A new artificial intelligence model can detect acromegaly, a rare hormonal disorder, with greater accuracy than experienced endocrinologists, according to research published this week by Kobe University. The model analyzes images of the back of the hand and clenched fist, offering a potential pathway for earlier diagnosis and reduced healthcare disparities.
Acromegaly, caused by excess growth hormone after the growth plates have closed, typically manifests in middle age with enlargement of the hands, feet, and facial features. Left untreated, the condition can lead to serious complications including type 2 diabetes, heart problems, and a reduced life expectancy of approximately ten years, according to medical definitions of the disease.
“Due to the fact that the condition progresses so slowly, and because This proves a rare disease, it is not uncommon to take up to a decade for it to be diagnosed,” explains Kobe University endocrinologist Fukuoka Hidenori. The Kobe University team’s research addresses this diagnostic delay by leveraging deep learning technology.
Researchers focused on hand images to mitigate privacy concerns associated with facial recognition technology. Kobe University graduate student Ohmachi Yuka explained the team’s rationale: “Trying to address this concern, we decided to focus on the hands, a body part we routinely examine alongside the face in clinical practice for diagnostic purposes, particularly because acromegaly often manifests changes in the hands.” They further refined their approach by using images of only the back of the hand and clenched fist, avoiding the unique patterns of palm lines.
The team trained and validated their AI model using over 11,000 images donated by 725 patients across 15 medical facilities in Japan. Published in the Journal of Clinical Endocrinology & Metabolism on February 27, 2026, the study demonstrates the model’s high sensitivity and specificity in identifying acromegaly. In direct comparison, the AI outperformed human endocrinologists evaluating the same images.
“Frankly, I was surprised that the diagnostic accuracy reached such a high level using only photographs of the back of the hand and the clenched fist,” said Ohmachi. “What struck me as particularly significant was achieving this level of performance without facial features, which makes this approach a great deal more practical for disease screening.”
The Kobe University team is now exploring the application of their model to detect other conditions visible through hand imagery, including rheumatoid arthritis, anemia, and finger clubbing. They envision the technology as a complement to clinical expertise, reducing diagnostic errors and facilitating earlier intervention. Fukuoka stated the team’s goal is to integrate the technology into comprehensive health check-ups, connecting potential cases to specialists and supporting physicians in regional healthcare settings.
