The competition between humans and machines in the field of medicine has recently captured even more attention, thanks to a striking case shared by Matteo Bassetti, the Director of Infectious Diseases at the San Martino Hospital in Genoa. In a post on Instagram, Bassetti shared the remarkable story of young Alex, a child who had been suffering from chronic pain for three years without receiving a diagnosis.
According to Bassetti, despite consulting 17 specialists, Alex’s mother remained desperate after no one could determine the cause of his pain. In a final attempt, she decided to input her son's symptoms and MRI data into ChatGPT. The AI suggested a possible diagnosis of tethered cord syndrome, a rare condition where the spinal cord becomes attached to surrounding tissues, leading to pain and neurological issues.
The mother took this suggestion to a neurosurgeon, who confirmed the diagnosis and performed surgery on Alex, ending his years of suffering. In this case, Bassetti points out, AI didn’t just outperform one doctor—it outperformed 17. He emphasizes the potential of artificial intelligence as a valuable tool that can complement medical expertise, and stresses the importance for the medical community to embrace these technologies.
This story highlights a broader trend revealed in recent studies, including one conducted by Stanford University and published in JAMA Network Open. The study involved 50 doctors and compared their diagnostic abilities with those of ChatGPT-4. The results were astonishing: ChatGPT-4 achieved an average diagnostic accuracy of 90% based on clinical case studies, while doctors who utilized AI support reached 76%, and those who didn’t used AI performed at 74%.
However, the study also pointed out some challenges in integrating AI into medical practice. Many healthcare professionals showed resistance to adopting AI-based diagnoses, particularly when the AI’s conclusions conflicted with their own. Moreover, most doctors were not fully aware of how to harness the full potential of AI tools.
The growing potential of AI in healthcare was also a key theme at the recent congress of the Italian Society of Cardiology (SIC). During the event, the first Italian Consensus Document on the use of AI in cardiology was presented. The document highlighted the immense promise of AI in monitoring high-risk cardiac patients. For instance, a large study published in Nature Medicine demonstrated that combining AI with ECG (electrocardiograms) could reduce three-month mortality rates by 31% by identifying patients most likely to experience fatal events.
Ciro Indolfi, past-president of SIC and Professor of Cardiology at the University of Cosenza, also noted the promising potential of AI in early diagnosis of heart conditions. An example from a study involving 362 patients showed that AI had a 99% accuracy rate in identifying serious cases of heart attack, with average evaluation times of just 37 seconds—about four times faster than human doctors. AI has also proven effective in analyzing Holter tests, telemonitoring patients with implanted defibrillators, and improving the accuracy of wearable devices used in early diagnosis.
Additionally, AI’s capabilities extend to the analysis of echocardiograms, MRIs, and CT scans, enhancing diagnosis of cardiomyopathies, valve dysfunctions, and coronary stenosis. AI has demonstrated an accuracy rate of over 98% in identifying thrombi and calcifications in angiographies.
While AI may never fully replace human doctors, its ability to enhance diagnostic skills and improve healthcare outcomes is undeniable. The future of medicine, it seems, lies in the collaboration between human expertise and artificial intelligence.
