Using AI to Find a Less-Invasive Alternative to Liver Biopsy

Blood test could help physicians diagnose and monitor common liver disease in at-risk patients. 

Boston, Mass. – Endocrinologists at Beth Israel Deaconess Medical Center (BIDMC) have developed a non-invasive test that can help doctors keep tabs on patients’ liver health. By measuring combinations of fats, hormones, carbohydrate and sugar molecules present in patients’ blood serum, the team of scientists used machine learning to develop models that can differentiate various stages of liver disease with near-perfect accuracy. The team reported its findings in the journal Metabolism, Clinical and Experimental, along with an accompanying editorial emphasizing the model’s potential impact.

Non-alcoholic fatty liver disease (NAFLD) is the most common chronic liver disease worldwide, affecting as many as one out of every three people, and is especially prevalent among patients with Type 2 diabetes and/or obesity. Patients with NAFL may never develop any further complications, but about 25 percent of people with NALFD will develop more serious liver inflammation called non-alcoholic steatohepatitis, or NASH, can eventually lead to cirrhosis or liver cancer. 

Today, physicians assess liver health with liver biopsy – a costly method that carries risks for the patient and is not always representative of the actual status of the entire liver. 

“It would be impossible to do biopsies in the 80 million at-risk Americans, and even if we did, it would result in tens of thousands of subjects suffering complications and about 16,000 deaths each year from complications,” said senior author Christos Mantzoros, MD, DSc, Director of the Human Nutrition Unit at BIDMC and Professor of Medicine at Harvard Medical School. “Finding easy to obtain, relatively inexpensive and reliable biomarkers which can be measured with less invasive techniques is an urgent, unmet need.”

Using five different machine learning techniques in two different platforms, Mantzoros and team generated novel predictive models that could diagnose patients’ liver health status with excellent accuracy. One model using just 10 lipids to detect the presence of liver fibrosis – the development of fibrous tissue in the organ which can be indicative of injury or disease – achieved 98 percent accuracy. 

Mantzoros and team analyzed blood serum drawn from 80 people with known liver health status as diagnosed by traditional liver biopsy evaluated by two independent pathologists. Of the 80 participants, 49 had no liver disease, 15 patients were diagnosed with the lower-risk NAFL and 16 patients were diagnosed with the more serious NASH. Next, the scientists analyzed differences in blood concentration of 366 lipids, 23 fatty acids, 62 glycans (carbohydrate- or sugar-based molecules), and four hormones – all known players in the development of liver disease – across the three groups of patients. 

“We measured as many circulating molecules as reasonably possible and then let machine learning and artificial intelligence pick the best sets of molecules that would most accurately predict outcomes,” said Mantzoros. “Although the number of subjects appears small given conventional study designs, employing powerful and novel artificial intelligence models allowed us to derive accurate results, as high as 98 percent in some cases. These models may serve as low-risk, cost-effective alternative method to liver biopsy for diagnosis and monitoring NALFD.”

Further studies will seek to validate the models in larger studies with more ethnically diverse participants, as well as include additional variables useful for enhancing the models’ predictive capabilities, such as patients’ genetic profile, clinical information including age and BMI. Such improved diagnostic tools may have the ability to also differentiate between subgroups within each category including stages of fibrosis.   

Co-authors include Nikolas Perakakis, MD, of BIDMC; Stergios A. Polyzos, MD, MSc, PhD, and Jannis Kountoouras, MD, PhD of Aristotle University of Thessaloniki; Alireza Yazdani, PhD, of Brown University; Aleix Sala-Vila, DPharm, PhD, of Instituto de Salud Carlos III; and, Athanasios D. Anastasilakis, MD, PhD, of 424 General Military Hospital, Thessaloniki.

This work was supported by the National Institutes of Health (NIH K24DK081913); DFG, the German Research Foundation –389891681 (PE 2432/2-1); and the Instituto de Salud Carlos Miguel Servet fellowship (grant CP II 17/00029). 

Dr. Mantzoros is consultant to Intarcia, is a grant recipient through BIDMC, and consultant to Novo Nordisk and Ansh Labs LLC and consultant to and shareholder of Coherus.