AI-Enhanced CT Scans May Detect Diabetes During Routine Exams

A new study suggests an opportunistic screening method could provide early diagnosis of diabetes and related health issues.

VesnaArt/Shutterstock

With more than 1.2 million new Type 2 diabetes cases diagnosed annually in the United States, early detection is crucial. New research on Tuesday suggests that artificial intelligence (AI)-enhanced computed tomography (CT) scans could revolutionize diabetes screening by utilizing incidental data from routine scans taken for other medical reasons. Imagine getting screened for diabetes during a CT scan intended for an entirely different health issue.

This opportunistic screening method could provide early diagnosis of diabetes and related health issues, potentially transforming routine health check-ups. However, experts caution that implementing this technology could present significant challenges and downsides.

About the Study

The study aimed to determine if existing CT scans could be used to gather valuable health information without requiring additional tests. Researchers focused on markers such as visceral and subcutaneous fat, muscle mass, and liver fat content.

The study notes that even a single CT image from the lower back can reveal critical details about these markers, potentially predicting events like cardiovascular disease and overall mortality.

“Given the significant burden of diabetes and its complications, we aimed to explore whether automated and precise imaging analyses could enhance early detection and risk stratification beyond conventional methods,” said Dr. Seungho Ryu, the study’s senior author, in a press release.

The study examined how well these CT-derived markers could predict Type 2 diabetes and other related health issues. Researchers analyzed data from more than 32,000 South Korean adults who had previously undergone health screenings, including positron emission tomography (PET)/CT scans.

At the start of the study, 6 percent of participants had diabetes, and 9 percent developed it during the follow-up period. The automated CT markers were effective at predicting new diabetes cases, with accuracy scores of 0.68 for men and 0.82 for women (where 1.0 indicates perfect accuracy).

The amount of visceral fat around the abdominal organs was the best predictor of diabetes. When combined with measures of muscle size, liver fat levels, and calcium buildup in the arteries, the predictions became even more accurate.

“By integrating these advanced imaging techniques into opportunistic health screenings, clinicians can identify individuals at high risk for diabetes and its complications more accurately and earlier than the current approach,” Ryu said. “This could lead to more personalized and timely interventions, ultimately improving patient outcomes.”

Advancements in Artificial Intelligence

Advancements in machine learning and AI have the potential to revolutionize the analysis of body composition from medical images, making the process faster and less dependent on manual effort. However, these technologies have yet to be fully integrated into everyday clinical practice.

Radiologist Dr. Perry J. Pickhardt, who wrote an accompanying editorial, told The Epoch Times that AI-opportunistic CT screening leverages incidental CT findings unrelated to the initial reason for the scan. This technology, he said, could identify unsuspected diseases such as cardiovascular issues, osteoporosis, sarcopenia, diabetes, and metabolic syndrome.

This is not the first time AI has been used to detect diabetes. A 2023 study in Nature Communications found that routine chest X-rays analyzed using an AI model can provide early warning signs for diabetes, even in patients who don’t meet typical risk guidelines.

Developed by a multi-institutional team, the AI model flagged elevated diabetes risk years before diagnosis by analyzing more than 270,000 x-rays and electronic health records, focusing on fatty tissue location. Validated on nearly 10,000 additional patients, the authors explained that this approach offers a cost-effective method for early detection.

“We need to capitalize on this ‘free’ and valuable info for the sake of our patients,” Dr. Pickhardt said.

Challenges and Downsides of Opportunistic Screening

In his editorial, Pickhardt explained that while AI-driven opportunistic CT screening shows promise, its implementation in clinical practice faces several challenges.

One significant obstacle is securing reimbursement that reflects the actual value of these screenings. Without proper compensation, integrating these advanced AI tools into routine practice may be financially unsustainable, he said. Additionally, there is a risk of backlash and misunderstanding from primary care providers if the rollout is not managed correctly.

Opportunistic screening will likely identify unsuspected findings or patients at high risk for disease. Appropriate referral networks must be established to help patients and providers manage these findings. Pickhardt noted that the increased workload for radiologists could complicate this process.

“Important clinical questions include whether relevant findings will be acted upon and whether interventions lead to better clinical outcomes,” Pickhardt wrote.

Is Imaging the Best Choice?

Dr. Richard Semelka, a seasoned radiologist, offered his view to The Epoch Times on using CT scans to identify diabetes risk. While recognizing the potential benefits of AI-driven opportunistic screening, he raises concerns about radiation exposure risks.

“I am very concerned about the radiation risk with all x-ray-based imaging, including CT and mammography,” Semelka said. He pointed out that the radiation dose from CT scans can increase the risk of malignancy, a serious consideration for screening purposes.

Semelka questions whether the benefits of CT imaging outweigh the risks and costs, primarily since methods like MRI can provide similar or better results without the cancer risk.

“How much better is this than using height, weight, BMI, and abdominal circumference to determine T2DM [Type 2 diabetes mellitus] risk?” he asked, suggesting that conventional methods may be more practical for routine screenings.

Despite his concerns, Semelka acknowledges that add-on or opportunistic scans without additional radiation can be beneficial. “An add-on or opportunistic scan that does not add any additional radiation is always a good idea,” he said. However, he criticized the use of PET-CT for general screening due to the high radiation dose.

While CT imaging offers advanced insights, particularly in measuring visceral and liver fat, Semelka advocates for cautious use. He believes that imaging should not be the sole method for screening unless it is part of a comprehensive approach that includes genetic testing for those with a high predisposition to certain diseases.

“The risk of the imaging procedure should not greatly exceed the risk of the potential diseases being looked for,” he said, emphasizing the need to balance the benefits and risks of using imaging technology in medical practice.