Illinois ECE ultrasound discovery becomes new tool for detecting early fatty liver disease

A new breakthrough by Illinois ECE researchers will make it easier to detect, prevent, and treat nonalcoholic fatty liver disease (NAFLD). The research team’s methods use noninvasive ultrasound, that could be used during a routine physical, to measure the amount of fat in the liver.  The discovery could have a major impact on the prognoses of millions of people suffering from NAFLD around the world.

The research, led by Electrical and Computer Engineering Research Professor William D O'Brien, Jr., Donald Biggar Willet Professor of Engineering, and Research Assistant Professor Aiguo Han, Food Science and Human Nutrition Professor John Erdman at UIUC, along with UCSD colleagues Professor Claude Sirlin (radiologist), Professor Rohit Loomba (hepatologist) and Professor Michael Andre (medical physicist), was recently published in two Radiology papers.

Early detection of liver fat changes offers the potential to halt or reverse increasing liver fat that could lead to disease which is why O’Brien and Erdman have spent several decades investigating how to improve early-detection techniques.

Currently, there is no routinely available, safe, accurate, and noninvasive capability that assesses the early stages of nonalcoholic fatty liver disease. However, the interdisciplinary team’s groundbreaking findings show that diagnostic ultrasound imaging offers the potential to change that.

The team has discovered that ultrasound systems, similar to those used regularly to assess the wellbeing of the fetus in utero, can quantify liver fat in a simple and accessible way. 

“Our research findings provide the scientific basis for future clinical capabilities for a low-cost and portable capability for liver evaluation that can be readily used during office visits with your clinician.  Liver fat could be quantified simply during routine physicals and potentially detect threatening changes in a way that is more accessible than ever before,” O’Brien explained.

NAFLD is the most common chronic liver disease worldwide and affects about 25% of the human population. For the US, that translates to about 82 million humans and worldwide about 2 billion persons. As liver fat increases from normal (less than 5%; non-NAFLD) to higher fat content values, liver conditions can progress from hepatic steatosis to nonalcoholic steatohepatitis to fibrosis to cirrhosis and even to hepatocellular carcinoma.

NAFLD is also an important risk factor for Metabolic Syndrome (MS) development. MS is a cluster of conditions that includes type 2 diabetes, heart disease, and stroke. Underlying conditions leading to MS are high blood pressure, high blood sugar, excess body fat around the waist, and abnormal blood lipids.    With over 3 million new cases yearly in the United States, prevention and management of MS are critical to health and longevity.  

“The availability of a quick, noninvasive measurement of liver fat allows clinicians to quickly assess and make recommendations to patients regarding dietary changes, physical activity, etc., to minimize risk of developing MS,” Han said. 

The team’s recent discovery is a combination of two novel studies. Assessment of Hepatic Steatosis and Nonalcoholic Fatty Liver Disease by Using Quantitative US” and “Noninvasive Diagnosis of Nonalcoholic Fatty Liver Disease and Quantification of Liver Fat with Radiofrequency Ultrasound Data Using One-dimensional Convolutional Neural Networks.” 

“Uniquely different processing strategies were used on two separate patient cohorts to yield comparable outcomes to demonstrate that noninvasive ultrasound examination of human livers not only diagnosed NAFLD but also quantified fat in the liver,” Erdman explained.

For both studies, MRI was used as the fat fraction reference standard to yield the liver’s fat content in terms of the proton density fat fraction (PDFF). 

“While MRI is safe and reliable, it is not routinely accessible in the United States, and less so worldwide for quantifying liver fat fraction,” O’Brien said.

The two cohorts used different subject populations (cohort 1. 102 participants recruited between August 2015 and February 2019, reported in the Radiology paper that used quantitative ultrasound to evaluate NAFLD; and cohort 2. 204 participants recruited between February 2012 and March 2014, reported in the Radiology paper that used deep learning techniques to evaluate NAFLD

The evaluation strategies of the two studies were quite different but yielded comparable outcomes: 

“The outcomes from both cohort studies were procedurally the same. One of the outcomes (a classifier) was the diagnosis of NAFLD (that is, differentiate with/without NAFLD; NAFLD is PDFF greater than or equal to 5%) and the other outcome was the quantification of the liver fat fraction (that is, a fat fraction estimator),” O’Brien explained.

Both classifiers performed well, both yielding 96% accuracy. Both fat fraction estimators also performed well, both yielding correlations with PDFF of 0.76 for cohort 1 and 0.85 for cohort 2.

The interdisciplinary team’s findings are now at the forefront of advancements in the field of radiology and the fight against NAFLD.  At least one of the diagnostic ultrasound companies has now included a built-in “fat fraction estimator” as part of their ultrasound system capability. “There have been no studies, to date, that have addressed the accuracy of this company’s estimator, but it is highly encouraging that a company believes there is a clinical need for quantitative assessment of the liver using ultrasound, “O’Brien said. 

The team’s research was funded by a 5-year grant from the National Institute of Health’s National Institute of Diabetes and Digestive and Kidney Diseases (R01DK106419). The research team is currently developing a competing renewal to continue the work of developing strategies for the diagnosis and detection of liver diseases using noninvasive ultrasound technologies.

O’Brien is affiliated with the Beckman Institute, CSL, Division of Nutritional Sciences, and HMNTL.