In a groundbreaking study from Stanford Medicine, researchers have demonstrated that artificial intelligence (AI) can significantly improve the accuracy of skin cancer diagnoses. This finding has implications for dermatologists, general practitioners, and less specialized medical personnel.
Published in npj Digital Medicine, the research highlights how intense learning-based AI algorithms assist healthcare practitioners by enhancing diagnostic precision. The study was led by Eleni Linos, MD, a professor of dermatology and epidemiology, along with Jiyeong Kim, Ph.D., and Isabelle Krakowski, MD, as the lead authors.
Deep learning involves training computers with vast amounts of data—hundreds of thousands of images of skin conditions. Over time, the computer learns to recognize patterns correlated with various skin diseases, including cancer. These AI systems do not replace the clinician but rather support them by offering diagnostic suggestions that the clinicians may review and decide upon.
The study’s findings are based on a review of 12 other studies involving more than 67,000 evaluations. It was found that without AI, healthcare workers had a sensitivity (the ability to identify those with the disease correctly) of about 75% and a specificity (the ability to identify those without the disease correctly) of about 81.5%. With AI, these figures improved to 81.1% sensitivity and 86.1% specificity.
Notably, the benefits of AI were more pronounced among medical students, nurse practitioners, and primary care doctors, who saw average improvements of around 13 points in sensitivity and 11 points in specificity. While dermatologists and dermatology residents already had high accuracy, their diagnostic effectiveness also saw a significant boost with AI assistance.
“This research underscores the value of AI in enhancing our diagnostic capabilities,” explained Linos. The technology not only helps improve the accuracy of diagnoses but also reduces diagnostic time, which can alleviate doctor fatigue and possibly enhance the quality of patient-doctor interactions.”
Under Linos’s leadership, the Stanford Center for Digital Health aims to explore further the integration of AI tools in healthcare, including understanding how doctors’ and patients’ perceptions of AI might affect its adoption and use.
Previous studies have indicated that a clinician’s confidence in their diagnosis, the AI’s confidence level, and the agreement between the clinician’s and the AI’s conclusions are crucial factors in whether the AI’s recommendations are incorporated into the final diagnosis. This suggests a complex interaction between humans and machines that impacts clinical decision-making.
Medical fields reliant on visual diagnostics, such as dermatology and radiology, stand to gain significantly from AI due to its ability to discern finer details than the human eye. However, the application of AI is not limited to image-based diagnoses; it shows promise in areas requiring nuanced symptom analysis and predictive modeling.
“As we look forward to broader applications of AI in medicine, it’s crucial to ensure that these technologies are implemented in ways that are beneficial to all patients and support the well-being of physicians,” Linos stated.
This study is part of an ongoing effort to harness technology to improve healthcare outcomes. Stanford Medicine’s initiative in promoting the intersection of technology and health continues to set benchmarks for the future of medical diagnostics and patient care.
Source: Medicalxpress and paper
Isabelle Krakowski et al, Human-AI interaction in skin cancer diagnosis: a systematic review and meta-analysis, npj Digital Medicine (2024). DOI: 10.1038/s41746-024-01031-w
Like this article? Keep up to date with AI news, apps, tools and get tips and tricks on how to improve with AI. Sign up to our Free AI Newsletter
Also, come check out our free AI training portal and community of business owners, entrepreneurs, executives and creators. Level up your business with AI ! New courses added weekly.
You can also follow us on X