Use of artificial intelligence for liver diseases: A survey from the EASL congress 2024

Abstract

Artificial intelligence (AI) methods enable humans to analyse large amounts of data, which would otherwise not be feasibly quantifiable. This is especially true for unstructured visual and textual data, which can contain invaluable insights into disease. The hepatology research landscape is complex and has generated large amounts of data to be mined. Many open questions can potentially be addressed with existing data through AI methods. However, the field of AI is sometimes obscured by hype cycles and imprecise terminologies. This can conceal the fact that numerous hepatology research groups already use AI methods in their scientific studies. In this review article, we aim to assess the contemporaneous use of AI methods in hepatology in Europe. To achieve this, we systematically surveyed all scientific contributions presented at the EASL Congress 2024. Out of 1,857 accepted abstracts (1,712 posters and 145 oral presentations), 6 presentations (∼4%) and 69 posters (∼4%) utilised AI methods. Of these, 55 posters were included in this review, while the others were excluded due to missing posters or incomplete methodologies. Finally, we summarise current academic trends in the use of AI methods and outline future directions, providing guidance for scientific stakeholders in the field of hepatology.

Keywords: Deep learning; MASLD; biomarkers; large language models; liver cancer; liver cirrhosis; liver fibrosis; machine learning; medical data; medical image analysis.

Fig. 1

Overview of potential applications of AI in addressing complex and challenging spectrum of liver diseases. Certain risk factors may lead to ALD, MASLD, or immune response, where AI may help predict further progression, thus stratifying patients by risk. Subsequent liver inflammation can progress to hepatic fibrosis and then cirrhosis, conditions marked by an increased risk of liver cancer: HCC, CCA or combined HCC-CCA. AI can assist in differential diagnosis and selecting optimal treatment. AI, artificial intelligence; ALD, alcohol-associated liver disease; ASH, alcohol-associated steatohepatitis; CCA, cholangiocarcinoma; ECM, extracellular matrix; MASH, metabolic dysfunction-associated steatohepatitis; MASLD, metabolic dysfunction-associated steatotic liver disease; PBC, primary biliary cholangitis; PSC, primary sclerosing cholangitis.

Fig. 2

Overview of common clinical data types for AI-based analysis and related studies from the EASL Congress 2024. (A) Many different clinical data types are suitable for AI analysis if appropriately prepared, e.g. images might need to undergo ROI selection. (B) Studies included in this review, categorised by application field. The inner donut chart indicates the distribution according to the data modality. Pink dots – machine learning; blue – deep learning; grey – unknown; black border – commercial; ∗ – young investigator. Histopathology slide is from The Cancer Genome Atlas, National Cancer Institute. ALF, acute liver failure; MASLD, metabolic dysfunction-associated steatotic liver disease; ROI, regions of interest.

Fig. 3

Overview of DL architectures, applications in clinical image analysis, and methods for explainability. (A) The most common architectures of DL models that can be utilised for any image data. (B) Example applications (not all from EASL Congress 2024) of DL models covered in the clinical image analysis section. (C) Foundation models trained in a self-supervised way can be utilised for various downstream prediction tasks in smaller cohorts. (D) Common explainability methods for 'black-box' DL models. The histopathology slide is from The Cancer Genome Atlas. Liver CT is from the open ‘DeepLesion’ dataset. CCA, cholangiocarcinoma; DL, deep learning; WSI, whole slide image.

Fig. 4

Overview of classical ML methods and their applications. (A) The most popular classical ML methods. (B) Example applications covered in this review utilising these methods (C) Decision tree-based models have greater explainability than deep learning models. ALD, alcohol-associated liver disease; ALF, acute liver failure; CCA, cholangiocarcinoma; CHB, chronic hepatitis B; CHC, chronic hepatitis C; LRE, liver-related events; LT, liver transplantation; MASLD, metabolic dysfunction-associated steatotic liver disease; MASH, metabolic dysfunction-associated steatohepatitis; ML, machine learning; PHLF, post-hepatectomy liver failure; SLD, steatotic liver disease. ∗Indicates a young investigator.

Fig. 5

Limitations of many studies incorporating AI and how to overcome them. (A) Many studies use rather small cohorts, lack external validation, and do not compare methods with published approaches, hindering clinical implementation. (B) Collaborative efforts are needed to build multicentre cohorts to develop reliable AI models that could be further validated in (pre)clinical trials. (C) Selecting the right evaluation metrics is important; for example, the precision-recall curve is better for imbalanced data as it highlights positive class performance and is not influenced by many TN. FN, false negatives; FP, false positives; TN, true negatives; TP, true positives.