Application of Artificial Intelligence (AI) in Medical Diagnosis Based on Liver Imaging
Fatty liver disease (FLD) is a common disease, with no or very mild symptoms, affecting approximately 1 out of 4 people. Excess fat builds up in the liver, causing health problems. Alcohol consumption may affect the onset or severity of this disease, but it is also prevalent in people that consume little or no alcohol. This version of FLD is referred to as non-alcoholic fatty liver disease (NAFLD).
No medicines are available to treat NAFLD; only changes in diet, alcohol consumption, weight loss, and exercise are recommended by medical professionals.
If NAFLD is not properly managed it can progress to complications. Cirrhosis, a condition where scar tissue builds up damaging the liver is the end stage of NAFLD. When the liver is scarred, high blood pressure in the liver can occur in blood vessels in the esophagus and lead to bleeding issues, or the scarring can also lead to liver cancer, which both can be fatal.
The research planned will create new tools for physicians to perform early diagnosis of liver disease from a non-invasive test, magnetic resonance imaging (MRI), and extending the learnings to the common tool of ultrasound. Usually, doctors detect, characterize, and monitor diseases by assessing pictures of the liver visually, but physicians are human, and are affected by their learning and biases, and may not be able to observe subtle changes. This project will develop innovations using artificial intelligence to recognize liver features and changes in those features automatically with the potential to assist physicians in making more accurate and reproducible imaging diagnosis significantly reducing a physicians' workload. Because fatty liver is so common, we believe that making a measurement tool that can be used with multiple common pieces of equipment (MRI and Ultrasound) will ultimately improve the public's health and enable new medicines to be discovered and developed.
The research project to develop the automated imaging characterization tools for researchers and physicians to use in assessing FLD is anticipated to take 24 to 36 months.