Estimating portal pressure from a research-fitted model
For my Clinical Technology research project (KTO) at TU Delft, I worked on a study into TIPS (a transjugular intrahepatic portosystemic shunt) and its effect on portal pressure. From the study data a linear model was fitted to predict the portal pressure gradient (PPG) after the TIPS procedure. I turned that model into a small, interactive web tool so the relationship between the imaging and clinical inputs and the predicted PPG can be explored directly in the browser.
The predictor uses the fitted linear model. Only the variables that actually carry weight in the equation are shown: the pre-treatment HVPG (hepatic venous pressure gradient), liver volume, the diameters of the portal, splenic and superior mesenteric veins, the presence of esophageal varices (slokdarmvarices), and a Liver/BMI ratio. BMI itself has no direct weight but is still required, because it feeds that ratio.
On the website the coefficients are fixed and read-only, so the formula cannot be changed — only the inputs. Every input updates the predicted PPG live, along with a clinical band (normal, elevated, or clinically significant portal hypertension).
Enter the imaging and clinical values below and the predicted PPG updates instantly. This is the read-only website version of the tool — the model itself runs entirely in your browser.
Example model — not clinically validated. The predicted values are indicative and must not be used for clinical decision-making.