LiPNet

Liver Perfusion Network model. Computes the three normalised loads a graft carries, z_P = z_F × z_R: pressure load = flow load × resistance load.

Pressure (after reperfusion)

Portal venous pressure PVP, mmHg Central venous pressure CVP, mmHg (blank = 5 mmHg assumed) Portal venous pressure after the planned change, mmHg

Normal gradient fixed at 5 mmHg (the coefficients were fitted on it).

Flow

Graft portal flow PVF, mL/min Graft weight, g (or GRWR + recipient weight) GRWR, % Recipient weight, kg Donor PVF per 100 g reference, mL/min/100 g (90 = healthy live donor) '

Reference level

Anchor for the absolute risk Outcome rate, % Average portocaval gradient of that cohort, mmHg (only for the approximate option)

Pressure load z_P

–

Flow load z_F

–

Effect of the change in gradient

–

Reading

Risk curve

The curve is anchored on the reference you chose; the markers are the current gradient and the gradient after the change. Without an anchor it shows the published groups, each centred on its own cohort.

What this does and does not give you. The effect of changing the gradient, as estimated by the hierarchical model reported in the paper, is transferable: OR 4.70 per unit of z_P (95% CI 1.38–16.25), 1.36 per mmHg, from 7 groups in 3 series. The level of risk is not: at the same gradient, published rates run from 0% to 17%. So the percentages shown are only as good as the baseline you supply. Research tool, not a medical device.

Send us this case

LiPNet needs individual measurements to be validated. If you keep PVP, CVP, flow and outcome for your recipients, five columns are enough. Nothing here leaves your browser: the button copies one tab-separated line you can paste into a spreadsheet.

Case identifier (your own code, no patient data) Outcome (leave any box unticked if not known) SFSS or early graft dysfunction graft loss within 90 days death within 90 days outcome known and none of the above occurred

The row has the 21 columns of data/cohort_template.tsv, in that order; fields the calculator does not know are left empty for you to fill. Values rejected on screen are exported empty, so the file always matches what you see.

What modulation does in the published series

Manoeuvre	Reported effect	Source
Splenic artery ligation	PVF/100 g 360 → 240 (−33%); PVP −4.4 mmHg on average with intentional control	Troisi 2003 Tab 4; Ogura 2010
Splenectomy	PVP 23.9 → 19.1 mmHg (−4.8); 21–33 → 14–19 mmHg in five patients	Wang 2014 Fig 1; Osman 2017
Hemiportocaval shunt	PVF/100 g 537 → 190 (−65%, then banded to ≈2× donor); PVP kept <20 with gradients 3–13 mmHg	Troisi 2005 Tab 4; Yamada 2008 Tab 4
Middle hepatic vein / outflow	PVF 3.9× donor with gradient 6 mmHg and no injury: pressure falls below flow	Chan 2011

Below z_P ≈ 1 (gradient under normal) the model predicts vessel regression: portal steal. Troisi calibrated shunts to about twice the donor flow; Kyoto reported steal with a 23.7% GV/SLV graft.

Data extracted from the original tables of Troisi 2003, Troisi 2005, Ou 2010, Vasavada 2014, Alim 2016, Chan 2011, Yagi 2005, Yagi 2006, Wang 2014, Osman 2017, Ogura 2010, Yamada 2008, Uemura 2016, Yao 2018, Kanetkar 2017, Ishizaki 2012 and Botha 2010. LiPNet, a dissipation-optimal perfusion network with fixed vascular maintenance cost (Vázquez-Victorio, Pérez-Calixto, Escutia-Guadarrama, Tovar, Pérez-Calixto, 2026). Source: https://github.com/Danpc11/LiPNet