Explainable Machine Learning for the Assessment of Donor Grafts in Liver Transplantation

Clinically, organ evaluation generally performed by the senior surgeons based on their experience and the visual and tactual inspection of the graft during procurement. However, it is proved that transplant surgeons intuition in the evaluation of donor risk and the estimation of steatosis is inconsistent and usually inaccurate. Besides, graft assessment is a dynamic process refer to amount of complex factors, which is considered to be an incredibly complicated relationship that is nonlinear in nature. Unfortunately, the classical statistic techniques in vogue such as multiple regression require the statistical assumption of independent and linear relationships between explanatory and outcome variables, and fail to analyse a large number of variables. We attempted to develop liver graft assessment models by predicting postoperative DGF using several ML techniques. Secondly, the best prediction model was selected by comparing the performance of different AI algorithms and logistic regression. Finally, we sought to explain the decision made by AI algorithms using a visualization algorithm based on the best prediction model, helping clinicians evaluate specific organ and whether to receive that may develop DGF postoperatively.

Study Overview

• Status: Completed
• Conditions: Liver Transplant; Complications
• Intervention / Treatment: Procedure: Liver transplantation

• Study Type: Observational
• Enrollment (Actual): 5636

Contacts and Locations

Study Locations

• China
  • Guangdong
    • Guangzhou, Guangdong, China
      • The Third Affiliated Hospital of Sun Yat-Sen University

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Non-Probability Sample

Study Population

Adult patients who underwent deceased donor liver transplantation

Description

Inclusion Criteria:

1. Age≥18 years-old
2. Underwent deceased donor liver transplantation

Exclusion Criteria:

1. Underwent living-donor LT;
2. Missing rates of data were more than 80%

Study Plan

How is the study designed?

Number of groups / cohorts

2

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
DGF; patients occured delayed graft function after liver transplantation	Procedure: Liver transplantation; Liver transplantation
non-DGF; patients NOT occured delayed graft function after liver transplantation	Procedure: Liver transplantation; Liver transplantation

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Delayed Graft Function (DGF)	defined as early graft dysfunction without the need for a second liver transplant or death	Within 7 days after liver transplantation

Collaborators and Investigators

• Sponsor: Third Affiliated Hospital, Sun Yat-Sen University
• Collaborators: the China Liver Transplant Registry

Study record dates

Study Major Dates

• Study Start (Actual): January 1, 2017
• Primary Completion (Actual): June 30, 2023
• Study Completion (Actual): June 30, 2024

Study Registration Dates

• First Submitted: July 28, 2024
• First Submitted That Met QC Criteria: July 30, 2024
• First Posted (Actual): August 2, 2024

Study Record Updates

• Last Update Posted (Actual): August 2, 2024
• Last Update Submitted That Met QC Criteria: July 30, 2024
• Last Verified: July 1, 2024

More Information

Terms related to this study

• Other Study ID Numbers: ZSSY-2403

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No