker Agent: Manages spoken interaction with the patient to maintain natural conversational flow.
- Planner Agent: Runs in the background to update differential diagnoses, management plans, and clinical goals.
- Perception Agent: Review streams of video and audio continuously to catch non-verbal signs and physical findings.
The multi-agent approach addresses a fundamental engineering challenge in medical AI: the trade-off between depth of clinical reasoning and conversational speed. Long pauses during medical interviews can damage patient rapport. By decoupling the real-time dialogue interface from background analytical processes, Google's design solves a crucial latency bottleneck in telehealth applications.
Google structured the human evaluation as a multi-arm randomised study where AMIE conducted real-time video consultations alongside a text-only baseline and 10 board-certified primary care physicians using the exact same interface. An independent panel of 20 experienced primary care physicians subsequently reviewed every recorded consultation using established clinical rubrics.
"Evaluators rated the video system higher on eliciting physical signs and proactively guiding actors through virtual examination manoeuvres than either the PCP group or text-only AMIE."
Google
Source: AI News
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment