SOPHIA ANTIPOLIS, France—Median Technologies, The Imaging Phenomics Company, recently announced the results of an artificial intelligence (AI)-based biomarker study which performed better than existing imaging biomarkers on evaluating the severity of hepatic fibrosis in non-alcoholic steatohepatitis (NASH) patients. If future studies hold true, doctors and medical professionals may one day use a new imaging biomarker extracted from MRI/MRE images to diagnose NASH early and save more lives.
Non-alcoholic fatty liver disease (NAFLD) affects 25 percent of the global population. For 20 percent of these patients, NAFLD leads to a more severe disease called NASH, according to the World Health Organization (WHO). Within the high-risk population of NASH patients, disease evolution can lead to cirrhosis and liver cancer.
“These initial results are exciting and show, once again, the relevance of our iBiopsy approach and the tremendous possibilities of AI-based technologies to extract hidden information from standard medical images and generate non-invasive biomarkers,” says Fredrik Brag, CEO and co-founder of Median. “These biomarkers are developed by considering the whole organ, in this case the liver. This is a particularly important point for the management of diffuse diseases such as NASH fibrosis, and an extremely important advantage when compared to biopsies, which give fibrotic scoring information solely based on a small sample.”
For the biopharmaceutical industry, “The use of a non-invasive test with strong discriminating power for the selection of patients to be included in NASH trials would make it possible to reduce the failure rates linked to patient recruitment, as well as decrease the costs of the trials,” he adds.
“NASH is a silent disease,” Brag tells DDN. “In the early fibrosis stages (F1 to F3), NASH evolution can be slowed down, even stopped by a change in lifestyle and eating habits, meaning that diagnosing the disease early can save patients’ lives. There is currently no cure that is efficient against NASH, and finding the disease early can have a huge impact for patients and a major medical economic impact as well.”
Models estimate a rise in annual US economic burden to $103 billion from direct medical care costs ($1,613 per patients, 52 million patients), and another $188 billion in societal costs related to NAFLD. The estimated global prevalence of NAFLD is as high as one billion.
In the United States and the European Union, 25 percent of the population has NAFLD, of which nearly 25 percent progress to NASH. NASH prevalence has increased steadily over the past few decades, as has the percentage of people with NASH who developed cirrhosis, liver failure, and/or hepatocellular carcinoma.
“Populations who have certain health conditions like obesity and conditions related to obesity—type 2 diabetes/pre-diabetes, high blood pressure, and abnormal levels of cholesterol—are at high risk of developing NAFLD, then NASH,” Brag remarks. “And some national and international recommendations exist providing guidelines for screening these high-risk populations. We are preparing the strategic regulatory paths in the US and EU, to bring iBiopsy to patients in the near future.
“Our next step is to validate our preliminary study results on larger and independent patient cohorts. We plan [for] new results to be published in 2021-22.”
A NASH prognosis mostly depends on the hepatic fibrosis grade, according to Brag. Hepatic biopsy is a method used to measure hepatic fibrosis severity. Due to its invasiveness, and sampling errors related to small sample size as well as spatial variation in degree of fibrosis, hepatic biopsy is not used as first indication for diagnostic purposes. In routine clinical practice, the assessment of the absence of advanced fibrosis of the liver is done by non-invasive tests such as blood tests and hepatic elastography. The biopsy is, however, essential for the diagnosis of advanced NASH.
In its early stages, the disease remains reversible by a change in eating habits and lifestyle. The clinical interest is therefore to distinguish accurately and non-invasively patients with early fibrosis from patients with advanced fibrosis at risk of progressing to cirrhosis and liver cancer.
In this clinical context, the objective of Median’s preliminary retrospective study was to quantify the ability of iBiopsy’s learning algorithms to discriminate between early and advanced fibrosis grade in NASH patients.
MRE and MRI images from a cohort of NASH patients with F2-F3 fibrosis grades (n=152 observations), based on the NASH CRN score (F0-F4), were used to model the relationships between liver image features and early and advanced fibrosis histological grades. The performance of iBiopsy testing characterized by the area under the curve (AUROC) is 0.90 with a specificity of 0.89 and a sensitivity of 0.86 for the diagnosis of advanced fibrosis (F3) based on MRE.
“In clinical routine, morphologic signs on MRI alone are unreliable and insufficient to detect even significant liver fibrosis,” Brag says. “However, using AI-based iBiopsy technologies applied on MRI, the prediction performance is encouraging with a specificity of 0.83 and a sensitivity of 0.72. We expect to further increase the accuracy through machine learning algorithm optimization.”