Precision increases in a project led by Germans Trias using AI to detect early facial changes in patients with acromegaly

• Acromegaly is a rare disease that causes facial and skeletal bone deformities, as well as heart complications, diabetes and cancer
• It is diagnosed around 10 years after its onset; its World Day is commemorated every 1 November

The international journal Pituitary has highlighted research led by Manel Puig, head of the Endocrine, Thyroid and Obesity Research Group at the Germans Trias i Pujol Research Institute (IGTP), which explores the potential of AI and machine learning algorithms to provide a revolutionary tool for the early diagnosis of acromegaly.

Acromegaly, which is marked worldwide every 1 November, is a rare disease caused by an excess of growth hormone secretion, and in over 99% of cases it is due to a usually benign pituitary tumour. It can affect anyone and is typically diagnosed from the age of 40, although cases can also appear in childhood, where if left undiagnosed, they can result in gigantism.

In addition to facial and skeletal deformities, which eventually become very noticeable, excess growth hormone can lead to serious alterations in other parts of the body: enlargement of the heart that can cause heart failure, a tendency to develop diabetes, sleep apnoea, and an increased risk of developing various tumours, particularly colon cancer.

The article published in the journal describes an AI-driven facial recognition system, called AcroFace, which could potentially detect acromegaly simply by analysing facial photographs. Specifically, it does so by analysing visual features (the appearance and texture of the face) and geometric features (measurements and distances between facial landmarks such as the eyes, nose and jaw).

In a preliminary trial of the project, tested on 118 people, the system correctly identified individuals with acromegaly with 93% accuracy, a success rate higher than that achieved in similar previous attempts, which did not exceed 86%. Although the researchers view these early results as promising, they stress that they must be confirmed once a pilot study in the general population is completed, one that will test thousands more individuals from different ethnic backgrounds. The research group led by Manel Puig is currently conducting this study, analysing 4,000 photographs from the general population, and expects to obtain initial results soon.

Acromegaly is a rare disease that often goes undiagnosed for a decade, and having a reliable early detection system could help people with the condition receive treatment years earlier, potentially through something as simple as a mobile app.

Thus, acromegaly could shift from being a late-diagnosed, disabling disease to "a condition that can be detected early and managed more easily, without troublesome comorbidities, and even serve as a model for other rare diseases with facial features", summarises Manel Puig. "We can all imagine a future where anyone could self-screen with a selfie, where doctors could perform routine checks with a quick photo, where the disease is detected ten years earlier and treated long before serious and irreversible damage occurs", he emphasises.