000180070 001__ 180070
000180070 005__ 20240229145604.0
000180070 0247_ $$2doi$$a10.1016/j.media.2022.102474
000180070 0247_ $$2pmid$$apmid:35588568
000180070 0247_ $$2ISSN$$a1361-8415
000180070 0247_ $$2ISSN$$a1361-8423
000180070 0247_ $$2ISSN$$a1361-8431
000180070 0247_ $$2altmetric$$aaltmetric:127726563
000180070 037__ $$aDKFZ-2022-01069
000180070 041__ $$aEnglish
000180070 082__ $$a610
000180070 1001_ $$aGhaffari Laleh, Narmin$$b0
000180070 245__ $$aBenchmarking weakly-supervised deep learning pipelines for whole slide classification in computational pathology.
000180070 260__ $$aAmsterdam [u.a.]$$bElsevier Science$$c2022
000180070 3367_ $$2DRIVER$$aarticle
000180070 3367_ $$2DataCite$$aOutput Types/Journal article
000180070 3367_ $$0PUB:(DE-HGF)16$$2PUB:(DE-HGF)$$aJournal Article$$bjournal$$mjournal$$s1653654265_14779
000180070 3367_ $$2BibTeX$$aARTICLE
000180070 3367_ $$2ORCID$$aJOURNAL_ARTICLE
000180070 3367_ $$00$$2EndNote$$aJournal Article
000180070 520__ $$aArtificial intelligence (AI) can extract visual information from histopathological slides and yield biological insight and clinical biomarkers. Whole slide images are cut into thousands of tiles and classification problems are often weakly-supervised: the ground truth is only known for the slide, not for every single tile. In classical weakly-supervised analysis pipelines, all tiles inherit the slide label while in multiple-instance learning (MIL), only bags of tiles inherit the label. However, it is still unclear how these widely used but markedly different approaches perform relative to each other. We implemented and systematically compared six methods in six clinically relevant end-to-end prediction tasks using data from N=2980 patients for training with rigorous external validation. We tested three classical weakly-supervised approaches with convolutional neural networks and vision transformers (ViT) and three MIL-based approaches with and without an additional attention module. Our results empirically demonstrate that histological tumor subtyping of renal cell carcinoma is an easy task in which all approaches achieve an area under the receiver operating curve (AUROC) of above 0.9. In contrast, we report significant performance differences for clinically relevant tasks of mutation prediction in colorectal, gastric, and bladder cancer. In these mutation prediction tasks, classical weakly-supervised workflows outperformed MIL-based weakly-supervised methods for mutation prediction, which is surprising given their simplicity. This shows that new end-to-end image analysis pipelines in computational pathology should be compared to classical weakly-supervised methods. Also, these findings motivate the development of new methods which combine the elegant assumptions of MIL with the empirically observed higher performance of classical weakly-supervised approaches. We make all source codes publicly available at https://github.com/KatherLab/HIA, allowing easy application of all methods to any similar task.
000180070 536__ $$0G:(DE-HGF)POF4-313$$a313 - Krebsrisikofaktoren und Prävention (POF4-313)$$cPOF4-313$$fPOF IV$$x0
000180070 588__ $$aDataset connected to CrossRef, PubMed, , Journals: inrepo02.dkfz.de
000180070 650_7 $$2Other$$aArtificial intelligence
000180070 650_7 $$2Other$$aComputational pathology
000180070 650_7 $$2Other$$aConvolutional neural networks
000180070 650_7 $$2Other$$aMultiple-Instance Learning
000180070 650_7 $$2Other$$aVision transformers
000180070 650_7 $$2Other$$aWeakly-supervised deep learning
000180070 7001_ $$aMuti, Hannah Sophie$$b1
000180070 7001_ $$aLoeffler, Chiara Maria Lavinia$$b2
000180070 7001_ $$aEchle, Amelie$$b3
000180070 7001_ $$aSaldanha, Oliver Lester$$b4
000180070 7001_ $$aMahmood, Faisal$$b5
000180070 7001_ $$aLu, Ming Y$$b6
000180070 7001_ $$aTrautwein, Christian$$b7
000180070 7001_ $$aLanger, Rupert$$b8
000180070 7001_ $$aDislich, Bastian$$b9
000180070 7001_ $$aBuelow, Roman D$$b10
000180070 7001_ $$aGrabsch, Heike Irmgard$$b11
000180070 7001_ $$0P:(DE-He78)90d5535ff896e70eed81f4a4f6f22ae2$$aBrenner, Hermann$$b12$$udkfz
000180070 7001_ $$0P:(DE-He78)c259d6cc99edf5c7bc7ce22c7f87c253$$aChang-Claude, Jenny$$b13$$udkfz
000180070 7001_ $$0P:(DE-He78)9b2a61b2abe4a64ca23b6783b7c4fe63$$aAlwers, Elizabeth$$b14$$udkfz
000180070 7001_ $$0P:(DE-He78)1e33961c8780aca9b76d776d1fdc1ebb$$aBrinker, Titus J$$b15$$udkfz
000180070 7001_ $$aKhader, Firas$$b16
000180070 7001_ $$aTruhn, Daniel$$b17
000180070 7001_ $$aGaisa, Nadine T$$b18
000180070 7001_ $$aBoor, Peter$$b19
000180070 7001_ $$0P:(DE-He78)6c5d058b7552d071a7fa4c5e943fff0f$$aHoffmeister, Michael$$b20$$udkfz
000180070 7001_ $$aSchulz, Volkmar$$b21
000180070 7001_ $$aKather, Jakob Nikolas$$b22
000180070 773__ $$0PERI:(DE-600)1497450-2$$a10.1016/j.media.2022.102474$$gVol. 79, p. 102474 -$$p102474$$tMedical image analysis$$v79$$x1361-8415$$y2022
000180070 909CO $$ooai:inrepo02.dkfz.de:180070$$pVDB
000180070 9101_ $$0I:(DE-588b)2036810-0$$6P:(DE-He78)90d5535ff896e70eed81f4a4f6f22ae2$$aDeutsches Krebsforschungszentrum$$b12$$kDKFZ
000180070 9101_ $$0I:(DE-588b)2036810-0$$6P:(DE-He78)c259d6cc99edf5c7bc7ce22c7f87c253$$aDeutsches Krebsforschungszentrum$$b13$$kDKFZ
000180070 9101_ $$0I:(DE-588b)2036810-0$$6P:(DE-He78)9b2a61b2abe4a64ca23b6783b7c4fe63$$aDeutsches Krebsforschungszentrum$$b14$$kDKFZ
000180070 9101_ $$0I:(DE-588b)2036810-0$$6P:(DE-He78)1e33961c8780aca9b76d776d1fdc1ebb$$aDeutsches Krebsforschungszentrum$$b15$$kDKFZ
000180070 9101_ $$0I:(DE-588b)2036810-0$$6P:(DE-He78)6c5d058b7552d071a7fa4c5e943fff0f$$aDeutsches Krebsforschungszentrum$$b20$$kDKFZ
000180070 9131_ $$0G:(DE-HGF)POF4-313$$1G:(DE-HGF)POF4-310$$2G:(DE-HGF)POF4-300$$3G:(DE-HGF)POF4$$4G:(DE-HGF)POF$$aDE-HGF$$bGesundheit$$lKrebsforschung$$vKrebsrisikofaktoren und Prävention$$x0
000180070 9141_ $$y2022
000180070 915__ $$0StatID:(DE-HGF)0160$$2StatID$$aDBCoverage$$bEssential Science Indicators$$d2021-01-28
000180070 915__ $$0StatID:(DE-HGF)0113$$2StatID$$aWoS$$bScience Citation Index Expanded$$d2021-01-28
000180070 915__ $$0StatID:(DE-HGF)0100$$2StatID$$aJCR$$bMED IMAGE ANAL : 2021$$d2022-11-18
000180070 915__ $$0StatID:(DE-HGF)0200$$2StatID$$aDBCoverage$$bSCOPUS$$d2022-11-18
000180070 915__ $$0StatID:(DE-HGF)0300$$2StatID$$aDBCoverage$$bMedline$$d2022-11-18
000180070 915__ $$0StatID:(DE-HGF)0600$$2StatID$$aDBCoverage$$bEbsco Academic Search$$d2022-11-18
000180070 915__ $$0StatID:(DE-HGF)0030$$2StatID$$aPeer Review$$bASC$$d2022-11-18
000180070 915__ $$0StatID:(DE-HGF)0199$$2StatID$$aDBCoverage$$bClarivate Analytics Master Journal List$$d2022-11-18
000180070 915__ $$0StatID:(DE-HGF)1160$$2StatID$$aDBCoverage$$bCurrent Contents - Engineering, Computing and Technology$$d2022-11-18
000180070 915__ $$0StatID:(DE-HGF)0150$$2StatID$$aDBCoverage$$bWeb of Science Core Collection$$d2022-11-18
000180070 915__ $$0StatID:(DE-HGF)9910$$2StatID$$aIF >= 10$$bMED IMAGE ANAL : 2021$$d2022-11-18
000180070 9201_ $$0I:(DE-He78)C070-20160331$$kC070$$lC070 Klinische Epidemiologie und Alternf.$$x0
000180070 9201_ $$0I:(DE-He78)C120-20160331$$kC120$$lPräventive Onkologie$$x1
000180070 9201_ $$0I:(DE-He78)HD01-20160331$$kHD01$$lDKTK HD zentral$$x2
000180070 9201_ $$0I:(DE-He78)C020-20160331$$kC020$$lC020 Epidemiologie von Krebs$$x3
000180070 9201_ $$0I:(DE-He78)C140-20160331$$kC140$$lNWG Digitale Biomarker in der Onkologie$$x4
000180070 980__ $$ajournal
000180070 980__ $$aVDB
000180070 980__ $$aI:(DE-He78)C070-20160331
000180070 980__ $$aI:(DE-He78)C120-20160331
000180070 980__ $$aI:(DE-He78)HD01-20160331
000180070 980__ $$aI:(DE-He78)C020-20160331
000180070 980__ $$aI:(DE-He78)C140-20160331
000180070 980__ $$aUNRESTRICTED