Google's Medical AI System AMIE Tested in Real-Time Video Consultations

Google Research and Google DeepMind have announced the first-of-its-kind case study demonstrating that the medical research AI system AMIE can conduct real-time clinical video consultations.

Ability to Interpret Visual and Auditory Cues

In a doctor's examination, communication is not limited to words alone but also encompasses details such as the patient's cough, gait, or signs of discomfort.

Powered by Gemini and Project Astra technologies, AMIE was developed with a multi-agent architecture capable of successfully interpreting these visual and auditory cues.

Real-Time Diagnosis and Physical Examination Guidance

The advanced AI system is capable of guiding virtual physical examinations in real time and demonstrates diagnostic reasoning skills.

Studies show that the system can support capabilities in a clinical setting at an expert level.

Simulated Study and Evaluations

Randomly selected simulated consultations were conducted with a group consisting of standardized patients and primary care physicians.

Clinical evaluators found AMIE to be successful in core clinical competencies such as history-taking, diagnostic accuracy, management appropriateness, and communication quality.

Future-Oriented Research Phase

Standardized patients stated that they preferred the video experience over the plain-text chat experience.

AMIE is currently still a research system and requires further research before it can be used responsibly in the real world.