AI Detects Acromegaly From Hand Images | Diagnosis

A decade-long diagnostic odyssey for acromegaly, a rare hormonal disorder, may soon be dramatically shortened thanks to a new AI model developed by Kobe University researchers. This isn’t just a marginal improvement in detection; the AI *outperforms* experienced endocrinologists, offering a potential paradigm shift in how we screen for and manage this debilitating condition. The breakthrough is particularly significant given the disease’s insidious progression and the potential for a 10-year reduction in life expectancy if left untreated.

Acromegaly arises from the overproduction of growth hormone, typically caused by a benign tumor on the pituitary gland. Its slow onset and often subtle initial symptoms – enlarged hands and feet, altered facial features – mean it frequently goes unrecognized for years. The current diagnostic process relies heavily on clinical observation, hormone level testing, and imaging scans, all of which require specialist expertise and can be time-consuming. The increasing prevalence of endocrine disorders, coupled with a global shortage of endocrinologists, has created a pressing need for more efficient diagnostic tools. Previous attempts to leverage AI for early detection have stumbled due to reliance on facial photographs, raising legitimate privacy concerns.

The Kobe University team cleverly circumvented this issue by focusing on the hands – a body part routinely examined in clinical practice for acromegaly. Crucially, they limited the image data to the back of the hand and clenched fist, avoiding the unique patterns of palm lines that could compromise individual privacy. This approach not only addressed ethical considerations but also enabled them to amass a substantial dataset of over 11,000 images from 725 patients across 15 Japanese medical facilities – a critical mass for robust AI training and validation. The resulting model, detailed in the Journal of Clinical Endocrinology & Metabolism, demonstrates remarkable accuracy, exceeding that of human specialists.

The Forward Look

This research isn’t an isolated success; it’s a harbinger of a broader trend. We can expect to see a surge in the development of “peripheral” AI diagnostics – tools that analyze readily available data (like hand images) to flag potential health issues. The Kobe University team is already planning to extend their model to detect other conditions, including rheumatoid arthritis, anemia, and finger clubbing. More importantly, this technology has the potential to address healthcare disparities. By integrating this AI into comprehensive health check-ups and providing support to non-specialist physicians in regional areas, access to timely diagnosis and treatment can be significantly improved. The next 18-24 months will be critical as the team seeks to validate the model in diverse populations and explore integration with existing electronic health record systems. Look for pilot programs in Japanese regional hospitals to begin in late 2026, with potential for wider adoption – and adaptation for other conditions – in the years that follow.

Reference: Ohmachi Y, Nishio M, Abe I, et al. Automatic acromegaly detection using deep learning on hand images: a multicenter observational study. J Clin Endocrinol Metab. Published online February 27, 2026:dgag027. doi: 10.1210/clinem/dgag027

This article has been republished from the following materials. Note: material may have been edited for length and content. For further information, please contact the cited source. Our press release publishing policy can be accessed here.