Researchers in Japan say they have developed an artificial intelligence model that can help detect acromegaly by analyzing photos of the back of the hand and a clenched fist—an approach designed to avoid using facial images or fingerprints. The team reports the system performed well in testing and could help clinicians identify potential cases earlier and refer patients for specialist evaluation.
Acromegaly is a rare endocrine disorder, usually developing in adulthood, in which the body produces too much growth hormone—most often due to a pituitary tumor. The condition can cause progressive enlargement of the hands and feet, changes in facial appearance, and broader effects on bone and organ growth.
Because symptoms often develop slowly, diagnosis can be delayed for years. “Because the condition progresses so slowly, and because it is a rare disease, it is not uncommon to take up to a decade for it to be diagnosed,” said Kobe University endocrinologist Hidenori Fukuoka.
While some research has explored using artificial intelligence to identify acromegaly from photographs, much of it has relied on facial images, which can raise privacy concerns. Seeking a less identifying alternative, the Kobe University-led group trained a deep-learning model using two types of hand photographs: an extended hand showing the back of the hand, and a clenched fist with the thumb positioned externally. The researchers said they avoided palm images because palm patterns can be used as biometric identifiers.
According to the study, 725 participants were enrolled across 15 medical facilities in Japan, contributing more than 11,000 images used to train and validate the model. The work was published in The Journal of Clinical Endocrinology & Metabolism under the title “Automatic Acromegaly Detection Using Deep Learning on Hand Images: A Multicenter Observational Study.”
The authors reported that the system achieved “very high sensitivity and specificity,” and that it outperformed experienced endocrinologists who were asked to evaluate the same photographs. Graduate student Yuka Ohmachi said she was surprised the model performed so well using only these hand views: “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 researchers emphasized that photographs alone are not sufficient for diagnosis, which typically requires clinical evaluation and hormone testing. Still, they said a screening-style tool could help prompt earlier follow-up, especially in settings where specialist access is limited. “We believe that, by further developing this technology, it could lead to creating a medical infrastructure during comprehensive health check-ups to connect suspected cases of hand-related disorders to specialists,” Fukuoka said.
The team said it plans to explore whether similar image-based methods could be adapted to help detect other conditions with visible signs in the hands, including rheumatoid arthritis, anemia and finger clubbing.
The research was funded by the Hyogo Foundation for Science Technology and involved collaborators from multiple institutions in Japan, including Fukuoka University and Nagoya University, the researchers said.
