A Machine Learning model developed by Stanford scientists may significantly improve the liver transplant process and reduce associated costs.
The number of people awaiting a liver transplant exceeds the number of available organs. Currently, a growing portion of liver donations are carried out through Donation After Circulatory Death (DCD).
In this type of donation, the time between the withdrawal of life support and the donor’s death must not exceed 30–45 minutes to prevent damage to the liver and minimize the risk of complications for the recipient. In such cases, the bile ducts in the liver may be damaged, which poses a serious risk of complications for the recipient, leading surgeons to often decline the use of such an organ.
The problem is that in these cases, at least half of the transplantation procedures are canceled because the organs are no longer viable for transplant. This can lead to the wasting of resources (such as expensive normothermic machine perfusion equipment).
Stanford researchers created a Machine Learning model that predicts whether the donor’s death will occur within the timeframe when their organs are still viable for transplantation.
The model utilizes the donor’s clinical information: sex, age, weight, vital signs (blood pressure, heart rate), blood tests, and other vital parameters.
The model reduced the rate of unsuccessful liver procurements (where surgical preparation has begun but the liver is not usable) by 60%.
The model predicts the time of death 75% more accurately than surgeons.
Ultimately, this Artificial Intelligence model makes it possible to accurately identify which organ will be viable before surgical preparation begins.
