Per reporter: Cai Ding Per editor: Lan Suying
After suffering from a strange disease for three years, he sought help from 17 doctors without success, and was finally successfully diagnosed by ChatGPT.
This is not a science fiction plot, but a real experience that happened to a 4-year-old boy in the United States. During the pandemic-induced lockdown, Courtney bought a "bouncing house" for her two children, but it wasn't long before her 4-year-old son, Alex, started feeling pain.
Courtney spent three years and took Alex to see 17 doctors, from pediatrics, dentistry, orthopedics and other outpatient departments to various specialists, and none of the doctors really accurately diagnosed the cause of Alex. It wasn't until earlier this year that Courtney finally got the answer she had been searching for for three years from an unlikely source, ChatGPT.
With the rapid development of artificial intelligence (AI), it is not new for patients to seek help from AI doctors before seeking medical treatment. Although there are reports that generative AI such as ChatGPT has a similar level compared to human doctors, many doctors believe that AI-based medical tools should also go through an approval process similar to the US Food and Drug Administration's (FDA) drug approval system.
17 doctors diagnosed without success, and ChatGPT accurately screened out the cause
The story also starts with a day in 2020. Alex's nanny told Courtney that Alex had to take a painkiller called Motrin every day or he would lose his temper from the pain. Alex then developed teeth grinding, which his parents linked to the pain, thought it might be caused by tooth change or tooth decay.
Alex was then taken by his parents to see a dentist. However, since the dentist did not find any problems after the examination, it was suggested that they see an orthodontist who specializes in treating airway obstruction given that Alex was grinding his teeth. This was the beginning of a three-year journey to medical treatment.
4-year-old Alex (Image: today.com)
The orthodontist noticed that Alex's upper jaw was too small, making it difficult to breathe, so he placed an expander on Alex. The treatment worked, and her mother thought Alex's illness was about to be cured.
However, Courtney soon discovered that Alex suddenly did not grow tall, so he went to the pediatrician, who thought that Alex may be infected with the effects of the new coronavirus, but Courtney was not satisfied with this explanation. At the beginning of 2021, Courtney took Alex for a review, and the pediatrician said at the time that Alex had "grown a little taller", but found that Alex's left and right feet were somewhat unbalanced and recommended that they choose physical therapy.
But before the physical therapy began, Alex had severe headaches again, and they were getting worse. Courtney took Alex to the otolaryngologist to see if he had sleep problems with his sinus or respiratory tract.
Courtney believes that no matter how many doctors they visit, experts will only deal with their respective areas of expertise. "No one wants to solve a bigger problem, and no one is even going to tell us what the outcome of the diagnosis is."
After all these twists and turns, Alex began physical therapy, and his physical therapist thought Alex might have a condition called Chiari malformation. According to the American Association of Neurosurgeons (AANS), it is a congenital condition that causes abnormalities in the brain at the interface between the skull and the spine.
After getting the results, Courtney began researching it and visiting more doctors: a neonatologist, a pediatric physician, an adult physician and an orthopedic surgeon, but the results were still very disappointing – no doctor gave her a definitive diagnosis.
During the three-year journey to the doctor, Courtney took Alex to 17 different doctors, but there was still no diagnosis that would explain Alex's symptoms.
The turning point happened earlier this year.
Exhausted and frustrated, Courtney signed up for a ChatGPT account and entered all the information about Alex's symptoms from three years of medical treatment, hoping to find a diagnosis from the artificial intelligence.
"I look at everything in Alex's MRI record line by line and enter it into ChatGPT." Courtney said. Eventually, Courtney got the answer to "tethered cord syndrome" and joined a Facebook parent communication group.
Finally, with ChatGPT's diagnosis, Courtney visited a new neurosurgeon who glanced at the MRI and gave the same conclusions as ChatGPT and pointed out the exact location of the tether.
"We saw a lot of doctors and even went to the emergency room at one point, and I've been trying. I sometimes spend the night in front of my computer, skimming through all the useful information. Therefore, when the advice given by ChatGPT is diagnosed with tethered cord syndrome, it makes perfect sense. Courtney said in an interview with TODAY.com.
AI medical tools should go through a similar drug approval process
The "Daily Economic News" reporter noticed that in fact, with the development of artificial intelligence, it is not a new thing for patients to turn to AI doctors.
According to National Public Radio (NPR), Riley Lyons, a fourth-year ophthalmology resident at Emory University School of Medicine, said he often finds patients seeking help from AI doctors like Google Doctor before coming to his door. Lyons and two ophthalmology colleagues recently evaluated the accuracy of the AI chatbot ChatGPT in diagnosing eye diseases. In June, they published a report in medRxiv, an online publisher of health sciences preprints, that ChatGPT performed quite well compared to human doctors who examined the same symptoms, and much better than the symptom checker on the popular health website WebMD.
Moreover, while ChatGPT occasionally gives outright wrong answers, the Emory University study reports that the latest version of ChatGPT makes no "seriously inaccurate" statements when faced with a standard set of eye symptoms.
Not only ChatGPT, but the medical big model Med-PaLM developed by Google and DeepMind researchers is already on par with real-world human clinicians — only 5.9 percent of Med-PaLM's answers were rated as likely to lead to "harmful" outcomes, similar to those generated by clinicians (5.7 percent).
Zheshang Securities pointed out in the research report that the application of large models such as ChatGPT in medical terminals has included five aspects: generating personalized medical reports, realizing online medical consultation, realizing automatic medical guidance, reviewing medical records and providing personalized medical advice.
Image source: Zheshang Securities
NPR reports that the accuracy of medical information provided by generative AI tools may indeed be improved compared to a simple Google search, but there are still many questions about how to integrate this new technology into the health care system and provide the same safeguards that have historically been used to introduce new drugs or medical devices.
Today, in healthcare, companies are also deploying generative AI in areas such as radiology and medical records.
However, for many healthcare professionals, chatbots come with a range of troubles, including issues related to privacy, security, bias, accountability, transparency, and the current lack of regulation. They argue that AI-based medical tools should go through an approval process similar to the U.S. Food and Drug Administration's (FDA) drug approval system, but it will be years away. Unfortunately, it's unclear how this mechanism would work for general-purpose AI like ChatGPT.
Daily economic news