Yun Zhong, from Aofeisi Temple
QbitAI | Public Account QbitAI
Can AI safely assess a patient’s medical needs before they visit? A study led by researchers from Google and BIDMC Medical Center, published today in The Lancet, provides an encouraging preliminary answer. This is the first time that Google’s research results have been published in the main issue of The Lancet.
In a real clinical study conducted at the outpatient primary care clinic of BIDMC, 98 patients used the Google research-level diagnostic AI chatbot AMIE before seeking emergency care. The supervising doctor monitored the entire interaction process in real time; according to predefined safety criteria, no conversation was interrupted during any round. According to clinical doctors’ feedback, in 75% of cases, the summaries generated by AI help them prepare for patient consultations in advance; in over half of the cases, the AI outputs changed the doctor’s diagnostic approach. The accuracy of the differential diagnosis results provided by AMIE with the doctor’s final diagnosis is 90%.
Although aimed at large-scale implementation, larger clinical trials are still needed to evaluate the performance of AI for patients. However, the results of this study indicate that AI has the potential to improve the relationship between doctors and patients and reduce the work stress of healthcare workers. As Google breaks through the limits of artificial intelligence in the fields of science and health, they have made meaningful progress.
The full research paper can be accessed at: https://goo.gle/4i7Fx9N