Abstract

Background & Aims Annotation of liver biopsies, for disease staging is increasingly aided by digital pathology, however existing systems do not quantify inflammation and steatosis within an anatomical framework. We developed an AI system to quantify portal tracts (PT) and disease features and their regional distribution in Metabolic Dysfunction-Associated Steatotic Liver Disease (MASLD)/Metabolic Dysfunction-Associated Steatohepatitis (MASH) and Autoimmune Hepatitis (AIH).

Methods Digitised images of haematoxylin and eosin-stained specimens were pooled from 4 clinical cohorts (n=390: 89 MASLD, 238 MASH, 63 AIH). Portal tracts, regional steatosis, interface hepatitis, portal and lobular inflammation were quantified using a proprietary AI system and scored by expert pathologists.

Results The percentage of steatosis was higher in MASH (7.5%) compared to MASLD (3.2%, p<0.0001). Lobular regions had larger steatotic vesicles (260 vs 190mm2, p<0.0001). AI-derived steatosis quantification correlated with manual grading (rs=0.72; p<0.0001). The inflammatory cell number (ICN) was 2-fold higher in AIH compared to MASLD/MASH in interface [390 vs 140; p<0.0001], portal [4600 vs 1500], and lobular [1500 vs 650] regions. Severity of portal inflammation from manual grading correlated well with ICN count at PT (rs=0.71; p<0.0001) but not lobular regions (rs=<0.1). For equivalent grades of portal inflammation, the ICN was up to 3-fold higher in AIH than in MASLD/MASH (rs=0.71; p<0.0001).

Conclusion Despite equivalent pathologist portal inflammation grades, the digital burden of inflammation was significantly higher in AIH than MASLD/MASH. This digital system quantifies PT, inflammation, and steatosis, providing powerful decision support for pathologists in AIH diagnosis and MASH staging.

LAY SUMMARY The detection of portal tracts is an important part of liver biopsy sample quality control and histological scoring. Using artificial intelligence, this study demonstrates a system that automatically detects and quantifies portal tracts and surrounding patterns of inflammation and steatosis. AI found that inflammation was in similar regions but was higher in autoimmune hepatitis than in metabolic dysfunction-associated steatohepatitis, despite similar grading from manual scoring. This AI system provides granular information that can aid biopsy grading and provide insights into liver disease progression and diagnosis.

Competing Interest Statement

DW, AM, CB, HTB, SL, KH, PA, RK, PW and CL are employees at Perspectum Ltd. KF, EF, TK and RG are consultants for Perspectum Ltd. All other co-authors have no conflicts of interest to declare relevant to this work. Manuscript includes use of data generated via collaboration with: Yokohama City University Hospital, CymaBay Therapeutics Inc (acquired by Gilead Sciences), University of Birmingham and Childrens Memorial Health Institute in Warsaw (IPCZD). Two trials of which data was utilised for this study were sponsored by Perspectum Ltd (Study 3&4).

Clinical Trial

NCT03551522, UMIN000026145, ISRCTN39463479, NCT03198104

Funding Statement

This paper presents data from independent research funded by Innovate UK (Project number: 101679) and the Eureka Eurostars 2 Grant (E!10124).

Author Declarations

I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained.

Yes

The details of the IRB/oversight body that provided approval or exemption for the research described are given below:

Study 1 was registered with registered with clinicaltrials.gov (NCT03551522). Study 2 was registered with UMIN clinical trials registry (UMIN000026145). Study 3 was registered with the ISRCTN registry (ISCRTN39463479) and the National Institute for Health Research portfolio (15912). The collection of study 3 pathology data was funded by Innovate UK (Project number: 101679). Study 4 was registered with clinicaltrials.gov (NCT03198104) and funded by the Eureka Eurostars 2 Grant (E!10124). All parent studies were conducted in accordance with the ethical principles of the Declaration of Helsinki 2013 and the Good Clinical Practice guidelines and all studies approved by ethical committees (Institutional review at trial sites for NCT03551522, Ethics committee at Yokohama City Hospital for UMIN000026145, West Midlands – Black Country Research Ethics Committee 14/WM/0010 for ISRCTN39463479, Komisja Bioetyczna przy Instytucie Pomnik-Centrum Zdrowia Dziecka 11/KBE/2016 for NCT03198104).

I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals.

Yes

I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance).

Yes

I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable.

Yes

List of abbreviationsAIArtificial IntelligenceAIHAutoimmune HepatitisDLDeep LearningH&EHaematoxylin and EosinIBInflammatory BurdenICNInflammatory Cell NumberSFSteatosis FractionMASLDMetabolic Dysfunction-associated Steatotic Liver DiseaseMASHMetabolic Dysfunction-associated SteatohepatitisMLMachine LearningMRIMagnetic Resonance ImagingNASNAFLD Activity ScorePDFFProton Density Fat-FractionPTPortal Tract(s)WSIWhole Slide Image(s)