October 29, 2024
A research collaboration led by Professor Marc Niethammer and involving researchers from the UNC Department of Computer Science and the UNC School of Medicine received a grant worth slightly less than $4 million from the National Institutes of Health to use machine learning techniques for earlier diagnosis of autoimmune diseases. In addition to computer science, the project involves experts in rheumatology, radiology, and biostatistics.
Systemic lupus erythematosus (SLE), commonly referred to simply as lupus, has a wide variety of possible symptoms which can vary from person to person and change over time. As a result, many cases of lupus are initially misdiagnosed as other diseases with similar symptoms. A diagnosis typically requires an experienced rheumatologist to assess multiple sources of data, including blood and urine tests, imaging tests, biopsy results, patient symptom descriptions, and physical examinations.
Because of the need for a specialist and a variety of medical data, it often takes years for lupus cases to be correctly diagnosed, and long diagnostic delays hinder early patient treatment and can increase the chances of organ damage that can shorten the patient’s lifespan. The disease is nine times more prevalent in women than men and disproportionately affects Black, Asian, and Hispanic women. These factors contribute to health equity issues, making it difficult for the most affected populations to get diagnosed as early as possible.
The multidisciplinary team of researchers will develop machine learning technology to assess many types of medical records, then provide patient-specific disease predictions and guide acquisition of additional data to help make earlier, accurate diagnoses. This new machine learning approach may ultimately also help physicians identify patients at greatest risk for the worst outcomes, enabling more aggressive treatment earlier. The researchers are specifically targeting lupus, but the processes will be generally applicable to diagnosis of other autoimmune diseases and to other multimodal machine learning tasks.
The project, which is funded for two years, is led by Niethammer, with support from fellow computer science faculty members Distinguished Professor Mohit Bansal, Assistant Professor Tianlong Chen, and Associate Professor Junier Oliva. Expertise from outside the department comes from Professor of Radiology Yueh Lee, Distinguished Professor of Medicine and Director of the UNC Rheumatology Lupus Clinic Saira Sheikh, and Professor of Biostatistics Hongtu Zhu.
Niethammer emphasized the potential of this research to improve the lives of those living with autoimmune diseases.
“Autoimmune diseases affect one in 10 people, and patients may needlessly suffer for years due to delays in diagnosis and referral delays to specialists,” Niethammer said. “We hope that our work can enable earlier diagnosis and ultimately earlier and better treatment for some of the worst cases, as well as help to make care more equitable.”
More information about this project can be found on the NIH website.