Advanced Radiomic Analysis Identifies Four Distinct Profiles of Lung Damage in Sarcoidosis Patients

National Jewish Health researchers use AI radiomics to identify four distinct sarcoidosis profiles, improving lung function prediction and diagnostic accuracy.

By: AXL Media

Published: Apr 10, 2026, 3:55 AM EDT

Source: Information for this report was sourced from National Jewish Health

Decoding the Complexity of Inflammatory Lung Disease

Sarcoidosis is a multifaceted inflammatory condition that affects approximately 150,000 Americans, primarily targeting the lungs through the formation of granulomas, inflammation, and potential scarring. Historically, pulmonologists have relied on the Scadding staging system and subjective visual reviews of chest CT scans to evaluate the extent of the disease. However, a new study led by researchers at National Jewish Health reveals that these traditional methods may fail to capture the full spectrum of the disease’s pathology. By applying radiomics—the quantitative extraction of hundreds of features from medical images—investigators have found a way to see patterns that remain invisible to the naked eye.

Four Distinct Radiomic Phenotypes Identified

The research team analyzed high-resolution CT scans from 320 participants enrolled in the Genomic Research in Alpha-1 Antitrypsin Deficiency and Sarcoidosis (GRADS) study, a major national cohort. Using machine-learning clustering techniques, they identified four unique patient groups based on the structural "texture" of their lung tissue. These groups ranged from individuals with minimal abnormalities to those exhibiting patterns consistent with extensive fibrosis or widespread inflammation. According to co-senior author Dr. Tasha Fingerlin, these profiles provide a much more detailed and reproducible way to categorize how the disease varies from one patient to another.

Objectivity Beyond Traditional Visual Staging

A significant finding of the study is that these four radiomic profiles accounted for variations in lung function even after researchers adjusted for traditional imaging assessments used in standard clinical practice. While current staging systems are helpful, they are often limited by inter-rater variability, where two specialists might interpret the same scan differently. Radiomics removes this subjectivity by using automated algorithms to measure physical features of the lung parenchyma. This objective approach allowed the team to map specific imaging signatures directly to clinical outcomes, such as forced vital capacity and other markers of respiratory health.