Contrastive Integrated Gradients: A Feature Attribution-Based Method for Explaining Whole Slide Image Classification Article Swipe
Anh Vu
,
Tuan L. Vo
,
Ngoc Bui
,
Akash Awasthi
,
Huy Q. Vo
,
Thanh-Huy Nguyen
,
Zhu Han
,
Chandra Mohan
·
YOU?
·
· 2025
· Open Access
·
· DOI: https://doi.org/10.48550/arxiv.2511.08464
YOU?
·
· 2025
· Open Access
·
· DOI: https://doi.org/10.48550/arxiv.2511.08464
Interpretability is essential in Whole Slide Image (WSI) analysis for computational pathology, where understanding model predictions helps build trust in AI-assisted diagnostics. While Integrated Gradients (IG) and related attribution methods have shown promise, applying them directly to WSIs introduces challenges due to their high-resolution nature. These methods capture model decision patterns but may overlook class-discriminative signals that are crucial for distinguishing between tumor subtypes. In this work, we introduce Contrastive Integrated Gradients (CIG), a novel attribution method that enhances interpretability by computing contrastive gradients in logit space. First, CIG highlights class-discriminative regions by comparing feature importance relative to a reference class, offering sharper differentiation between tumor and non-tumor areas. Second, CIG satisfies the axioms of integrated attribution, ensuring consistency and theoretical soundness. Third, we propose two attribution quality metrics, MIL-AIC and MIL-SIC, which measure how predictive information and model confidence evolve with access to salient regions, particularly under weak supervision. We validate CIG across three datasets spanning distinct cancer types: CAMELYON16 (breast cancer metastasis in lymph nodes), TCGA-RCC (renal cell carcinoma), and TCGA-Lung (lung cancer). Experimental results demonstrate that CIG yields more informative attributions both quantitatively, using MIL-AIC and MIL-SIC, and qualitatively, through visualizations that align closely with ground truth tumor regions, underscoring its potential for interpretable and trustworthy WSI-based diagnostics
Related Topics
Concepts
No concepts available.
Metadata
- Type
- preprint
- Landing Page
- http://arxiv.org/abs/2511.08464
- https://arxiv.org/pdf/2511.08464
- OA Status
- green
- OpenAlex ID
- https://openalex.org/W4416332680
All OpenAlex metadata
Raw OpenAlex JSON
- OpenAlex ID
-
https://openalex.org/W4416332680Canonical identifier for this work in OpenAlex
- DOI
-
https://doi.org/10.48550/arxiv.2511.08464Digital Object Identifier
- Title
-
Contrastive Integrated Gradients: A Feature Attribution-Based Method for Explaining Whole Slide Image ClassificationWork title
- Type
-
preprintOpenAlex work type
- Publication year
-
2025Year of publication
- Publication date
-
2025-11-11Full publication date if available
- Authors
-
Anh Vu, Tuan L. Vo, Ngoc Bui, Akash Awasthi, Huy Q. Vo, Thanh-Huy Nguyen, Zhu Han, Chandra MohanList of authors in order
- Landing page
-
https://arxiv.org/abs/2511.08464Publisher landing page
- PDF URL
-
https://arxiv.org/pdf/2511.08464Direct link to full text PDF
- Open access
-
YesWhether a free full text is available
- OA status
-
greenOpen access status per OpenAlex
- OA URL
-
https://arxiv.org/pdf/2511.08464Direct OA link when available
- Cited by
-
0Total citation count in OpenAlex
Full payload
| id | https://openalex.org/W4416332680 |
|---|---|
| doi | https://doi.org/10.48550/arxiv.2511.08464 |
| ids.doi | https://doi.org/10.48550/arxiv.2511.08464 |
| ids.openalex | https://openalex.org/W4416332680 |
| fwci | |
| type | preprint |
| title | Contrastive Integrated Gradients: A Feature Attribution-Based Method for Explaining Whole Slide Image Classification |
| biblio.issue | |
| biblio.volume | |
| biblio.last_page | |
| biblio.first_page | |
| is_xpac | False |
| apc_list | |
| apc_paid | |
| language | |
| locations[0].id | pmh:oai:arXiv.org:2511.08464 |
| locations[0].is_oa | True |
| locations[0].source.id | https://openalex.org/S4306400194 |
| locations[0].source.issn | |
| locations[0].source.type | repository |
| locations[0].source.is_oa | True |
| locations[0].source.issn_l | |
| locations[0].source.is_core | False |
| locations[0].source.is_in_doaj | False |
| locations[0].source.display_name | arXiv (Cornell University) |
| locations[0].source.host_organization | https://openalex.org/I205783295 |
| locations[0].source.host_organization_name | Cornell University |
| locations[0].source.host_organization_lineage | https://openalex.org/I205783295 |
| locations[0].license | |
| locations[0].pdf_url | https://arxiv.org/pdf/2511.08464 |
| locations[0].version | submittedVersion |
| locations[0].raw_type | text |
| locations[0].license_id | |
| locations[0].is_accepted | False |
| locations[0].is_published | False |
| locations[0].raw_source_name | |
| locations[0].landing_page_url | http://arxiv.org/abs/2511.08464 |
| locations[1].id | doi:10.48550/arxiv.2511.08464 |
| locations[1].is_oa | True |
| locations[1].source.id | https://openalex.org/S4306400194 |
| locations[1].source.issn | |
| locations[1].source.type | repository |
| locations[1].source.is_oa | True |
| locations[1].source.issn_l | |
| locations[1].source.is_core | False |
| locations[1].source.is_in_doaj | False |
| locations[1].source.display_name | arXiv (Cornell University) |
| locations[1].source.host_organization | https://openalex.org/I205783295 |
| locations[1].source.host_organization_name | Cornell University |
| locations[1].source.host_organization_lineage | https://openalex.org/I205783295 |
| locations[1].license | |
| locations[1].pdf_url | |
| locations[1].version | |
| locations[1].raw_type | article |
| locations[1].license_id | |
| locations[1].is_accepted | False |
| locations[1].is_published | |
| locations[1].raw_source_name | |
| locations[1].landing_page_url | https://doi.org/10.48550/arxiv.2511.08464 |
| indexed_in | arxiv, datacite |
| authorships[0].author.id | https://openalex.org/A5113493557 |
| authorships[0].author.orcid | |
| authorships[0].author.display_name | Anh Vu |
| authorships[0].author_position | first |
| authorships[0].raw_author_name | Vu, Anh Mai |
| authorships[0].is_corresponding | False |
| authorships[1].author.id | https://openalex.org/A5109622472 |
| authorships[1].author.orcid | https://orcid.org/0009-0001-8545-4719 |
| authorships[1].author.display_name | Tuan L. Vo |
| authorships[1].author_position | middle |
| authorships[1].raw_author_name | Vo, Tuan L. |
| authorships[1].is_corresponding | False |
| authorships[2].author.id | https://openalex.org/A5112210778 |
| authorships[2].author.orcid | https://orcid.org/0000-0002-4345-6003 |
| authorships[2].author.display_name | Ngoc Bui |
| authorships[2].author_position | middle |
| authorships[2].raw_author_name | Bui, Ngoc Lam Quang |
| authorships[2].is_corresponding | False |
| authorships[3].author.id | https://openalex.org/A5088671507 |
| authorships[3].author.orcid | https://orcid.org/0000-0002-8383-0637 |
| authorships[3].author.display_name | Akash Awasthi |
| authorships[3].author_position | middle |
| authorships[3].raw_author_name | Awasthi, Akash |
| authorships[3].is_corresponding | False |
| authorships[4].author.id | https://openalex.org/A5060238453 |
| authorships[4].author.orcid | https://orcid.org/0009-0004-3130-5066 |
| authorships[4].author.display_name | Huy Q. Vo |
| authorships[4].author_position | middle |
| authorships[4].raw_author_name | Vo, Huy Quoc |
| authorships[4].is_corresponding | False |
| authorships[5].author.id | https://openalex.org/A5103260009 |
| authorships[5].author.orcid | https://orcid.org/0009-0003-9355-5434 |
| authorships[5].author.display_name | Thanh-Huy Nguyen |
| authorships[5].author_position | middle |
| authorships[5].raw_author_name | Nguyen, Thanh-Huy |
| authorships[5].is_corresponding | False |
| authorships[6].author.id | https://openalex.org/A5063667378 |
| authorships[6].author.orcid | https://orcid.org/0000-0002-6606-5822 |
| authorships[6].author.display_name | Zhu Han |
| authorships[6].author_position | last |
| authorships[6].raw_author_name | Han, Zhu |
| authorships[6].is_corresponding | False |
| authorships[7].author.id | https://openalex.org/A5042585597 |
| authorships[7].author.orcid | https://orcid.org/0000-0001-7896-5740 |
| authorships[7].author.display_name | Chandra Mohan |
| authorships[7].author_position | middle |
| authorships[7].raw_author_name | Mohan, Chandra |
| authorships[7].is_corresponding | False |
| has_content.pdf | True |
| has_content.grobid_xml | False |
| is_paratext | False |
| open_access.is_oa | True |
| open_access.oa_url | https://arxiv.org/pdf/2511.08464 |
| open_access.oa_status | green |
| open_access.any_repository_has_fulltext | False |
| created_date | 2025-11-13T00:00:00 |
| display_name | Contrastive Integrated Gradients: A Feature Attribution-Based Method for Explaining Whole Slide Image Classification |
| has_fulltext | False |
| is_retracted | False |
| updated_date | 2025-11-28T11:27:13.757401 |
| primary_topic | |
| cited_by_count | 0 |
| locations_count | 2 |
| best_oa_location.id | pmh:oai:arXiv.org:2511.08464 |
| best_oa_location.is_oa | True |
| best_oa_location.source.id | https://openalex.org/S4306400194 |
| best_oa_location.source.issn | |
| best_oa_location.source.type | repository |
| best_oa_location.source.is_oa | True |
| best_oa_location.source.issn_l | |
| best_oa_location.source.is_core | False |
| best_oa_location.source.is_in_doaj | False |
| best_oa_location.source.display_name | arXiv (Cornell University) |
| best_oa_location.source.host_organization | https://openalex.org/I205783295 |
| best_oa_location.source.host_organization_name | Cornell University |
| best_oa_location.source.host_organization_lineage | https://openalex.org/I205783295 |
| best_oa_location.license | |
| best_oa_location.pdf_url | https://arxiv.org/pdf/2511.08464 |
| best_oa_location.version | submittedVersion |
| best_oa_location.raw_type | text |
| best_oa_location.license_id | |
| best_oa_location.is_accepted | False |
| best_oa_location.is_published | False |
| best_oa_location.raw_source_name | |
| best_oa_location.landing_page_url | http://arxiv.org/abs/2511.08464 |
| primary_location.id | pmh:oai:arXiv.org:2511.08464 |
| primary_location.is_oa | True |
| primary_location.source.id | https://openalex.org/S4306400194 |
| primary_location.source.issn | |
| primary_location.source.type | repository |
| primary_location.source.is_oa | True |
| primary_location.source.issn_l | |
| primary_location.source.is_core | False |
| primary_location.source.is_in_doaj | False |
| primary_location.source.display_name | arXiv (Cornell University) |
| primary_location.source.host_organization | https://openalex.org/I205783295 |
| primary_location.source.host_organization_name | Cornell University |
| primary_location.source.host_organization_lineage | https://openalex.org/I205783295 |
| primary_location.license | |
| primary_location.pdf_url | https://arxiv.org/pdf/2511.08464 |
| primary_location.version | submittedVersion |
| primary_location.raw_type | text |
| primary_location.license_id | |
| primary_location.is_accepted | False |
| primary_location.is_published | False |
| primary_location.raw_source_name | |
| primary_location.landing_page_url | http://arxiv.org/abs/2511.08464 |
| publication_date | 2025-11-11 |
| publication_year | 2025 |
| referenced_works_count | 0 |
| abstract_inverted_index.a | 73, 98 |
| abstract_inverted_index.In | 64 |
| abstract_inverted_index.We | 150 |
| abstract_inverted_index.by | 80, 92 |
| abstract_inverted_index.in | 3, 19, 84, 164 |
| abstract_inverted_index.is | 1 |
| abstract_inverted_index.of | 114 |
| abstract_inverted_index.to | 36, 41, 97, 143 |
| abstract_inverted_index.we | 67, 123 |
| abstract_inverted_index.CIG | 88, 110, 152, 179 |
| abstract_inverted_index.and | 26, 106, 119, 130, 137, 171, 188, 190, 207 |
| abstract_inverted_index.are | 57 |
| abstract_inverted_index.but | 51 |
| abstract_inverted_index.due | 40 |
| abstract_inverted_index.for | 9, 59, 205 |
| abstract_inverted_index.how | 134 |
| abstract_inverted_index.its | 203 |
| abstract_inverted_index.may | 52 |
| abstract_inverted_index.the | 112 |
| abstract_inverted_index.two | 125 |
| abstract_inverted_index.(IG) | 25 |
| abstract_inverted_index.WSIs | 37 |
| abstract_inverted_index.both | 184 |
| abstract_inverted_index.cell | 169 |
| abstract_inverted_index.have | 30 |
| abstract_inverted_index.more | 181 |
| abstract_inverted_index.that | 56, 77, 178, 194 |
| abstract_inverted_index.them | 34 |
| abstract_inverted_index.this | 65 |
| abstract_inverted_index.weak | 148 |
| abstract_inverted_index.with | 141, 197 |
| abstract_inverted_index.(WSI) | 7 |
| abstract_inverted_index.(lung | 173 |
| abstract_inverted_index.Image | 6 |
| abstract_inverted_index.Slide | 5 |
| abstract_inverted_index.These | 45 |
| abstract_inverted_index.While | 22 |
| abstract_inverted_index.Whole | 4 |
| abstract_inverted_index.align | 195 |
| abstract_inverted_index.build | 17 |
| abstract_inverted_index.helps | 16 |
| abstract_inverted_index.logit | 85 |
| abstract_inverted_index.lymph | 165 |
| abstract_inverted_index.model | 14, 48, 138 |
| abstract_inverted_index.novel | 74 |
| abstract_inverted_index.shown | 31 |
| abstract_inverted_index.their | 42 |
| abstract_inverted_index.three | 154 |
| abstract_inverted_index.trust | 18 |
| abstract_inverted_index.truth | 199 |
| abstract_inverted_index.tumor | 62, 105, 200 |
| abstract_inverted_index.under | 147 |
| abstract_inverted_index.using | 186 |
| abstract_inverted_index.where | 12 |
| abstract_inverted_index.which | 132 |
| abstract_inverted_index.work, | 66 |
| abstract_inverted_index.(CIG), | 72 |
| abstract_inverted_index.(renal | 168 |
| abstract_inverted_index.First, | 87 |
| abstract_inverted_index.Third, | 122 |
| abstract_inverted_index.access | 142 |
| abstract_inverted_index.across | 153 |
| abstract_inverted_index.areas. | 108 |
| abstract_inverted_index.axioms | 113 |
| abstract_inverted_index.cancer | 158, 162 |
| abstract_inverted_index.class, | 100 |
| abstract_inverted_index.evolve | 140 |
| abstract_inverted_index.ground | 198 |
| abstract_inverted_index.method | 76 |
| abstract_inverted_index.space. | 86 |
| abstract_inverted_index.types: | 159 |
| abstract_inverted_index.yields | 180 |
| abstract_inverted_index.(breast | 161 |
| abstract_inverted_index.MIL-AIC | 129, 187 |
| abstract_inverted_index.Second, | 109 |
| abstract_inverted_index.between | 61, 104 |
| abstract_inverted_index.capture | 47 |
| abstract_inverted_index.closely | 196 |
| abstract_inverted_index.crucial | 58 |
| abstract_inverted_index.feature | 94 |
| abstract_inverted_index.measure | 133 |
| abstract_inverted_index.methods | 29, 46 |
| abstract_inverted_index.nature. | 44 |
| abstract_inverted_index.nodes), | 166 |
| abstract_inverted_index.propose | 124 |
| abstract_inverted_index.quality | 127 |
| abstract_inverted_index.regions | 91 |
| abstract_inverted_index.related | 27 |
| abstract_inverted_index.results | 176 |
| abstract_inverted_index.salient | 144 |
| abstract_inverted_index.sharper | 102 |
| abstract_inverted_index.signals | 55 |
| abstract_inverted_index.through | 192 |
| abstract_inverted_index.MIL-SIC, | 131, 189 |
| abstract_inverted_index.TCGA-RCC | 167 |
| abstract_inverted_index.analysis | 8 |
| abstract_inverted_index.applying | 33 |
| abstract_inverted_index.cancer). | 174 |
| abstract_inverted_index.datasets | 155 |
| abstract_inverted_index.decision | 49 |
| abstract_inverted_index.directly | 35 |
| abstract_inverted_index.distinct | 157 |
| abstract_inverted_index.enhances | 78 |
| abstract_inverted_index.ensuring | 117 |
| abstract_inverted_index.metrics, | 128 |
| abstract_inverted_index.offering | 101 |
| abstract_inverted_index.overlook | 53 |
| abstract_inverted_index.patterns | 50 |
| abstract_inverted_index.promise, | 32 |
| abstract_inverted_index.regions, | 145, 201 |
| abstract_inverted_index.relative | 96 |
| abstract_inverted_index.spanning | 156 |
| abstract_inverted_index.validate | 151 |
| abstract_inverted_index.Gradients | 24, 71 |
| abstract_inverted_index.TCGA-Lung | 172 |
| abstract_inverted_index.WSI-based | 209 |
| abstract_inverted_index.comparing | 93 |
| abstract_inverted_index.computing | 81 |
| abstract_inverted_index.essential | 2 |
| abstract_inverted_index.gradients | 83 |
| abstract_inverted_index.introduce | 68 |
| abstract_inverted_index.non-tumor | 107 |
| abstract_inverted_index.potential | 204 |
| abstract_inverted_index.reference | 99 |
| abstract_inverted_index.satisfies | 111 |
| abstract_inverted_index.subtypes. | 63 |
| abstract_inverted_index.CAMELYON16 | 160 |
| abstract_inverted_index.Integrated | 23, 70 |
| abstract_inverted_index.challenges | 39 |
| abstract_inverted_index.confidence | 139 |
| abstract_inverted_index.highlights | 89 |
| abstract_inverted_index.importance | 95 |
| abstract_inverted_index.integrated | 115 |
| abstract_inverted_index.introduces | 38 |
| abstract_inverted_index.metastasis | 163 |
| abstract_inverted_index.pathology, | 11 |
| abstract_inverted_index.predictive | 135 |
| abstract_inverted_index.soundness. | 121 |
| abstract_inverted_index.AI-assisted | 20 |
| abstract_inverted_index.Contrastive | 69 |
| abstract_inverted_index.attribution | 28, 75, 126 |
| abstract_inverted_index.carcinoma), | 170 |
| abstract_inverted_index.consistency | 118 |
| abstract_inverted_index.contrastive | 82 |
| abstract_inverted_index.demonstrate | 177 |
| abstract_inverted_index.diagnostics | 210 |
| abstract_inverted_index.information | 136 |
| abstract_inverted_index.informative | 182 |
| abstract_inverted_index.predictions | 15 |
| abstract_inverted_index.theoretical | 120 |
| abstract_inverted_index.trustworthy | 208 |
| abstract_inverted_index.Experimental | 175 |
| abstract_inverted_index.attribution, | 116 |
| abstract_inverted_index.attributions | 183 |
| abstract_inverted_index.diagnostics. | 21 |
| abstract_inverted_index.particularly | 146 |
| abstract_inverted_index.supervision. | 149 |
| abstract_inverted_index.underscoring | 202 |
| abstract_inverted_index.computational | 10 |
| abstract_inverted_index.interpretable | 206 |
| abstract_inverted_index.understanding | 13 |
| abstract_inverted_index.distinguishing | 60 |
| abstract_inverted_index.qualitatively, | 191 |
| abstract_inverted_index.visualizations | 193 |
| abstract_inverted_index.differentiation | 103 |
| abstract_inverted_index.high-resolution | 43 |
| abstract_inverted_index.quantitatively, | 185 |
| abstract_inverted_index.Interpretability | 0 |
| abstract_inverted_index.interpretability | 79 |
| abstract_inverted_index.class-discriminative | 54, 90 |
| cited_by_percentile_year | |
| countries_distinct_count | 0 |
| institutions_distinct_count | 8 |
| citation_normalized_percentile |