Ensuring medical AI safety: interpretability-driven detection and mitigation of spurious model behavior and associated data

Deep neural networks are increasingly employed in high-stakes medical applications, despite their tendency for shortcut learning in the presence of spurious correlations, which can have potentially fatal consequences in practice. Whereas a multitude of works address either the detection or mitigatio...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:Machine learning Ročník 114; číslo 9; s. 206
Hlavní autoři: Pahde, Frederik, Wiegand, Thomas, Lapuschkin, Sebastian, Samek, Wojciech
Médium: Journal Article
Jazyk:angličtina
Vydáno: New York Springer US 01.09.2025
Springer Nature B.V
Témata:
ISSN:0885-6125, 1573-0565
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Abstract Deep neural networks are increasingly employed in high-stakes medical applications, despite their tendency for shortcut learning in the presence of spurious correlations, which can have potentially fatal consequences in practice. Whereas a multitude of works address either the detection or mitigation of such shortcut behavior in isolation, the Reveal2Revise approach provides a comprehensive bias mitigation framework combining these steps. However, effectively addressing these biases often requires substantial labeling efforts from domain experts. In this work, we review the steps of the Reveal2Revise framework and enhance it with semi-automated interpretability-based bias annotation capabilities. This includes methods for the sample- and feature-level bias annotation, providing valuable information for bias mitigation methods to unlearn the undesired shortcut behavior. We show the applicability of the framework using four medical datasets across two modalities, featuring controlled and real-world spurious correlations caused by data artifacts. We successfully identify and mitigate these biases in VGG16, ResNet50, and contemporary Vision Transformer models, ultimately increasing their robustness and applicability for real-world medical tasks. Our code is available at https://github.com/frederikpahde/medical-ai-safety .
AbstractList Deep neural networks are increasingly employed in high-stakes medical applications, despite their tendency for shortcut learning in the presence of spurious correlations, which can have potentially fatal consequences in practice. Whereas a multitude of works address either the detection or mitigation of such shortcut behavior in isolation, the Reveal2Revise approach provides a comprehensive bias mitigation framework combining these steps. However, effectively addressing these biases often requires substantial labeling efforts from domain experts. In this work, we review the steps of the Reveal2Revise framework and enhance it with semi-automated interpretability-based bias annotation capabilities. This includes methods for the sample- and feature-level bias annotation, providing valuable information for bias mitigation methods to unlearn the undesired shortcut behavior. We show the applicability of the framework using four medical datasets across two modalities, featuring controlled and real-world spurious correlations caused by data artifacts. We successfully identify and mitigate these biases in VGG16, ResNet50, and contemporary Vision Transformer models, ultimately increasing their robustness and applicability for real-world medical tasks. Our code is available at https://github.com/frederikpahde/medical-ai-safety.Deep neural networks are increasingly employed in high-stakes medical applications, despite their tendency for shortcut learning in the presence of spurious correlations, which can have potentially fatal consequences in practice. Whereas a multitude of works address either the detection or mitigation of such shortcut behavior in isolation, the Reveal2Revise approach provides a comprehensive bias mitigation framework combining these steps. However, effectively addressing these biases often requires substantial labeling efforts from domain experts. In this work, we review the steps of the Reveal2Revise framework and enhance it with semi-automated interpretability-based bias annotation capabilities. This includes methods for the sample- and feature-level bias annotation, providing valuable information for bias mitigation methods to unlearn the undesired shortcut behavior. We show the applicability of the framework using four medical datasets across two modalities, featuring controlled and real-world spurious correlations caused by data artifacts. We successfully identify and mitigate these biases in VGG16, ResNet50, and contemporary Vision Transformer models, ultimately increasing their robustness and applicability for real-world medical tasks. Our code is available at https://github.com/frederikpahde/medical-ai-safety.
Deep neural networks are increasingly employed in high-stakes medical applications, despite their tendency for shortcut learning in the presence of spurious correlations, which can have potentially fatal consequences in practice. Whereas a multitude of works address either the detection or mitigation of such shortcut behavior in isolation, the Reveal2Revise approach provides a comprehensive bias mitigation framework combining these steps. However, effectively addressing these biases often requires substantial labeling efforts from domain experts. In this work, we review the steps of the Reveal2Revise framework and enhance it with semi-automated interpretability-based bias annotation capabilities. This includes methods for the sample- and feature-level bias annotation, providing valuable information for bias mitigation methods to unlearn the undesired shortcut behavior. We show the applicability of the framework using four medical datasets across two modalities, featuring controlled and real-world spurious correlations caused by data artifacts. We successfully identify and mitigate these biases in VGG16, ResNet50, and contemporary Vision Transformer models, ultimately increasing their robustness and applicability for real-world medical tasks. Our code is available at https://github.com/frederikpahde/medical-ai-safety.
Deep neural networks are increasingly employed in high-stakes medical applications, despite their tendency for shortcut learning in the presence of spurious correlations, which can have potentially fatal consequences in practice. Whereas a multitude of works address either the detection or mitigation of such shortcut behavior in isolation, the Reveal2Revise approach provides a comprehensive bias mitigation framework combining these steps. However, effectively addressing these biases often requires substantial labeling efforts from domain experts. In this work, we review the steps of the Reveal2Revise framework and enhance it with semi-automated interpretability-based bias annotation capabilities. This includes methods for the sample- and feature-level bias annotation, providing valuable information for bias mitigation methods to unlearn the undesired shortcut behavior. We show the applicability of the framework using four medical datasets across two modalities, featuring controlled and real-world spurious correlations caused by data artifacts. We successfully identify and mitigate these biases in VGG16, ResNet50, and contemporary Vision Transformer models, ultimately increasing their robustness and applicability for real-world medical tasks. Our code is available at https://github.com/frederikpahde/medical-ai-safety .
ArticleNumber 206
Author Pahde, Frederik
Samek, Wojciech
Lapuschkin, Sebastian
Wiegand, Thomas
Author_xml – sequence: 1
  givenname: Frederik
  surname: Pahde
  fullname: Pahde, Frederik
  organization: Fraunhofer Heinrich Hertz Institute
– sequence: 2
  givenname: Thomas
  surname: Wiegand
  fullname: Wiegand, Thomas
  organization: Fraunhofer Heinrich Hertz Institute, Technische Universität Berlin, BIFOLD – Berlin Institute for the Foundations of Learning and Data
– sequence: 3
  givenname: Sebastian
  surname: Lapuschkin
  fullname: Lapuschkin, Sebastian
  email: sebastian.lapuschkin@hhi.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Technological University Dublin
– sequence: 4
  givenname: Wojciech
  surname: Samek
  fullname: Samek, Wojciech
  email: wojciech.samek@hhi.fraunhofer.de
  organization: Fraunhofer Heinrich Hertz Institute, Technische Universität Berlin, BIFOLD – Berlin Institute for the Foundations of Learning and Data
BackLink https://www.ncbi.nlm.nih.gov/pubmed/40814399$$D View this record in MEDLINE/PubMed
BookMark eNp9kk1v1DAYhC1URLeFP8ABWeLCJeCvJA4XVFWlrVSJC5wtx36zdZXYwXa2WvHn8e6WAj30ZFl-ZjSj8Qk68sEDQm8p-UgJaT8lSrpOVITVFWkkF9X9C7SidcsrUjf1EVoRKeuqoaw-Ricp3RFCWCObV-hYEEkF77oV-nXh0xKdX-MJrDN6xGfXOOkB8vYzdj5DnCNk3bvR5W1lo9uAxxYymOyCx9pbPLns1np_DQNOc7ELS8JTsDDiHm71xoW4J3VKwTidwWKrs36NXg56TPDm4TxFP75efD-_qm6-XV6fn91URrRtrizlQA1ncuhsw8TQW6Cak4b2UlsKohlaSlrZDFxLyxugjHBBB8k703eScX6Kvhx856UvLQ34HPWo5ugmHbcqaKf-f_HuVq3DRlHGBW_5zuHDg0MMPxdIWU0uGRhH7aF0VZzxriSgNSvo-yfoXViiL_12lJS0FYQU6t2_kR6z_BmmAOwAmBhSijA8IpSo3frqsL4q66v9-uq-iOQTkXF5v0yp5cbnpfwgTfPuN0D8G_sZ1W-9f8bh
CitedBy_id crossref_primary_10_1111_jsr_70044
Cites_doi 10.1109/JBHI.2020.3022989
10.1609/aaai.v38i19.30096
10.1109/ICCV51070.2023.01851
10.1038/s42256-020-0212-3
10.1162/coli_a_00422
10.1038/s42256-025-01084-w
10.1023/A:1022627411411
10.1109/CVPR.2019.00065
10.1145/3514221.3517886
10.1007/978-3-031-73016-0_20
10.24963/ijcai.2017/371
10.1109/CVPR52729.2023.00266
10.5281/zenodo.4414861
10.1109/ISBI.2018.8363547
10.1038/s41467-023-39902-7
10.1371/journal.pmed.1002683
10.1109/ICCV.2017.74
10.1038/s41597-020-0495-6
10.18653/v1/2020.acl-main.647
10.1109/CVPRW63382.2024.00357
10.1007/978-3-031-43895-0_56
10.1109/CVPR.2016.319
10.1371/journal.pone.0130140
10.1109/CVPRW50498.2020.00378
10.23915/distill.00024.001
10.1109/ICNN.1995.488997
10.1016/j.ejca.2019.04.001
10.21105/joss.00861
10.1109/CVPR.2018.00910
10.1038/s42256-023-00711-8
10.1109/CVPRW63382.2024.00353
10.1038/s41467-019-08987-4
10.1016/j.compbiomed.2024.108525
10.1007/978-3-031-44067-0_26
10.1145/342009.335388
10.1016/j.inffus.2021.07.015
10.1109/CVPR.2009.5206848
10.1016/B978-0-44-340553-2.00008-3
10.1038/s42256-020-00257-z
10.1038/sdata.2018.161
10.1038/s41597-020-00622-y
10.1073/pnas.1907375117
10.23915/distill.00007
10.1038/s42256-021-00338-7
10.1609/aaai.v33i01.3301590
10.1109/CVPR.2016.90
10.1016/j.media.2021.102305
10.1609/aaai.v35i13.17389
10.1109/TPAMI.2012.120
10.1111/j.1469-1809.1936.tb02137.x
ContentType Journal Article
Copyright The Author(s) 2025
The Author(s) 2025.
The Author(s) 2025. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
The Author(s) 2025 2025
Copyright_xml – notice: The Author(s) 2025
– notice: The Author(s) 2025.
– notice: The Author(s) 2025. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.
– notice: The Author(s) 2025 2025
DBID C6C
AAYXX
CITATION
NPM
7SC
8FD
JQ2
L7M
L~C
L~D
7X8
5PM
DOI 10.1007/s10994-025-06834-w
DatabaseName Springer Nature OA Free Journals
CrossRef
PubMed
Computer and Information Systems Abstracts
Technology Research Database
ProQuest Computer Science Collection
Advanced Technologies Database with Aerospace
Computer and Information Systems Abstracts – Academic
Computer and Information Systems Abstracts Professional
MEDLINE - Academic
PubMed Central (Full Participant titles)
DatabaseTitle CrossRef
PubMed
Computer and Information Systems Abstracts
Technology Research Database
Computer and Information Systems Abstracts – Academic
Advanced Technologies Database with Aerospace
ProQuest Computer Science Collection
Computer and Information Systems Abstracts Professional
MEDLINE - Academic
DatabaseTitleList MEDLINE - Academic
Computer and Information Systems Abstracts
CrossRef


PubMed
Database_xml – sequence: 1
  dbid: NPM
  name: PubMed
  url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed
  sourceTypes: Index Database
– sequence: 2
  dbid: 7X8
  name: MEDLINE - Academic
  url: https://search.proquest.com/medline
  sourceTypes: Aggregation Database
DeliveryMethod fulltext_linktorsrc
Discipline Computer Science
EISSN 1573-0565
ExternalDocumentID PMC12343733
40814399
10_1007_s10994_025_06834_w
Genre Journal Article
GrantInformation_xml – fundername: HORIZON EUROPE European Research Council
  grantid: 101093003
  funderid: https://doi.org/10.13039/100019180
– fundername: Horizon 2020
  grantid: 965221
  funderid: https://doi.org/10.13039/501100007601
– fundername: Fraunhofer-Institut für Nachrichtentechnik, Heinrich-Hertz-Institut HHI (1050)
– fundername: Bundesministerium für Bildung und Forschung
  grantid: 01IS18025A, 01IS180371I
  funderid: https://doi.org/10.13039/501100002347
– fundername: Deutsche Forschungsgemeinschaft
  grantid: KI-FOR 5363 − project ID: 459422098
  funderid: https://doi.org/10.13039/501100001659
GroupedDBID -~C
-~X
.4S
.86
.DC
.VR
06D
0R~
0VY
199
1N0
203
29M
2J2
2JN
2JY
2KG
2KM
2LR
2~H
30V
4.4
406
408
409
40D
40E
5GY
5VS
67Z
6NX
78A
8TC
8UJ
95-
95.
95~
96X
AAAVM
AABHQ
AACDK
AAHNG
AAIAL
AAJBT
AAJKR
AANZL
AAPKM
AARTL
AASML
AATNV
AATVU
AAUYE
AAWCG
AAYIU
AAYQN
AAYZH
ABAKF
ABBBX
ABBRH
ABBXA
ABDBE
ABDZT
ABECU
ABFSG
ABFTV
ABHLI
ABHQN
ABIVO
ABJNI
ABJOX
ABKCH
ABKTR
ABMNI
ABMQK
ABNWP
ABQBU
ABRTQ
ABSXP
ABTEG
ABTHY
ABTKH
ABTMW
ABWNU
ABXPI
ACAOD
ACDTI
ACGFS
ACGOD
ACHSB
ACHXU
ACKNC
ACMDZ
ACMLO
ACNCT
ACOKC
ACOMO
ACPIV
ACSTC
ACZOJ
ADHHG
ADHIR
ADIMF
ADKFA
ADKNI
ADKPE
ADRFC
ADTPH
ADURQ
ADYFF
ADZKW
AEFQL
AEGAL
AEGNC
AEJHL
AEJRE
AEMSY
AENEX
AEOHA
AEPYU
AESKC
AETLH
AEVLU
AEXYK
AEZWR
AFBBN
AFDZB
AFHIU
AFLOW
AFOHR
AFQWF
AFWTZ
AFZKB
AGAYW
AGDGC
AGJBK
AGMZJ
AGQEE
AGQMX
AGRTI
AGWIL
AGWZB
AGYKE
AHAVH
AHBYD
AHKAY
AHPBZ
AHSBF
AHWEU
AHYZX
AIAKS
AIGIU
AIIXL
AILAN
AITGF
AIXLP
AJRNO
AJZVZ
ALMA_UNASSIGNED_HOLDINGS
ALWAN
AMKLP
AMXSW
AMYLF
AMYQR
AOCGG
ARAPS
ARMRJ
ASPBG
ATHPR
AVWKF
AXYYD
AYFIA
AYJHY
AZFZN
B-.
BA0
BENPR
BGNMA
BSONS
C6C
CS3
CSCUP
DDRTE
DL5
DNIVK
DPUIP
DU5
EBLON
EBS
EIOEI
ESBYG
F5P
FEDTE
FERAY
FFXSO
FIGPU
FNLPD
FRRFC
FWDCC
GGCAI
GGRSB
GJIRD
GNWQR
GQ7
GQ8
GXS
HCIFZ
HF~
HG5
HG6
HMJXF
HQYDN
HRMNR
HVGLF
HZ~
I09
IHE
IJ-
IKXTQ
ITM
IWAJR
IXC
IZIGR
IZQ
I~X
I~Y
I~Z
J-C
J0Z
JBSCW
JCJTX
JZLTJ
KDC
KOV
LAK
LLZTM
M4Y
MA-
N9A
NB0
NPVJJ
NQJWS
NU0
O93
O9G
O9I
O9J
OAM
P19
P2P
P9O
PF-
PT4
QM1
QN7
QOK
QOS
R89
R9I
RHV
RNS
ROL
RPX
RSV
S16
S1Z
S27
S3B
SAP
SCO
SDH
SHX
SISQX
SJYHP
SNE
SNPRN
SNX
SOHCF
SOJ
SPISZ
SRMVM
SSLCW
STPWE
SZN
T13
TAE
TN5
TSG
TSK
TSV
TUC
TUS
U2A
UG4
UOJIU
UTJUX
VC2
W23
W48
WH7
WIP
WK8
YLTOR
Z45
Z8Z
ZMTXR
-Y2
1SB
2.D
28-
2P1
2VQ
5QI
6TJ
88I
8AO
8FE
8FG
AAEWM
AAOBN
AARHV
AAYTO
AAYXX
ABQSL
ABULA
ABUWG
ACBXY
ADHKG
ADMLS
AEBTG
AEFIE
AEKMD
AFEXP
AFFHD
AFGCZ
AFKRA
AGQPQ
AJBLW
AMVHM
ARCSS
AZQEC
BBWZM
BDATZ
BGLVJ
BPHCQ
CAG
CCPQU
CITATION
COF
DWQXO
EJD
FINBP
FSGXE
GNUQQ
H13
I-F
ITG
ITH
K6V
K7-
KOW
M2P
MVM
N2Q
NDZJH
O9-
OVD
P62
PHGZM
PHGZT
PQGLB
PQQKQ
PROAC
Q2X
QF4
QO4
R4E
RNI
RZC
RZE
S26
S28
SCJ
SCLPG
T16
TEORI
UZXMN
VFIZW
NPM
XJT
7SC
8FD
JQ2
L7M
L~C
L~D
7X8
5PM
ID FETCH-LOGICAL-c477t-d13e1c328f9d624fbde1a3061b8ad1e46f710786f3a8d36e120341f839cb98233
IEDL.DBID RSV
ISICitedReferencesCount 1
ISICitedReferencesURI http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001548481600001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
ISSN 0885-6125
IngestDate Tue Nov 04 02:05:02 EST 2025
Fri Aug 15 19:38:56 EDT 2025
Wed Nov 05 08:58:48 EST 2025
Mon Aug 18 01:33:02 EDT 2025
Tue Nov 18 21:13:29 EST 2025
Sat Nov 29 07:33:12 EST 2025
Wed Sep 03 02:47:09 EDT 2025
IsDoiOpenAccess true
IsOpenAccess true
IsPeerReviewed true
IsScholarly true
Issue 9
Keywords Explainable artificial intelligence
Bias mitigation
Data annotation
Interpretability
Spurious correlations
Language English
License The Author(s) 2025.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
LinkModel DirectLink
MergedId FETCHMERGED-LOGICAL-c477t-d13e1c328f9d624fbde1a3061b8ad1e46f710786f3a8d36e120341f839cb98233
Notes ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
Editors: Annalisa Appice, Giuseppina Andresini, Przemyslaw Biecek, Christian Wressnegger.
OpenAccessLink https://link.springer.com/10.1007/s10994-025-06834-w
PMID 40814399
PQID 3238817400
PQPubID 54194
ParticipantIDs pubmedcentral_primary_oai_pubmedcentral_nih_gov_12343733
proquest_miscellaneous_3239786152
proquest_journals_3238817400
pubmed_primary_40814399
crossref_primary_10_1007_s10994_025_06834_w
crossref_citationtrail_10_1007_s10994_025_06834_w
springer_journals_10_1007_s10994_025_06834_w
PublicationCentury 2000
PublicationDate 2025-09-01
PublicationDateYYYYMMDD 2025-09-01
PublicationDate_xml – month: 09
  year: 2025
  text: 2025-09-01
  day: 01
PublicationDecade 2020
PublicationPlace New York
PublicationPlace_xml – name: New York
– name: United States
– name: Dordrecht
PublicationTitle Machine learning
PublicationTitleAbbrev Mach Learn
PublicationTitleAlternate Mach Learn
PublicationYear 2025
Publisher Springer US
Springer Nature B.V
Publisher_xml – name: Springer US
– name: Springer Nature B.V
References R Achtibat (6834_CR2) 2023; 5
6834_CR70
Y Belinkov (6834_CR8) 2022; 48
S Bach (6834_CR5) 2015; 10
6834_CR35
6834_CR79
6834_CR78
R Achanta (6834_CR3) 2012; 34
6834_CR37
6834_CR36
M Dreyer (6834_CR30) 2024; 38
6834_CR31
6834_CR75
6834_CR33
6834_CR32
6834_CR76
TJ Brinker (6834_CR9) 2019; 113
6834_CR39
B Cassidy (6834_CR22) 2022; 75
RA Fisher (6834_CR38) 1936; 7
6834_CR60
R Wightman (6834_CR80) 2019
6834_CR62
6834_CR61
6834_CR68
6834_CR67
6834_CR26
CJ Anders (6834_CR4) 2022; 77
6834_CR25
6834_CR69
6834_CR20
6834_CR64
6834_CR63
6834_CR66
6834_CR21
6834_CR65
S Lapuschkin (6834_CR51) 2019; 10
6834_CR28
D Slijepcevic (6834_CR71) 2021; 3
6834_CR27
P Tschandl (6834_CR77) 2018; 5
6834_CR50
6834_CR13
6834_CR12
6834_CR56
C Olah (6834_CR59) 2017; 2
N Strodthoff (6834_CR74) 2020; 25
6834_CR14
6834_CR53
6834_CR52
N Otsu (6834_CR57) 1975; 11
6834_CR11
6834_CR55
6834_CR10
6834_CR54
J Crabbé (6834_CR24) 2022; 35
6834_CR17
6834_CR19
R Geirhos (6834_CR41) 2020; 2
H Borgli (6834_CR16) 2020; 7
C Olah (6834_CR58) 2020; 5
P Wagner (6834_CR81) 2024; 176
6834_CR7
S Santurkar (6834_CR73) 2021; 34
6834_CR82
6834_CR40
6834_CR84
6834_CR6
6834_CR46
A Brown (6834_CR15) 2023; 14
6834_CR45
P Schramowski (6834_CR72) 2020; 2
6834_CR1
6834_CR48
AJ DeGrave (6834_CR29) 2021; 3
6834_CR47
6834_CR42
6834_CR86
6834_CR85
R Zhang (6834_CR87) 2021; 35
6834_CR44
6834_CR43
D Bau (6834_CR18) 2020; 117
D Erhan (6834_CR34) 2009; 1341
C Cortes (6834_CR23) 1995; 20
6834_CR49
P Wagner (6834_CR83) 2020; 7
References_xml – volume: 25
  start-page: 1519
  issue: 5
  year: 2020
  ident: 6834_CR74
  publication-title: IEEE Journal of Biomedical and Health Informatics
  doi: 10.1109/JBHI.2020.3022989
– volume: 38
  start-page: 21046
  year: 2024
  ident: 6834_CR30
  publication-title: AAAI
  doi: 10.1609/aaai.v38i19.30096
– ident: 6834_CR56
  doi: 10.1109/ICCV51070.2023.01851
– volume: 2
  start-page: 476
  issue: 8
  year: 2020
  ident: 6834_CR72
  publication-title: Nature Machine Intelligence
  doi: 10.1038/s42256-020-0212-3
– volume: 48
  start-page: 207
  issue: 1
  year: 2022
  ident: 6834_CR8
  publication-title: Computational Linguistics
  doi: 10.1162/coli_a_00422
– ident: 6834_CR27
  doi: 10.1038/s42256-025-01084-w
– ident: 6834_CR50
– volume: 20
  start-page: 273
  year: 1995
  ident: 6834_CR23
  publication-title: Machine Learning
  doi: 10.1023/A:1022627411411
– ident: 6834_CR47
  doi: 10.1109/CVPR.2019.00065
– ident: 6834_CR54
– ident: 6834_CR64
  doi: 10.1145/3514221.3517886
– ident: 6834_CR82
  doi: 10.1007/978-3-031-73016-0_20
– ident: 6834_CR44
– ident: 6834_CR60
– ident: 6834_CR66
  doi: 10.24963/ijcai.2017/371
– ident: 6834_CR39
  doi: 10.1109/CVPR52729.2023.00266
– year: 2019
  ident: 6834_CR80
  publication-title: GitHub
  doi: 10.5281/zenodo.4414861
– ident: 6834_CR21
  doi: 10.1109/ISBI.2018.8363547
– ident: 6834_CR68
– volume: 14
  start-page: 4314
  issue: 1
  year: 2023
  ident: 6834_CR15
  publication-title: Nature Communications
  doi: 10.1038/s41467-023-39902-7
– ident: 6834_CR6
– ident: 6834_CR35
– ident: 6834_CR12
– ident: 6834_CR85
  doi: 10.1371/journal.pmed.1002683
– ident: 6834_CR31
– ident: 6834_CR69
  doi: 10.1109/ICCV.2017.74
– volume: 35
  start-page: 2590
  year: 2022
  ident: 6834_CR24
  publication-title: NeurIPS
– volume: 7
  start-page: 1
  issue: 1
  year: 2020
  ident: 6834_CR83
  publication-title: Scientific Data
  doi: 10.1038/s41597-020-0495-6
– ident: 6834_CR65
  doi: 10.18653/v1/2020.acl-main.647
– ident: 6834_CR7
  doi: 10.1109/CVPRW63382.2024.00357
– ident: 6834_CR26
– ident: 6834_CR61
  doi: 10.1007/978-3-031-43895-0_56
– ident: 6834_CR78
– ident: 6834_CR86
  doi: 10.1109/CVPR.2016.319
– volume: 10
  start-page: 0130140
  issue: 7
  year: 2015
  ident: 6834_CR5
  publication-title: PLoS ONE
  doi: 10.1371/journal.pone.0130140
– ident: 6834_CR43
– ident: 6834_CR17
  doi: 10.1109/CVPRW50498.2020.00378
– volume: 5
  start-page: 00024
  issue: 3
  year: 2020
  ident: 6834_CR58
  publication-title: Distill
  doi: 10.23915/distill.00024.001
– ident: 6834_CR53
  doi: 10.1109/ICNN.1995.488997
– volume: 113
  start-page: 47
  year: 2019
  ident: 6834_CR9
  publication-title: European Journal of Cancer
  doi: 10.1016/j.ejca.2019.04.001
– ident: 6834_CR1
– ident: 6834_CR63
– ident: 6834_CR52
  doi: 10.21105/joss.00861
– ident: 6834_CR40
  doi: 10.1109/CVPR.2018.00910
– ident: 6834_CR42
– ident: 6834_CR36
– volume: 5
  start-page: 1006
  issue: 9
  year: 2023
  ident: 6834_CR2
  publication-title: Nature Machine Intelligence
  doi: 10.1038/s42256-023-00711-8
– ident: 6834_CR25
  doi: 10.1109/CVPRW63382.2024.00353
– volume: 10
  start-page: 1096
  issue: 1
  year: 2019
  ident: 6834_CR51
  publication-title: Nature Communications
  doi: 10.1038/s41467-019-08987-4
– ident: 6834_CR67
– ident: 6834_CR84
– ident: 6834_CR70
– volume: 176
  year: 2024
  ident: 6834_CR81
  publication-title: Computers in Biology and Medicine
  doi: 10.1016/j.compbiomed.2024.108525
– ident: 6834_CR19
– ident: 6834_CR79
– ident: 6834_CR75
– volume: 3
  start-page: 1
  issue: 2
  year: 2021
  ident: 6834_CR71
  publication-title: ACM Transactions on Computing for Healthcare (HEALTH)
– ident: 6834_CR55
  doi: 10.1007/978-3-031-44067-0_26
– ident: 6834_CR11
  doi: 10.1145/342009.335388
– volume: 77
  start-page: 261
  year: 2022
  ident: 6834_CR4
  publication-title: Information Fusion
  doi: 10.1016/j.inffus.2021.07.015
– ident: 6834_CR28
  doi: 10.1109/CVPR.2009.5206848
– ident: 6834_CR32
  doi: 10.1016/B978-0-44-340553-2.00008-3
– ident: 6834_CR14
– volume: 2
  start-page: 665
  issue: 11
  year: 2020
  ident: 6834_CR41
  publication-title: Nature Machine Intelligence
  doi: 10.1038/s42256-020-00257-z
– volume: 5
  start-page: 1
  issue: 1
  year: 2018
  ident: 6834_CR77
  publication-title: Scientific Data
  doi: 10.1038/sdata.2018.161
– volume: 7
  start-page: 283
  issue: 1
  year: 2020
  ident: 6834_CR16
  publication-title: Scientific data
  doi: 10.1038/s41597-020-00622-y
– ident: 6834_CR37
– ident: 6834_CR62
– volume: 117
  start-page: 30071
  issue: 48
  year: 2020
  ident: 6834_CR18
  publication-title: Proceedings of the National Academy of Sciences
  doi: 10.1073/pnas.1907375117
– ident: 6834_CR33
– volume: 2
  start-page: 7
  issue: 11
  year: 2017
  ident: 6834_CR59
  publication-title: Distill
  doi: 10.23915/distill.00007
– volume: 3
  start-page: 610
  issue: 7
  year: 2021
  ident: 6834_CR29
  publication-title: Nature Machine Intelligence
  doi: 10.1038/s42256-021-00338-7
– volume: 1341
  start-page: 1
  issue: 3
  year: 2009
  ident: 6834_CR34
  publication-title: University of Montreal
– ident: 6834_CR10
– ident: 6834_CR48
  doi: 10.1609/aaai.v33i01.3301590
– ident: 6834_CR76
– ident: 6834_CR46
  doi: 10.1109/CVPR.2016.90
– volume: 75
  year: 2022
  ident: 6834_CR22
  publication-title: Medical Image Analysis
  doi: 10.1016/j.media.2021.102305
– ident: 6834_CR49
– ident: 6834_CR45
– volume: 35
  start-page: 11682
  year: 2021
  ident: 6834_CR87
  publication-title: AAAI
  doi: 10.1609/aaai.v35i13.17389
– volume: 34
  start-page: 2274
  issue: 11
  year: 2012
  ident: 6834_CR3
  publication-title: IEEE TPAMI
  doi: 10.1109/TPAMI.2012.120
– ident: 6834_CR20
– volume: 34
  start-page: 23359
  year: 2021
  ident: 6834_CR73
  publication-title: NeurIPS
– volume: 11
  start-page: 23
  issue: 285–296
  year: 1975
  ident: 6834_CR57
  publication-title: Automatica
– volume: 7
  start-page: 179
  issue: 2
  year: 1936
  ident: 6834_CR38
  publication-title: Annals of Eugenics
  doi: 10.1111/j.1469-1809.1936.tb02137.x
– ident: 6834_CR13
SSID ssj0002686
Score 2.474371
Snippet Deep neural networks are increasingly employed in high-stakes medical applications, despite their tendency for shortcut learning in the presence of spurious...
SourceID pubmedcentral
proquest
pubmed
crossref
springer
SourceType Open Access Repository
Aggregation Database
Index Database
Enrichment Source
Publisher
StartPage 206
SubjectTerms Annotations
Artifact identification
Artificial Intelligence
Artificial neural networks
Automation
Bias
Computer Science
Control
Datasets
Identification
Localization
Machine Learning
Mechatronics
Medical equipment
Melanoma
Natural Language Processing (NLP)
Neural networks
Robotics
Simulation and Modeling
Subject specialists
Title Ensuring medical AI safety: interpretability-driven detection and mitigation of spurious model behavior and associated data
URI https://link.springer.com/article/10.1007/s10994-025-06834-w
https://www.ncbi.nlm.nih.gov/pubmed/40814399
https://www.proquest.com/docview/3238817400
https://www.proquest.com/docview/3239786152
https://pubmed.ncbi.nlm.nih.gov/PMC12343733
Volume 114
WOSCitedRecordID wos001548481600001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D
hasFullText 1
inHoldings 1
isFullTextHit
isPrint
journalDatabaseRights – providerCode: PRVAVX
  databaseName: SpringerLINK Contemporary 1997-Present
  customDbUrl:
  eissn: 1573-0565
  dateEnd: 99991231
  omitProxy: false
  ssIdentifier: ssj0002686
  issn: 0885-6125
  databaseCode: RSV
  dateStart: 19970101
  isFulltext: true
  titleUrlDefault: https://link.springer.com/search?facet-content-type=%22Journal%22
  providerName: Springer Nature
link http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwnV1Jb9UwEB514cCF0rIFSuVK3MBSvCRxuFWoFUhVVbHp3SLHi3hSm1e9pFQVf56xs5RHAYkeI0-c2J7xfCOPvwF4laEXNU6XFN2LpbLOalqzPKfSIXYQVnhjIs_scXFyomaz8nS4FNaO2e7jkWTcqX-57BZpbHkWCqUISa_WYRPdnQoFGz5--jrtvzyP9R3RfDIa_PdwVebPfay6o1sY83aq5G_npdENHW3dbQAP4cEAO8lBryfbsOaaHdgaSzqQwcIfwY_Dpo0XF8l5f4JDDj6QVnvXXb8l8ylBMWbUXlO7DHslsa6LCV0N0Y0l5_OetgMfF560F9jd4rIlseQOGVkBoqQeVMNZEhJVH8OXo8PP797ToT4DNbIoOmqZcMwIrnxpcy59bR3TGIKwWmnLnMw9wpdC5V5oZUXuGE_RZ3qEZKYuFRfiCWw0i8Y9A2KNZHVqveQG48Uy1aGzEOoVrjSZyBJg4zJVZiAvDzU0zqob2uUwuxXObhVnt7pK4PX0zkVP3fFP6d1x9avBjNtKIKBRGLOlaQL7UzMaYDhV0Y3D6QsyGIkjMOQJPO2VZfqcRMAVAr4E1IoaTQKB3Hu1pZl_iyTfiCgC65RI4M2oTTf_9fdhPP8_8Rdwn0eFDElzu7DRLS_dS7hnvnfzdrkH68VM7UXr-gk5zyFn
linkProvider Springer Nature
linkToHtml http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwnV1bTxUxEJ4omsiLeMUF1Jr4pk22l73xRggE4vHEKBreNt1ewklkDzm7SIh_nmn3AkfURB83ne1u25nON-n0G4C3CXpRbVVB0b0YKqukohVLUyotYgdhhNM68MxOsuk0Pz4uPvWXwpoh2304kgw79Y3LboHGlie-UIqQ9OIu3JPosTxj_ucv38b9l6ehviOaT0K9_-6vyvy-j2V3dAtj3k6V_OW8NLih_bX_G8AjeNjDTrLT6cljuGPrJ7A2lHQgvYU_hZ97dRMuLpLT7gSH7BySRjnbXm6T2ZigGDJqL6lZ-L2SGNuGhK6aqNqQ01lH24GPc0eaM-xuft6QUHKHDKwAQVL1qmEN8Ymqz-Dr_t7R7gHt6zNQLbOspYYJy7TguStMyqWrjGUKQxBW5cowK1OH8CXLUydUbkRqGY_RZzqEZLoqci7Ec1ip57V9AcRoyarYOMk1xotFrHxnPtTLbKETkUTAhmUqdU9e7mtofC-vaZf97JY4u2WY3fIignfjO2cddcdfpbeG1S97M25KgYAmx5gtjiN4MzajAfpTFVVbnD4vg5E4AkMewXqnLOPnJAIuH_BFkC-p0Sjgyb2XW-rZSSD5RkThWadEBO8Hbbr-rz8PY-PfxF_Dg4Ojj5Nycjj9sAmrPCinT6DbgpV2cW5fwn39o501i1fBxq4A_zUjYw
linkToPdf http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV1Zb9QwEB5BQagvlLuhBYzEG1hNbOfirSpdUVGtKnGob5HjQ6xEvatNSlX1z3fsHO1SQEI8Rp44sTOT-Uae-QbgTYpeVBlZUnQvmoo6rWmdZBkVBrED19wqFXhmD_PptDg-Lo-uVfGHbPfhSLKrafAsTa7dWWi7c63wLVDastQ3TeGCnt2GO8In0vt4_fO38V_MstDrEU0ppd6X92Uzv59j1TXdwJs30yZ_OTsNLmmy8f-LeQD3ezhKdjv9eQi3jHsEG0OrB9Jb_mO42HdNKGgkJ93JDtk9II20pj1_T2Zj4mLItD2neun_oUSbNiR6OSKdJiezjs4DL-eWNAucbn7akNCKhwxsAUFS9ipjNPEJrE_g62T_y95H2vdtoErkeUt1wk2iOCtsqTMmbK1NIjE0SepC6sSIzCKsyYvMcllonpmExehLLUI1VZcF4_wprLm5M5tAtBJJHWsrmMI4soyln8yHgLkpVcrTCJLhk1WqJzX3vTV-VFd0zH53K9zdKuxudRbB2_GeRUfp8Vfp7UETqt68m4oj0CkwlovjCF6Pw2iY_rRFOoPb52UwQkfAyCJ41inO-DiBQMwHghEUKyo1CnjS79URN_seyL8RaXg2Kh7Bu0Gzrt7rz8t4_m_ir-De0YdJdXgw_bQF6yzops-r24a1dnlqXsBd9bOdNcuXwdwuAe-NLEc
openUrl ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Ensuring+medical+AI+safety%3A+interpretability-driven+detection+and+mitigation+of+spurious+model+behavior+and+associated+data&rft.jtitle=Machine+learning&rft.au=Pahde%2C+Frederik&rft.au=Wiegand%2C+Thomas&rft.au=Lapuschkin%2C+Sebastian&rft.au=Samek%2C+Wojciech&rft.date=2025-09-01&rft.pub=Springer+US&rft.issn=0885-6125&rft.eissn=1573-0565&rft.volume=114&rft.issue=9&rft_id=info:doi/10.1007%2Fs10994-025-06834-w&rft.externalDocID=10_1007_s10994_025_06834_w
thumbnail_l http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=0885-6125&client=summon
thumbnail_m http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=0885-6125&client=summon
thumbnail_s http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=0885-6125&client=summon