Researchers at Kobe University made a computer program that can find a rare disease by looking at pictures of the back of a hand and a closed fist. The method does not need any face photos, so it keeps patients’ privacy safe while still being very accurate.
The disease is called acromegaly. It happens when the body makes too much growth hormone. People with acromegaly get larger hands and feet, and their bones and organs can grow too much. The changes happen slowly, so doctors often miss it for many years.
If the disease is not treated, it can cause serious health problems and shorten life by about ten years. "It can take up to ten years to get a diagnosis," says endocrinologist Hidenori Fukuoka. He adds that AI tools are now being tried to spot the disease earlier.
Privacy‑First AI Using Hand Images
Most AI studies use face pictures, but faces can raise privacy worries. The Kobe team decided to look at hands instead because doctors already check hands when they examine patients for acromegaly.
Graduate student Yuka Ohmachi explains that they only photographed the back of the hand and a clenched fist. They avoided palm pictures because palm lines are unique and could identify a person. This careful choice helped them collect many images: over 11,000 hand photos from 725 patients at 15 hospitals across Japan.
AI Beats Experienced Doctors
The researchers published their findings in the Journal of Clinical Endocrinology & Metabolism. Their AI showed very high sensitivity (catching the disease) and specificity (avoiding false alarms). When they compared the AI to seasoned endocrinologists, the computer program performed better.
Ohmachi says, "I was surprised that we could get such high accuracy using only hand photos, and without any facial features. This makes the tool very practical for screening."
Looking Ahead to More Hand‑Based Tests
The team hopes to teach the AI to recognize other conditions that change the hands, such as rheumatoid arthritis, anemia, and finger clubbing.
Helping Doctors and Patients
In real clinics, doctors use many clues—medical history, lab tests, and physical exams—to diagnose. The AI is meant to support doctors, not replace them. It can alert doctors to possible hand‑related problems early, especially in places where specialists are hard to find.
Fukuoka adds, "With more development, this tool could become part of routine health checks, guiding patients to the right specialist and reducing gaps in care for rural communities."
The project received support from the Hyogo Foundation for Science Technology and involved many Japanese universities and hospitals.