An Artificial Intelligence Solution for Noninvasively Identifying Portal Hypertension?

The most accurate method of diagnosing clinically significant portal hypertension (CSPH) in patients with cirrhosis is by measuring hepatic venous pressure gradient (HVPG). However, this technique is invasive and not readily available in all clinical settings.

In the current study, researchers developed and validated deep convolutional neural network (CNN) models using abdominal scans (liver and spleen) from computed tomography (CT) and magnetic resonance imaging (MRI) to identify CSPH (defined as HVPG ≥10 mmHg). They used data from two prospective, multicenter trials conducted in China and Turkey in which CT or MRI scans were performed within 14 days of HVPG measurement via transjugular catheterization. Two cohorts, divided into CT and MRI …