AI Model Slashes Liver Transplant Cancellations by 60%

The introduction of a machine learning-based model by researchers at Stanford Medicine has the potential to significantly reduce the number of canceled liver transplants. This innovative approach predicts whether a donor’s liver will remain viable for transplantation, with the aim of decreasing cancellations by up to 60%.

Currently, the demand for liver transplants far exceeds the supply of available organs. Approximately half of the time, when a match is found with a donor who has died after cardiac arrest, the transplant must be canceled. This typically occurs in cases of donation after circulatory death, where the time frame for successful organ recovery is critical. Specifically, the interval between the removal of life support and the donor’s death must not exceed 30 to 45 minutes to minimize complications for the recipient.

The new model developed by the Stanford team outperforms traditional surgeon judgment, effectively addressing the challenge of predicting donor viability. Kazunari Sasaki, MD, a clinical professor of abdominal transplantation and the senior author of the study, stated, “By identifying when an organ is likely to be useful before any preparations for surgery have started, this model could make the transplant process more efficient.”

Improving Efficiency in Organ Donation

For patients suffering from end-stage liver disease, a transplant remains the best treatment option. The discrepancy between the number of individuals needing a liver and the available donors has been narrowing, thanks in part to advancements such as normothermic machine perfusion. This technique maintains organs at optimal temperatures and oxygen levels during transport from donor to recipient, allowing livers from donation after circulatory death to be utilized for transplant.

While most organ donations originate from donors who have experienced brain death, the increasing number of donations following circulatory death is encouraging. Sasaki remarked, “The number of liver transplants keeps going up because of donation after circulatory death, and the waitlist is getting smaller.” He envisions a future where all who require liver transplants can receive organs from deceased donors.

The challenge with donation after circulatory death primarily revolves around timing. As the donor’s condition declines, the blood supply to organs can diminish, leading to potential liver damage. If the donor’s death occurs more than 30 minutes after blood flow begins to decrease, the viability of the liver for transplantation diminishes significantly.

Advanced Machine Learning Techniques

To predict the precise time of death, the model analyzes a range of clinical information from the donor, including age, gender, body mass index, blood pressure, and vital signs. The researchers tested various machine learning algorithms to identify the most accurate predictor of death within the critical time frame. The most effective algorithm demonstrated a 75% accuracy rate in predicting death, surpassing both existing tools and the average accuracy of surgeons at 65%.

The model’s design allows for customization, accommodating different surgical preferences and hospital protocols. For instance, it can calculate the time of death based on when life support is removed or when agonal breathing begins. Additionally, the research team has created a natural language interface that integrates donor medical records into the prediction model.

Despite the model’s advancements, some missed opportunities for transplantation still occur, with both the model and surgeon judgment yielding a missed opportunity rate of just over 15%. However, with ongoing improvements, the researchers anticipate further reductions in missed opportunities. “We are now working on decreasing the missed opportunity rate because it is in the patients’ best interest that those who need transplants receive them,” stated Sasaki.

The study, which showcases the collaboration of institutions including Kyoto University and several U.S. transplant centers, highlights the potential of artificial intelligence in enhancing the efficiency of organ donation and transplantation. Future iterations of the model are also being explored for use in heart and lung transplants, indicating a broader application of this technology in the medical field.

The findings from the study are published in Lancet Digital Health, and the research continues to evolve, promising a positive outlook for patients awaiting life-saving organ transplants.