"This is the first time that artificial intelligence has succeeded in imitating the clinical reasoning of a physician […] to make a diagnosis ". Professor Kang Zhang, from the University of California at San Diego, is formal: the algorithm he has developed with his team can detect diseases in children with the same efficiency as a doctor. The results of their research were published on Monday by the journal Nature Medicine.
– Nature Medicine (@NatureMedicine) February 12, 2019
"A very high level of relevance"
Based on artificial intelligence, this program was developed from data from over 1.3 million pediatric consultations at a Chinese medical center. The system is based on automatic natural language processing (NLP), that is, the machine can analyze spontaneous speech as opposed to predefined words and sentences.
She was able to synthesize all kinds of medical data: blood test results, X-rays, but also the description of the symptoms, the medical history of the patient or the written notes taken by the doctor.
According to the researchers, the algorithm has achieved a "very high level of relevance" for the diagnosis of common diseases such as influenza (94%), chickenpox (93%) or hand-foot-borne infectious disease (97%). ). But the tool has also been effective in recognizing life-threatening diseases such as
bacterial meningitis (93%).
AI "will not replace doctors"
"Artificial intelligence is going to play a growing role in the future. With enough training, I think she will be able to perform most of the duties of a doctor, believes Professor Kang Zhang. It will not replace doctors (…) but it can help them do their job better, faster and with lower costs. "
According to the press release of the specialized journal Nature Medicine, this type of algorithm could "help physicians analyze large volumes of data, strengthen the evaluation of diagnoses and provide decision support in case of diagnostic uncertainty".