**How AI Can Spot Diseases That Doctors Aren’t Looking For**
Artificial Intelligence (AI) is revolutionizing medical diagnostics and treatments, particularly in the realm of opportunistic screening. It involves the systematic use of AI technology to examine medical scans, such as CT scans, to identify diseases that might otherwise go undetected until they become more problematic. This evolving integration of AI in healthcare is noteworthy as it highlights the potential to enhance early detection of diseases, leading to improved treatment outcomes.
A compelling case is that of 58-year-old Will Studholme, who visited an NHS hospital in Oxford suffering from gastrointestinal symptoms in 2023. While he initially expected a diagnosis related to his symptoms, he was surprised to learn during further examinations that he had osteoporosis, a condition generally associated with aging and characterized by weak bones susceptible to fractures. This diagnosis arose indirectly after Studholme underwent an abdominal CT scan as part of his evaluation, which AI technology subsequently analyzed. The AI identified a collapsed vertebra—a marker often indicative of osteoporosis. Further assessments confirmed his diagnosis, and he was placed on an annual infusion treatment to bolster his bone density. Will Studholme candidly expressed his feeling of luck, indicating that without AI, he doubted the condition would have been discovered in time for effective treatment.
Traditionally, radiologists may spot incidental findings on imaging, such as tumors or other abnormalities. However, the application of AI for opportunistic screening—an automated process—marks a new frontier in patient care. “This approach of screening is just beginning,” noted Perry Pickhardt, a professor at the University of Wisconsin-Madison and a developer of AI algorithms for such applications. The term “opportunistic” signifies that the technology utilizes imaging conducted for other medical reasons, thereby unleashing invaluable information that can identify treatable conditions before symptoms manifest.
Prof. Pickhardt further elucidated the significance of this development, saying, “We can avoid a lot of the lack of prevention that we have missed out on previously.” Indeed, many chronic diseases remain undiagnosed through regular check-ups and tests; thus, leveraging existing imaging data provides an innovative solution.
In discussions with Miriam Bredella, a radiologist at NYU Langone, she emphasized that CT scans contain a wealth of unexamined data regarding body tissues and organs that could yield life-saving information. Manually reviewing these scans to extract valuable insights would take substantial time; thus, AI eases that burden while also helping to mitigate biases. For instance, osteoporosis is often perceived as affecting primarily thin, older white women, leading healthcare providers to overlook other demographics. AI, by contrast, remains agnostic to such biases and ensures a broader scope of detection.
The potential of AI extends beyond osteoporosis, as it is also being utilized to detect conditions like heart disease, fatty liver disease, and diabetes. The algorithms are particularly focused on analyzing scans from various imaging modalities, including mammograms and chest x-rays, thereby broadening the horizons of opportunistic imaging. It is essential that the training datasets for these algorithms encompass diverse ethnic backgrounds to ensure fair deployment across various patient populations.
The AI technology that assessed Studholme’s scan, created by the Israeli firm Nanox.AI, is one of few dedicated to opportunistic screening. Oxford NHS hospitals began trialling its osteoporosis-focused algorithm in 2018, leading to a significant uptick in the identification of patients with vertebral fractures, which could then be investigated for osteoporosis.
Despite the promise of AI-driven diagnostics, experts like Sebastien Ourselin from Kings College London caution against potential pitfalls, notably the increased demand on healthcare systems due to patients flagged for further testing and subsequent management. He underscored the necessity of having robust services in place to handle this surge in patient numbers.
In conclusion, artificial intelligence is transforming the landscape of medical diagnostics by surfacing diseases that healthcare providers may not prioritize during initial assessments. As demonstrated through Will Studholme’s experience, the capacity for AI to identify early-stage conditions presents a significant opportunity to prevent disease progression and facilitate targeted treatment regimens. As the technology advances and becomes more integrated into healthcare systems, it could very well redefine our approach to disease detection and management, ultimately improving patient outcomes while driving efficiencies in clinical practice.
