Artificial Intelligence Applications in Hepatology

Abstract

Over the past 2 decades, the field of hepatology has witnessed major developments in diagnostic tools, prognostic models, and treatment options making it one of the most complex medical subspecialties. Through artificial intelligence (AI) and machine learning, computers are now able to learn from complex and diverse clinical datasets to solve real-world medical problems with performance that surpasses that of physicians in certain areas. AI algorithms are currently being implemented in liver imaging, interpretation of liver histopathology, noninvasive tests, prediction models, and more. In this review, we provide a summary of the state of AI in hepatology and discuss current challenges for large-scale implementation including some ethical aspects. We emphasize to the readers that most AI-based algorithms that are discussed in this review are still considered in early development and their utility and impact on patient outcomes still need to be assessed in future large-scale and inclusive studies. Our vision is that the use of AI in hepatology will enhance physician performance, decrease the burden and time spent on documentation, and reestablish the personalized patient-physician relationship that is of utmost importance for obtaining good outcomes. Over the past 2 decades, the field of hepatology has witnessed major developments in diagnostic tools, prognostic models, and treatment options making it one of the most complex medical subspecialties. Through artificial intelligence (AI) and machine learning, computers are now able to learn from complex and diverse clinical datasets to solve real-world medical problems with performance that surpasses that of physicians in certain areas. AI algorithms are currently being implemented in liver imaging, interpretation of liver histopathology, noninvasive tests, prediction models, and more. In this review, we provide a summary of the state of AI in hepatology and discuss current challenges for large-scale implementation including some ethical aspects. We emphasize to the readers that most AI-based algorithms that are discussed in this review are still considered in early development and their utility and impact on patient outcomes still need to be assessed in future large-scale and inclusive studies. Our vision is that the use of AI in hepatology will enhance physician performance, decrease the burden and time spent on documentation, and reestablish the personalized patient-physician relationship that is of utmost importance for obtaining good outcomes. Technological advancements have created the unique opportunity to use artificial intelligence (AI) and more specifically machine learning (ML) in clinical medicine. With available computational power, AI has the potential to transform patient care without losing the patient-centric, physician-guided approach of traditional clinical medicine. This has become even more evident during the COVID-19 pandemic, which provided unprecedented advancements in technology acceptance and availability in all areas of society and in particular in the health care system. Although AI is considered the overarching term that details the rational exploitation of data by a machine, ML more specifically describes the building of models that learn from available data to improve the prediction or performance related to a specific task without actually programming.1Mintz Y. Brodie R. Introduction to artificial intelligence in medicine.Minim Invasive Ther Allied Technol. 2019; 28: 73-81Crossref PubMed Scopus (160) Google Scholar The ability of ML algorithms to predict outcomes can be exploited based on labeled (supervised) or unlabeled (unsupervised) data. By using the ML algorithm over time and providing more training, the desired output becomes progressively more accurate. Deep learning (DL) refines and narrows ML by using multiple neuronal networks that mimic the human neurologic system to analyze, identify, and learn from complex datasets.2Esteva A. Robicquet A. Ramsundar B. et al.A guide to deep learning in healthcare.Nat Med. 2019; 25: 24-29Crossref PubMed Scopus (1343) Google Scholar The neural networks are organized in multiple layers where the signal travels from the first layer (input) to the last layer (output) after going through multiple intervening layers. A few important issues have to be considered when aiming to implement AI in the clinical environment today (Figure 1). Beyond investments in technology in the health care sector, the quality of the data that are used to develop algorithms and predict outcome is most critical. In the field of hepatology research, several large prospective studies that are aimed to explore outcome are actively recruiting and will provide the quality and robustness of the data that are required.3Sanyal A.J. Shankar S.S. Calle R.A. et al.Non-invasive biomarkers of nonalcoholic steatohepatitis: the FNIH NIMBLE project.Nat Med. 2022; 28: 430-432Crossref PubMed Scopus (12) Google Scholar,4Hardy T. Wonders K. Younes R. et al.The European NAFLD Registry: a real-world longitudinal cohort study of nonalcoholic fatty liver disease.Contemp Clin Trials. 2020; 98106175Crossref PubMed Scopus (43) Google Scholar The enormous potential to account for a large number of variables in complex databases and determine the likelihood of specific outcomes in a very short time, will markedly outperform a single physician’s capability that operates at the level of personal experience and medical education. Despite high expectations by many stakeholders and receptivity toward AI in the general society and among medical professionals, the level of implementation in clinical practice today is relatively low.5Xiang Y. Zhao L. Liu Z. et al.Implementation of artificial intelligence in medicine: status analysis and development suggestions.Artif Intell Med. 2020; 102101780https://doi.org/10.1016/j.artmed.2019.101780Crossref PubMed Scopus (33) Google Scholar There are several limitations that must be taken into account to allow for a safe application of AI and a higher penetration into clinical care. The assembly of high-quality representative data sets that eliminate unwanted and unconscious biases is a prerequisite for building ML models that do not perpetuate health care disparities. The inability of AI algorithms to account for information gained from a direct patient-physician interaction is an inherent limitation. The AI algorithms will never be able to substitute for a physician’s direct interaction with their patients. We view AI as a complementary tool to significantly enhance patient-provider interaction and patient care. For AI’s integration into hepatology clinical practice, multiple currently open questions need to be addressed including the quality of data synthesized, operational procedures, data and systems safety, and ethical challenges. Ethical challenges may arise from clinical decision making based on AI-generated diagnostic algorithms that are not readily recapitulated through medical reasoning. As seen with automated driving, a critical question arises around responsibility and liability in the context of a decision that is based on an AI algorithm. Similarly in medicine, the consequences of false-positive and false-negative results that are generated by AI are far reaching for patients and their providers. This is exacerbated by the fact that AI algorithms are comprised of complex interconnected structures with numerous parameters and a “black box” nature, offering little understanding of their inner working. Explainable AI is a set of processes that allows humans to comprehend the output created by ML algorithms, which help develop trust in the system and meet adherence to regulatory requirements.6Nazir S. Dickson D.M. Akram M.U. Survey of explainable artificial intelligence techniques for biomedical imaging with deep neural networks.Comput Biol Med. 2023; 156106668Crossref PubMed Scopus (3) Google Scholar The challenges in hepatology are to allow for a safe and evidence-based implementation of AI to support clinical decision making. Several AI approaches have been used in hepatology with a focus on identification of cases (through imaging and noninvasive tests [NIT]), augmentation of histologic analysis, and prediction of outcome (Table 1). This review article focuses on current developments in AI/ML with potential applications in hepatology and defines areas of research that should be addressed in the future. To prepare for this manuscript, a literature search by the using the PubMed and the search and The the most to provide an of where AI may impact the practice of clinical of AI in application in of AI to based on for automated and liver in data by and imaging and algorithms to of histologic algorithms to predict to algorithms to determine the of and predict of for of data data and of patients using ML to predict the of and patients with to need for high-quality representative datasets to eliminate the potential for to predict models to predict from liver ML to predict tool to predict outcomes in patients with of prospective AI-based clinical that the of AI models in clinical outcomes for artificial neural machine nonalcoholic noninvasive in a artificial neural machine nonalcoholic noninvasive a in the of liver The ability to liver and in to or as is Although is the technology in imaging imaging to to liver and and imaging to allow for of With a of liver is and the of more through liver has data to T. et by future and in PubMed Scopus Google Scholar The of data that are from imaging for using a of the available data is actually used to an that the clinical decision making. available research data generated on the use of AI in of liver neural networks have a very high of the of based on A. et al.Implementation of deep learning algorithm for fatty liver in PubMed Scopus Google Scholar In a more analysis, of data an the of when with the in even in the of to imaging A. et of nonalcoholic fatty liver and of liver with data using neural 2020; PubMed Scopus Google Scholar of or by in a study that used from cases at the of using a the of from to for the histologic by the of a of and the of et liver from Biol Google Scholar for the of a deep an of for from liver et diagnostic performance for liver using deep a 2020; Scopus Google Scholar this to the diagnostic and of with and to et diagnostic performance for liver using deep a 2020; Scopus Google Scholar The of AI and ML in diagnostic for liver do not the inherent limitations that specific imaging the use of imaging to not a high for of on diagnostic of identification of areas of fatty PubMed Scopus Google Scholar In a analysis using a automated algorithm and with as the to and of and et on using a automated deep learning in using as a PubMed Scopus Google Scholar even and imaging the to of The more imaging outperform in the diagnostic of and of is an automated that for imaging In a and an with the histologic in of nonalcoholic fatty liver A. et to and in liver 2022; PubMed Scopus (12) Google Scholar In a study on parameters on with to early K. A. S. et of analysis and machine learning for of liver in with and 2020; PubMed Scopus Google Scholar a automated liver even and for automated liver K. A. T. et and liver and using a neural 2019; Scopus Google Scholar of in imaging is the augmentation of to of liver health and even the to the is not in the context of liver The ability of to in the of the liver with the of on liver R. A. et on in patients with 2022; PubMed Scopus (3) Google Scholar one clinical application where AI can in the of an imaging will be to the of to the to the of in the T. imaging to liver an need in the NAFLD 2022; PubMed Scopus Google Scholar Despite the results of AI algorithms in liver imaging, several issues their use including in data by imaging and to the of et intelligence in liver and 2022; Google Scholar Several need to be taken to the and analysis of imaging The details of algorithm development including the datasets and should be to into the clinical is still considered the for nonalcoholic and the of the histologic of and by the has been to be and to major and S. D.M. K. et of on and on interpretation of nonalcoholic fatty liver PubMed Scopus Google Scholar In the and provided a of including the of of 2 histologic as for of without of or of by or more without of A.J. et and clinical for nonalcoholic PubMed Scopus Google Scholar the on histologic has it to patients for and to to with from clinical the of among et of liver has for clinical 2020; PubMed Scopus Google Scholar This has created an opportunity to use AI/ML algorithms to develop for of the histologic of that are to by the algorithm on that are by (Figure In et S. et and of in 2020; PubMed Scopus Google Scholar one of the first studies on using ML to in liver as a to provide of The ML algorithm with when with by 2 The an AI-based to liver and determine in patients with good to the generated and the with a of from to S. et and of in 2020; PubMed Scopus Google Scholar used data from patients with to develop a of and with high the and the from for to for R. et machine of and in from patients with nonalcoholic fatty liver 2020; PubMed Google Scholar In the study to the used liver from large clinical to and a deep to of histologic A. L. et al.A machine learning approach of liver and in PubMed Scopus Google Scholar The ML parameters clinical as to and and by of more complex as and the of to a Deep to in from to the of treatment and with noninvasive as the liver and liver by The at a ML to predict by using and from a The a with traditional by and able to patients with with good of and in the and et al.A machine learning approach to liver in PubMed Scopus (12) Google Scholar The of the is the fact that can be from a liver without the need for and human to In a of at et histologic ML in a analysis of patients from the This provided a cohort of patients with with liver and data including and clinical outcomes. This analysis of and A tool provided an automated of histologic related to related to related to and related to The a ML which significantly with as a for and for for and significantly of for and for created 2 which significantly the of of for and of for and which patients in with an of data a that ML can be to liver to important data that are to from et of machine learning histologic with and the development of in patients with 2022; Google Z. et learning liver with in nonalcoholic 2023; PubMed Scopus Google Scholar several AI/ML have been to liver in the context of in the to provide of histologic on liver automated from and analysis using on histologic of the integration of AI in the interpretation of the in of and will be important to have of in of data data and of that will future of ML algorithms have been generated that can on liver and provide for of after T. et intelligence for the and clinical of 2022; PubMed Scopus Google Scholar For et S. K. et deep learning for the of liver Biol Med. PubMed Scopus Google Scholar a to on liver histopathology, which results with traditional important tool for is of by to and treatment et Y. S. et al.A for in Scopus Google Scholar a neural for that high in of patients with experience at et B. et after using deep learning on 2020; PubMed Scopus Google Scholar 2 models based on for after and performance of the models in with that used and et R. A. et learning of from PubMed Scopus Google Scholar and a system from that patients into and to their are the of clinical decision making in many patients with liver The interpretation of can be complex based on clinical and experience and and augmentation of the interpretation of by AI and ML has been in several studies. A study from the used ML to and based on used variables including direct and et learning techniques from in patients with and liver Clin 2022; PubMed Scopus Google Scholar of and cases of the ML algorithm an performance with an of an of physicians using the variables in an diagnostic with an of the of et learning techniques from in patients with and liver Clin 2022; PubMed Scopus Google Scholar A approach taken to improve the of identification of patients with NAFLD and AI algorithm that clinical data and studies available in health care is the algorithm. using the of and and on patients with and in the using an of The performance of this to predict cases with histologic good with an of et of a machine learning to predict of nonalcoholic 28: PubMed Scopus Google Scholar the number of variables to using and the to predict still et of a machine learning to predict of nonalcoholic 28: PubMed Scopus Google Scholar In a analysis from the of NAFLD from cases by using clinical variables including and the in the cohort Through of clinical data with and a analysis that aimed to liver on in used and and an of A. et and the of fatty liver a machine learning and study in the Med. 2020; PubMed Scopus Google Scholar ML models including and artificial neural implemented to predict the histologic of and the of using in patients with The ML models traditional for et learning models are to noninvasive tests in of NAFLD and 2023; PubMed Scopus (3) Google Scholar with the for the by and the for and being and The ML readily available clinical as and the of This can be implemented in the system to patients with that may from In the as as the in with in and and algorithms are not to substitute for a clinical will allow to cases that are not as of liver and be in care to support the of cases (Figure of the challenges in is the identification of patients with cases are to using clinical variables and In a large cohort with in a clinical a including and patients with and with an of et and of the nonalcoholic fatty liver to from Med. 2022; PubMed Scopus (3) Google Scholar approach that the to patients with based on that are on The in patients that liver at the and The from with an of based on an analysis of of approaches to the development of in the care of patients when hepatology not be et of the an deep learning in 2022; PubMed Scopus Google Scholar A approach aiming at liver in a care by implementation of automated analysis in a system. The liver study of when liver et liver a of automated and of liver in 2019; PubMed Scopus Google Scholar Through this of liver by with In an automated and in a implementation in the through et liver a of automated and of liver in 2019; PubMed Scopus Google Scholar of by AI is T. et intelligence for the and clinical of 2022; PubMed Scopus Google Scholar In a large-scale analysis using clinical data and in patients with in the the ability of neural models to predict an of et of a deep learning to predict in patients with 2020; Scopus Google Scholar this neural used variables and longitudinal et of a deep learning to predict in patients with 2020; Scopus Google Scholar Several traditional have been to in with the for being the most et and among patients on the Med. PubMed Scopus Google Scholar to predict which may be short of a for most patients with for for understanding of can patients and and of care and improve decision making for ML models have potential to enhance the and improve with traditional models their have their et et and of a machine learning to predict 2020; Scopus Google Scholar used a approach that ML algorithms in a large cohort from to predict and a the that used variables that implemented in an for and with the for being for and for are in patients with and major for and the health care system making prediction of patients at for from a high for patients with PubMed Google Scholar a study that AI models including and support machine to predict with of and K. et of and using machine learning in PubMed Scopus Google Scholar the fact that in are by the complex of and that may not be by the data. In the the use of AI to data from and variables may help improve and et of and at with development in patients with 2022; 28: PubMed Scopus Google Scholar the number of patients in need for liver and the of is an need to and to in of the of liver PubMed Scopus Google Scholar models, as and the outcomes after liver do not the complex among and the that are to A. The a of PubMed Scopus Google Scholar is an opportunity for AI/ML techniques to improve based on and variables that are at the time of et L. Y. B. et algorithms predict after liver PubMed Scopus Google Scholar ML models and to predict from in a that and available and their ML models with traditional liver and that their to and outcomes after liver of with and The of in the and the can be considered to provide of that the ML be used to support decision making in a review the use of ML to predict outcome liver S. et artificial intelligence for of in liver a 2020; PubMed Scopus Google Scholar studies the and outcomes from liver with being the most used ML Despite are several challenges to ML models for the of when the algorithm including variables that are not considered or the of and the fact that the ML algorithms for outcomes on the from which and may not as on a AI has the potential to transform the of clinical care by providing and to treatment For more on the of AI in we readers to an review by et T. et intelligence for the and clinical of 2022; PubMed Scopus Google Scholar Several studies have AI techniques to longitudinal data to predict in including an ML algorithm that of which significantly the for A. et learning algorithms outperform models in development of PubMed Scopus Google Scholar where ML algorithm may improve care the of and to et et by artificial neural networks for patients after PubMed Scopus Google Scholar an that clinical variables to predict and of patients with and good with of an to predict patient at after for with performance of A. et after for using a neural a 2020; PubMed Scopus Google Scholar is a liver that is with and higher for and L. A. et practice on and 2023; PubMed Scopus Google Scholar the of the and the over a short of time, models that can the of clinical is a high et for clinical in review and results from an PubMed Scopus Google Scholar The an ML to predict or in 2 large of patients with et outcomes of the a and study using machine 2020; PubMed Scopus Google Scholar clinical variables the of and the of in and able to predict with traditional of with for and for ML to a large of patients with to predict the outcome of or A. et learning in a approach for 2022; PubMed Scopus Google Scholar ML of patients with with and and that of with is with in the which can in to and the development of et and of practice by the for the of PubMed Scopus Google Scholar early is and can be by using AI/ML approaches to large In a study ML algorithms in the and which of cases and to determine which of and algorithm provided the performance in the of T. machine approach for at of 2020; PubMed Scopus Google Scholar The using and the and The for the and the traditional tool of liver is now the for liver in the A. et of liver in analysis of for liver and and PubMed Scopus Google Scholar and most patients are after K. et and by of liver in a study of 2019; PubMed Scopus Google Scholar the identification of at an provide the opportunity to through treatment and the of as and 2 et L. et biomarkers for liver Med. 2022; 28: PubMed Scopus Google Scholar the use of by ML to biomarkers for the and of early The ML models a of that with high of and clinical tests, as the and The of the development of major liver outcomes and study from used a AI approach to in with et intelligence in liver patients in 2022; PubMed Scopus Google Scholar The used clinical variables that available in a prospective cohort of of at for The AI models patients with liver more of and to Despite the potential for AI-based models to improve care for patients with liver several challenges need to be addressed models from research to There is a that algorithms in health care may and health and biases Z. B. et in an algorithm used to the health of 2019; PubMed Scopus Google Scholar Several biases can be through AI algorithms including biases related to the of the clinical research biases in data and and biases in algorithm development and A. et intelligence in and to clinical practice health 2022; PubMed Scopus Google Scholar To in AI use in health need to be to data and and clinical health early in the of algorithm development is of utmost In ethical and regulatory need to be by the and and as the of the and status of the patient used in the AI This will of certain and the of the The is the development of an intelligence approach that the physician decision making with the AI to eliminate and health A. et in machine learning to health Med. PubMed Scopus Google Scholar To implement AI in prospective AI-based clinical are to the of AI models in clinical outcomes for patients. In the several have the impact of algorithms on and during et system and a prospective 2019; PubMed Scopus Google Scholar In we that AI/ML models will an in the identification and of liver in the future. providing for physicians on the of AI and to the output of AI models is of In investments in health data safety, and ethical studies of AI are to the level of trust with physicians and patient to clinical decision making. review review review

Ähnliche Arbeiten