Artificial Intelligence Enhances Detection of Hidden Belly Fat Using Bone Scans

Researchers at Edith Cowan University (ECU) have developed an innovative artificial intelligence (AI) algorithm capable of estimating visceral fat levels from routine bone density scans used to detect spine fractures. Visceral fat, the deep fat nestled around internal organs, is a significant risk factor for health problems such as heart disease, diabetes, and certain cancers. Traditionally, assessing visceral fat requires expensive imaging techniques like MRI or CT scans, which also expose patients to higher radiation levels, limiting their routine use.

ECU's new approach leverages lateral spine Dual-energy X-ray Absorptiometry (DXA) scans—commonly performed to diagnose osteoporosis and spine conditions—and re-purposes these images for opportunistic screening of visceral fat. The AI model analyzes the scans to accurately predict visceral fat content without the need for additional tests, providing a quick, cost-effective, and less invasive method for health risk assessment.

"Have you heard of hidden fat that cloaks your organs? That's visceral fat, a major contributor to health risks," explained Ph.D. student Arooba Maqsood. She emphasized that obesity significantly impacts global health, not only causing increased morbidity and mortality but also imposing enormous economic burdens. For instance, Australia's health-related costs related to obesity were estimated to reach A$39 billion in 2019, projected to soar to A$228 billion by 2060, representing about 3.5% of the nation's gross domestic product. Globally, nearly 3.7 million deaths annually are linked to obesity.

Current standard measures like body mass index (BMI), waist circumference, and waist-to-hip ratio are limited in their ability to distinguish between different types of body fat, leading to inconsistencies in obesity assessment and related health risk evaluations. While advanced imaging like MRI and CT can precisely measure visceral fat, their high costs and radiation exposure restrict widespread use.

The ECU team’s machine learning model is trained on thousands of existing scans and aims to incorporate more diverse datasets worldwide to improve accuracy. Dr. Syed Zulqarnain Gilani, the lead AI scientist at ECU, highlighted that expanding the dataset will help the model become more effective across different populations.

Maqsood plans to present her research at MICCAI 2025, the International Conference on Medical Image Computing and Computer Assisted Interventions, held in Korea from September 23–27. This advancement could revolutionize the early detection and management of obesity-related health risks by utilizing routine imaging for comprehensive health assessments.

Source: https://medicalxpress.com/news/2025-09-machine-hidden-fat-routine-bone.html