AI reliably diagnoses Marfan syndrome from a simple facial photograph.

AI Accurately Diagnoses a Genetic Condition From Facial Photographs

August 12, 2024

by Sean McCabe

A Yale School of Medicine team reports in a new study, published in the journal Heliyon, that an artificial intelligence (AI) model was able to reliably diagnose people living with Marfan syndrome from a simple facial photograph.

Marfan syndrome is a genetic disorder, affecting about 1 in 3,000 people, which impacts the body's connective tissues. "Patients living with Marfan Syndrome are usually very tall and thin," said John Elefteriades, MD, professor of surgery at Yale School of Medicine and senior author of the study. "They have long faces and are prone to spine and joint issues. However, many are not diagnosed."

Marfan syndrome increases the risk for aortic dissection, where the aorta splits suddenly after becoming enlarged. "It is often lethal, and when the patient survives, urgent surgery is needed," Elefteriades said. "Being able to identify individuals from a photograph with AI will enhance diagnosis and enable protective therapies."

In their pilot study, researchers assembled 672 facial photographs of people with and without Marfan syndrome. A Convolutional Neural Network was trained on 80% of the photographs, then asked to identify the other 20% as Marfan or non-Marfan faces. The model successfully distinguished between Marfan and non-Marfan faces with 98.5% accuracy.

Researchers say they plan to make the tool available online in the future. "We are planning to extend this work beyond this initial pilot project," said Elefteriades. "We anticipate that many individuals may self-test once we put the test online."

"Yale School of Medicine faculty and students are leading the way in developing novel applications of AI to recognize and diagnose diseases, including rare diseases, earlier when we can have the greatest impact," said Nancy J. Brown, MD, dean of Yale School of Medicine.

Sandip Mukherjee, Mohammad A. Zafar, and Bulat Ziganshin were also authors on the study. Danny Saksenberg of Emerge, who also has an appointment at Yale, conducted the AI analysis.

https://medicine.yale.edu/news-article/ai-accurately-diagnoses-a-genetic-condition-from-facial-photographs/#:

Saksenberg D, Mukherjee S, Zafar MA, Ziganshin B, Elefteriades JA. Pilot study exploring artificial intelligence for facial-image-based diagnosis of Marfan syndrome. Heliyon. 2024 Jun 28;10(13):e33858. doi: 10.1016/j.heliyon.2024.e33858. PMID: 39055814; PMCID: PMC11269824.

Highlights
•
We explore Artificial Intelligence for detection of Marfan Disease (MFS) from simple facial images.
•
AI proves extremely effective at detection of MFS.
•
Overall accuracy is 98.5 % (0 false positives, 2 % false negatives).
•
Clinical usefulness is anticipated.

Abstract

Background

Marfan Syndrome (MFS), a genetic disorder impacting connective tissue, manifests in a wide array of phenotypes which can affect numerous bodily systems, especially the thoracic aorta. The syndrome often presents distinct facial features that potentially allow for diagnostic clinical recognition. Herein, we explore the potential of Artificial Intelligence (AI) in diagnosing Marfan syndrome from ordinary facial images, as assessed by overall accuracy, F1 score, and area under the ROC curve.

Methods

This study explores the utilization of Convolutional Neural Networks (CNN) for MFS identification through facial images, offering a novel, non-invasive, automated, and computerized diagnostic approach. The research examines the accuracy of Neural Networks in the diagnosis of Marfan Disease from ordinary on-line facial images. The model was trained on 80 % of 672 facial images (182 Marfan and 490 control). The other 20 % of images were used as the test set.

Results

Overall accuracy was 98.5 % (0 % false positive, 2 % false negative). F1 score was 97 % for Marfan facies and 99 % for non-Marfan facies. Area under the ROC curve was 100 %.

Conclusion

An Artificial Intelligence (AI) program was able to distinguish Marfan from non-Marfan facial images (from ordinary on-line photographs) with an extremely high degree of accuracy. Clinical usefulness of this program is anticipated. However, due to the limited and preliminary nature of this work, this should be viewed as only a pilot study.