An AI Model Based on Smartphone-derived Multimodality Images to Evaluate Portal Hypertension in Patients With Cirrhosis (CHESS2203)

Overview

Portal hypertension contributed to the main complications of liver cirrhosis. Currently, hepatic venous pressure gradient (HVPG) was the reference standard for evaluating portal pressure in patients with cirrhosis. However, the practice of HVPG is limited to require the extensive experience and highly specialized centers. In recent years, non-invasive methods were proposed to predict the degree of cirrhotic portal hypertension. Liver stiffness is currently the most widely used method for noninvasive assessment of portal hypertension. The renewing Baveno VII recommended that liver stiffness ≥ 25 kPa by transient elastography is sufficient to identify clinically significant portal hypertension (specificity and positive predictive value \> 90%). Although liver stiffness has a good predictive value for evaluation of clinically significant portal hypertension, it is difficult to apply in primary hospitals due to expensive equipment. Recently, a multicenter study has shown that artificial intelligence analysis based on ocular images can aid to screening and diagnosis hepatobiliary diseases. The patented technology of collecting and analyzing diagnostic images of Traditional Chinese Medicine (TCM) based on mobile phone terminals has been realized. This technology mainly includes image acquisition, quality control and analysis, and clinical information collection. Liver cirrhosis belongs to the diseases of bulging and accumulation in TCM, and the most common symptoms are the liver and gallbladder damp-heat and liver stagnation and spleen deficiency. The main contents of inspection diagnosis in TCM for liver disease include the images of the tongue, eye and palms. In our study, the patented technology of TCM based on artificial intelligence is applied to establish a precise evaluation model of traditional Chinese and western medicine for portal hypertension with cirrhosis by combining the macroscopic characteristics of images and microscopic pathological indicators.

Portal hypertension contributed to the main complications of liver cirrhosis. Currently, hepatic venous pressure gradient (HVPG) was the reference standard for evaluating portal pressure in patients with cirrhosis. However, the practice of HVPG is limited to require the extensive experience and highly specialized centers. In recent years, non-invasive methods were proposed to predict the degree of cirrhotic portal hypertension. Liver stiffness is currently the most widely used method for noninvasive assessment of portal hypertension. The renewing Baveno VII recommended that liver stiffness ≥ 25 kPa by transient elastography is sufficient to identify clinically significant portal hypertension (specificity and positive predictive value \> 90%). Although liver stiffness has a good predictive value for evaluation of clinically significant portal hypertension, it is difficult to apply in primary hospitals due to expensive equipment. Recently, a multicenter study has shown that artificial intelligence analysis based on ocular images can aid to screening and diagnosis hepatobiliary diseases. The patented technology of collecting and analyzing diagnostic images of Traditional Chinese Medicine based on mobile phone terminals has been realized. This technology mainly includes image acquisition, quality control and analysis, and clinical information collection. Liver cirrhosis belongs to the diseases of bulging and accumulation in Traditional Chinese Medicine, and the most common symptoms are the liver and gallbladder damp-heat and liver stagnation and spleen deficiency. The main contents of inspection diagnosis in Traditional Chinese Medicine for liver disease include the images of the tongue, eye and palms. In our study, the patented technology of Traditional Chinese Medicine based on artificial intelligence is applied to establish a precise evaluation model of traditional Chinese and western medicine for portal hypertension with cirrhosis by combining the macroscopic characteristics of images and microscopic pathological indicators.