Predicting liver failure after liver surgery using deep learning

Usability and Clinical Effectiveness of an Interpretable Deep Learning Framework (VAE-MILP) Using Counterfactual Explanations and Layerwise Relevance Propagation Framework for Post-Hepatectomy Liver Failure Prediction

This study tests if a new deep learning tool can help doctors predict liver failure after surgery in patients with liver cancer.

Quick facts

• Study type: Observational
• Enrollment: 80 (estimated)
• Sex: All
• Sponsor: Maastricht University (other)
• Locations: 1 site (Guangzhou, Guangdong)
• Trial ID: NCT06031818 on ClinicalTrials.gov

What this trial studies

This observational study aims to evaluate the usability and clinical effectiveness of an interpretable deep learning framework called VAE-MLP for predicting post-hepatectomy liver failure (PHLF) in patients with hepatocellular carcinoma (HCC). The study will involve clinicians and radiologists assessing the model's predictions under two conditions: with and without model explanations. By utilizing two-dimensional shear wave elastography (2D-SWE) images, the framework seeks to enhance the interpretability of predictions, addressing the 'black box' issue commonly associated with deep learning models. The goal is to improve clinical decision-making and foster trust in AI-assisted predictions.

Who should consider this trial

Good fit: Ideal candidates for this study are patients with treatment-naive and resectable hepatocellular carcinoma who have a performance status score of 0-1.

Not a fit: Patients who have not undergone liver resection or have a pathological diagnosis other than HCC may not benefit from this study.

Why it matters

Potential benefit: If successful, this study could lead to more accurate predictions of liver failure after surgery, improving patient outcomes and clinical decision-making.

How similar studies have performed: While the use of deep learning in medical imaging is growing, this specific approach to predicting post-hepatectomy liver failure using an interpretable model is relatively novel and has not been extensively tested in prior studies.

Eligibility criteria

Inclusion Criteria:

1. patients with treatment-naive and resectable HCC;
2. performance status Eastern Cooperative Oncology Group (PS) score 0-1.

Exclusion Criteria:

1. liver resection was not performed;
2. pathological diagnosis of non-HCC;
3. failure in liver stiffness measurement defined as the elastography color map was less than 75% filled or interquartile range (IQR)/median \> 30%;
4. immune-active chronic hepatitis indicated by an elevation of alanine aminotransferase (ALT) levels ≥ 2×upper limit of normal (ULN);
5. obstructive jaundice or dilated intrahepatic bile ducts with a diameter of \>3 mm;
6. hypoalbuminemia, hyperbilirubinemia, or coagulopathy not related to the liver.

Where this trial is running

Guangzhou, Guangdong

• The First Affiliated Hospital of Sun Yat-Sen University — Guangzhou, Guangdong, China (RECRUITING)

Study contacts

• Principal investigator: Philippe Lambin — Maastricht University
• Study coordinator: Xian Zhong

Conditions: Post-hepatectomy Liver Failure, Hepatocellular Carcinoma, Artificial Intelligence, deep learning, interpretability, post-hepatectomy liver failure, hepatocellular carcinoma, shear wave elastography