From Jorge D. Faccinetti – Chairman and cofounder – Could artificial intelligence and machine learning recognize the early, subtle changes in people’s faces associated with the hormonal disease acromegaly? New research spearheaded by Dr. Manel Puig-Domingo and outlined in a recent paper published in the journal PITUITARY raises the possibility that AI algorithms and machine learning could provide a game-changing approach to the early diagnosis of this complex disease, which now takes an average of 6 to 10 years to diagnose. The paper outlines an AI-driven facial recognition system, AcroFace, that could detect acromegaly by analyzing facial photographs.
Read the article and podcast in Spanish here
In a nutshell, the paper describes highly sophisticated systems that analyze visual features (what the face looks like—texture, appearance) and geometric features (the measurements and distances between facial landmarks like eyes, nose, and jaw). Without getting too technical, the program combines CNNs (Convolutional Neural Networks) for analyzing facial images and an SVM (Support Vector Machine) for processing geometric measurements to produce the results. Initial results are outlined in the paper, which you can read here. For those of you interested in the technological aspects, CNNs are a type of artificial neural network used for image recognition and processing. CNNs identify patterns in images to detect features, starting with simple ones like edges and progressively combining them to recognize more complex patterns. An SVM is a machine learning algorithm. Here’s a link to more information on SVM.
Download Dr. Puig Domingo’s paper here
While the study has some limitations—using a relatively small database (118 patients) and testing only Caucasian subjects, requiring validation across diverse populations —the system was able to identify facial changes up to 10 years before diagnosis. Clearly, it could potentially revolutionize early detection. There are other challenges, as the program still misses some cases and flags healthy people as having the disease.
On the “good news” side, the program correctly identified people with acromegaly with an accuracy of 93%. The researchers acknowledge this is promising but not yet ready for prime time. It has to be tested in thousands more people in the general population to prove that the system can detect undiagnosed cases and be validated across different ethnic groups. Currently, Dr. Puig-Domingo and his teams are conducting a study in the general population in collaboration with a large company based in Barcelona. The ACROFACE system has analyzed 4,000 employees as part of a screening program for acromegaly. Results are slated to be released in a few months.
These early findings present a system that is 93% accurate at detecting a rare disease that usually goes undiagnosed for a decade and could eventually help thousands of people get treatment years earlier—potentially through something as simple as a smartphone app. Notably, the paper notes that previous attempts achieved 81-86% accuracy, representing a significant improvement over earlier efforts. To read more, you can download the paper here.
If further research goes forward and the program is validated and deployed, it promises to save years of undetected disease. It will undoubtedly prevent the development of complications from related conditions, transform acromegaly from a late-diagnosed crippling disease to an early-detected, more manageable condition without the pesky comorbidities, and serve as a model for other rare diseases with facial features.
On Acromegaly Day, we can all imagine a future where anyone can check themselves with a selfie, doctors can screen routinely with a quick photo, we catch it in year 1 instead of year 10, and treat people before serious, irreversible damage takes hold.
In a similar effort, Dr. Michael Cusimano, Professor of Neurosurgery, Education and Public Health in the Division of Neurosurgery at St. Michael’s Hospital, University of Toronto, aims to develop and evaluate a system that can detect acromegaly and other pituitary diseases before they lead to significant physical changes—such as alterations to the face or hands—and before patients develop complications such as diabetes, high blood pressure, and sleep apnea. Patients participating in the study are asked to complete a detailed survey, upload images, and provide documentation of acromegaly tumors or other pituitary tumors. This link takes you to the study page, where you can learn more and register if you wish to participate. For more information, please reach out to Dr. Cusimano or Melissa Fazari, research assistant at acromegaly-research@smh.ca
More on Acromegaly
Acromegaly is often caused by a benign tumor (adenoma) on the pituitary gland, leading to increased growth hormone production. The excess growth hormone causes bones and tissues to enlarge, leading to abnormal growth, especially in the hands, feet, and facial features. The condition develops slowly, often over several years, and can be challenging to diagnose in its early stages. Reducing the time it takes to recognize the disease, presently 6 to 10 years, would mean significant improvements in quality of life and substantial cost reductions. According to NIH studies, acromegaly ranges between 2.8 and 13.7 cases per 100,000 people and is most often diagnosed in middle age.
About Dr. Manel Puig-Domingo
Dr. Manel Puig-Domingo is the Head of the Service of Endocrinology and Nutrition at Germans Trias i Pujol University Hospital and Professor of Endocrinology at the Autonomous University of Barcelona.
Read and listen to prior articles and podcasts with Dr. Puig on this subject: https://pituitaryworldnews.org/?s=Manel+Puig+Domingo+
Read more about the facial recognition effort in Canada: https://pituitaryworldnews.org/on-acromegaly-day-technology-promises-early-diagnosis/
© 2025, J D Faccinetti. All rights reserved.