Jordan Shaked

Artificial intelligence (AI) is shaping the future of health care, offering new tools for earlier diagnosis of disease and more precise tracking of treatment outcomes. In a new Yale-led study, 

, researchers used a type of AI technology called deep neural network (DNN) analysis to decipher skin involvement and treatment response in patients with systemic sclerosis.

Systemic sclerosis (SSc), or scleroderma, is a chronic autoimmune condition characterized by the overproduction of collagen, a protein that gives tissues strength and structure. The overproduction of collagen can lead to thickening and hardening of the skin and other areas, significantly impacting quality of life.

“Patients often feel double the stress because systemic sclerosis can affect their internal organs and their outward appearance, making the disease very public,” says 

, associate professor of medicine (rheumatology, allergy, and immunology) and primary investigator of the study. “Earlier diagnosis would allow for earlier lifestyle changes and treatment—before internal organ damage occurs—leading to longer, healthier lives.”

The current gold standard for skin thickness assessment in SSc clinical trials is the semi-quantitative modified Rodnan skin score (mRSS). Although the tool is widely used, it has some significant limitations, according to 

, a research assistant in Hinchcliff’s lab and the study’s lead author.

“The mRSS measures dermal thickness through a pinch test, requires long intervals to detect meaningful changes, and can be confounded by factors like obesity and edema,” Gunes says. “Our goal in this study wasn’t to replace the mRSS, but to find complementary methods that are quantitative and reproducible, and that could potentially shorten the length of clinical trials, which often last a year.”

For the study, researchers used deep neural networks to analyze skin biopsies from patients with SSc and generated a “fibrosis score” for each sample. The team is the first to apply AI to SSc skin biopsies.

AI approaches are developing rapidly, and we are experimenting with new methods that may help measure the three components of SSc skin disease: inflammation, vascular abnormalities, and fibrosis. The hope is that AI models can be trained to detect early clinical disease using skin biopsies or chest computed tomography scans so treatments can be initiated to prevent organ damage.
Monique Hinchcliff, MD, MS

The study aimed to evaluate how the deep neural network-derived fibrosis score compared to the mRSS in an SSc clinical trial and to identify which histologic features the DNN detects and quantifies. Researchers found that the DNN fibrosis score showed a weak correlation with the traditional mRSS, and that different histologic features were associated with changes in each measure.

“The low correlation between the mRSS and the fibrosis scores suggests that AI may be capturing skin features beyond what clinicians can detect through a simple pinch test,” Gunes says.

Since the mRSS and fibrosis scores appear to measure distinct pathological features upon histological analysis, it is possible that combining the two approaches may be better than using either one in isolation, she adds.

The researchers hope their findings will help streamline clinical trials, accelerate global recruitment, and improve participant diversity, ultimately enhancing the generalizability of SSc trial results.

Hinchcliff believes that AI will continue to advance earlier diagnosis. “AI approaches are developing rapidly, and we are experimenting with new methods that may help measure the three components of SSc skin disease: inflammation, vascular abnormalities, and fibrosis,” she says. “The hope is that AI models can be trained to detect early clinical disease using skin biopsies or chest computed tomography scans so treatments can be initiated to prevent organ damage.”

The research reported in this news article was supported

by the National Institutes of Health (awards R01AR073270, 1R01GM141309, K23AR075112, R01HL164758, W81XWH2210163, R01HL159620, R01AG083735, and R01AG062109) and Yale University. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Additional support was provided by Sanofi-Aventis U.S., LLC.

The Section of Rheumatology, Allergy and Immunology is dedicated to providing care for patients with rheumatic, allergic, and immunologic disorders; educating future generations of thought leaders in the field; and conducting research into fundamental questions of autoimmunity and immunology. To learn more, visit

Articles

Artificial intelligence (AI) is revolutionizing healthcare by enabling earlier disease diagnosis and precise treatment tracking. A Yale-led study utilized deep neural network (DNN) analysis to assess skin involvement and treatment response in systemic sclerosis (SSc) patients. The study compared DNN-derived fibrosis scores with the modified Rodnan skin score (mRSS), revealing a weak correlation, suggesting AI captures additional skin features. Combining both methods may enhance clinical trial efficiency and diversity. AI's potential in early disease detection and treatment initiation could prevent organ damage in SSc patients.

AI revolutionizes systemic sclerosis diagnosis, enhancing treatment tracking and potentially improving patient outcomes through innovative deep neural network analysis.