A RECENT study has introduced a new artificial intelligence tool called GraftIQ, which may help improve the diagnosis and care of liver transplant patients. Liver transplantation can be life-saving for people with end-stage liver disease, but even after a successful operation, patients are still at risk of graft injury, which may lead to serious complications or even loss of the transplanted organ.
Currently, liver graft problems are often flagged by abnormal blood tests, but these are non-specific and do not reveal the exact cause of the issue. A liver biopsy is still considered the best way to diagnose graft injury, but it carries risks such as bleeding and infection, and can delay treatment decisions. Doctors often have to rely on experience and best judgement while waiting for biopsy results, highlighting the urgent need for accurate, non-invasive diagnostic tools.
The researchers behind GraftIQ used a hybrid approach that combines machine learning with expert clinical knowledge. The model analyses demographic information, lab results, and other clinical data to predict the cause of graft injury across six possible conditions. Unlike traditional ‘black box’ neural networks, GraftIQ also provides explanations by identifying which clinical factors are most relevant to each diagnosis. For example, raised liver enzymes were strongly associated with rejection, one of the most common post-transplant complications.
The model was tested on international patient data from the US, Germany, and Singapore, and consistently performed well, with high accuracy scores across all datasets. In one case, it correctly predicted rejection with 81% certainty, even when other conditions were also possible.
The study authors hope GraftIQ will allow for faster, more precise diagnosis without immediate need for biopsy, enabling quicker treatment and better long-term outcomes. A clinician-facing dashboard is in development, allowing healthcare providers to input patient data and receive instant diagnostic insights. If adopted in clinical practice, this technology could reduce dependence on invasive tests and help protect transplanted livers from avoidable damage.
Reference
Sharma D et al. GraftIQ: Hybrid multi-class neural network integrating clinical insight for multi-outcome prediction in liver transplant recipients. Nat Commun. 2025;16(1):4943.
