Albrecht, ThomasThomasAlbrechtRossberg, AnnikAnnikRossbergAlbrecht, Jana DorotheaJana DorotheaAlbrechtNicolay, Jan PeterJan PeterNicolayStraub, Beate KatharinaBeate KatharinaStraubGerber, Tiemo SvenTiemo SvenGerberAlbrecht, MichaelMichaelAlbrechtBrinkmann, FritzFritzBrinkmannCharbel, AlphonseAlphonseCharbelSchwab, ConstantinConstantinSchwabSchreck, JohannesJohannesSchreckBrobeil, AlexanderAlexanderBrobeilFlechtenmacher, ChristaChristaFlechtenmachervon Winterfeld, MoritzMoritzvon WinterfeldKöhler, Bruno ChristianBruno ChristianKöhlerSpringfeld, ChristophChristophSpringfeldMehrabi, ArianebArianebMehrabiSinger, StephanStephanSingerVogel, Monika NadjaMonika NadjaVogelNeumann, OlafOlafNeumannStenzinger, AlbrechtAlbrechtStenzingerSchirmacher, PeterPeterSchirmacherWeis, Cleo-AronCleo-AronWeisRoessler, StephanieStephanieRoesslerKather, Jakob NikolasJakob NikolasKatherGoeppert, Frank BenjaminFrank BenjaminGoeppert2024-10-252024-10-252023-11https://boris-portal.unibe.ch/handle/20.500.12422/169207BACKGROUND & AIMS Diagnosis of adenocarcinoma in the liver is a frequent scenario in routine pathology and has a critical impact on clinical decision-making. However, rendering a correct diagnosis can be challenging and often requires the integration of clinical, radiological, and immunohistochemical information. We present a deep learning model (HEPNET) to distinguish intrahepatic cholangiocarcinoma (iCCA) from colorectal liver metastasis (CRM) as the most frequent primary and secondary forms of liver adenocarcinoma with clinical-grade accuracy using hematoxylin and eosin-stained whole-slide images. METHODS HEPNET was trained on 714 589 image tiles from 456 patients who were randomly selected in a stratified manner from a pool of 571 patients who underwent surgical resection or biopsy at Heidelberg University Hospital. Model performance was evaluated on a hold-out internal test set comprising 115 patients and externally validated on 159 patients recruited at Mainz University Hospital. RESULTS On the hold-out internal test set, HEPNET achieved an area under the receiver operating characteristic curve (AUROC) of 0.994 (95% CI 0.989-1.000) and an accuracy of 96.522% (95% CI 94.521-98.694%) at the patient level. Validation on the external test set yielded an AUROC of 0.997 (95% CI 0.995-1.000), corresponding to an accuracy of 98.113% (95% CI 96.907-100.000%). HEPNET surpassed the performance of six pathology experts with different levels of experience in a reader study of 50 patients (P=.0005), boosted the performance of resident pathologists to the level of senior pathologists, and reduced potential downstream analyses. CONCLUSION Here, we provide a ready-to-use tool with a clinical-grade performance that may facilitate routine pathology by rendering a definitive diagnosis and guiding ancillary testing. The incorporation of HEPNET into pathology laboratories may optimize the diagnostic workflow, complemented by test-related labor and cost savings.enDigital pathology artificial intelligence biliary tract cancer intestinal cancer500 - Science::570 - Life sciences; biology600 - Technology::610 - Medicine & healtharticle10.48350/1853803756265710.1053/j.gastro.2023.07.026