Artificial intelligence in liver diseases: Improving diagnostics, prognostics and response prediction

Abstract

Clinical routine in hepatology involves the diagnosis and treatment of a wide spectrum of metabolic, infectious, autoimmune and neoplastic diseases. Clinicians integrate qualitative and quantitative information from multiple data sources to make a diagnosis, prognosticate the disease course, and recommend a treatment. In the last 5 years, advances in artificial intelligence (AI), particularly in deep learning, have made it possible to extract clinically relevant information from complex and diverse clinical datasets. In particular, histopathology and radiology image data contain diagnostic, prognostic and predictive information which AI can extract. Ultimately, such AI systems could be implemented in clinical routine as decision support tools. However, in the context of hepatology, this requires further large-scale clinical validation and regulatory approval. Herein, we summarise the state of the art in AI in hepatology with a particular focus on histopathology and radiology data. We present a roadmap for the further development of novel biomarkers in hepatology and outline critical obstacles which need to be overcome.

Keywords: AI, artificial intelligence; Artificial intelligence; CNN, convolutional neural network; DICOM, Digital Imaging and Communications in Medicine; HCC, hepatocellular carcinoma; ML, machine learning; MVI, microvascular invasion; NAFLD, non-alcoholic fatty liver disease; NASH, non-alcoholic steatohepatitis; TACE, transarterial chemoembolisation; TRIPOD, Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis; WSIs, whole slide images; deep learning; diagnostic support system; imaging; machine learning; multimodal data integration.

Fig. 1

Digital pathology and radiology using artificial intelligence for management of liver diseases. (A) Number of studies by country of the first author. (B) Number of studies by prediction of the models. (C) Number of studies by liver disease. (D) Number of studies stratified by the clinical input data used. Raw data for this figure is available in Tables S1 and S2. Methodological details are available in the supplementary materials and methods. (E) Cumulative number of published original studies per half-year from 2010 to mid-2021. (F) Cumulative number of published original studies per half-year by research field. (G) Cumulative number of published original studies per half-year by either deep learning or handcrafted feature extraction. CCA, cholangiocarcinoma; CLD, chronic liver disease; HCC, hepatocellular carcinoma; NAFLD, non-alcoholic fatty liver disease; NASH, non-alcoholic steatohepatitis.

Fig. 2

Radiology image analysis workflows. (A) Handcrafted feature extraction, also referred to as “radiomics,” is an established image analysis technique in radiology image analysis. Alternatively, deep learning, in the form of neural networks, can be used to learn features and predict a target label in a supervised fashion (end-to-end analysis). (B) Common tasks in radiology image analysis are segmentation, classification and prognostication.

Fig. 3

Data types in hepatology and multimodal learning. Inner area: data types routinely used for clinical decision making in hepatology. Blue circle: type of digital data. Yellow circle: suitable machine learning methods to analyse this data. NLP, natural language processing. Icon source: openmoji.org (CC-BY-SA 4.0 license).