{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,6]],"date-time":"2026-07-06T12:19:33Z","timestamp":1783340373472,"version":"3.54.6"},"reference-count":35,"publisher":"SAGE Publications","license":[{"start":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T00:00:00Z","timestamp":1782864000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by-nc\/4.0\/"},{"start":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T00:00:00Z","timestamp":1782864000000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/journals.sagepub.com\/page\/policies\/text-and-data-mining-license"}],"content-domain":{"domain":["journals.sagepub.com"],"crossmark-restriction":true},"short-container-title":["Technol Cancer Res Treat"],"published-print":{"date-parts":[[2026,7]]},"abstract":"<jats:sec>\n                    <jats:title>Introduction<\/jats:title>\n                    <jats:p>Determining the HER2 status accurately is a critical determinant in breast cancer treatment planning. Manual scoring remains highly susceptible to inter- and intra-observer variability, particularly in diagnostically ambiguous cases (HER2 1+\/2+). Furthermore, contemporary deep learning models frequently lack the robustness and interpretability required for safe clinical integration.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>\n                      We propose a hybrid ensemble framework that synergizes EVA-02-Large, Vision Transformer-Base, and ConvNeXt-V2-Nano architectures via an adaptive late-fusion mechanism. The model was trained on the expert-annotated HER2-IHC-40x dataset (\n                      <jats:inline-formula>\n                        <mml:math xmlns:mml=\"http:\/\/www.w3.org\/1998\/Math\/MathML\" overflow=\"scroll\">\n                          <mml:mrow>\n                            <mml:mi>N<\/mml:mi>\n                            <mml:mo>=<\/mml:mo>\n                            <mml:mn>10<\/mml:mn>\n                            <mml:mo>,<\/mml:mo>\n                            <mml:mn>997<\/mml:mn>\n                          <\/mml:mrow>\n                        <\/mml:math>\n                      <\/jats:inline-formula>\n                      patches) and rigorously evaluated across ten independent trials. The experimental protocol incorporated comprehensive perturbation analysis across 20+ image corruption types and multi-perspective Explainable AI (XAI) assessment using Grad-CAM, Grad-CAM++, and Layer-CAM to validate alignment with pathologist-recognized morphological features.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>\n                      The fused ensemble achieved state-of-the-art weighted accuracy (\n                      <jats:inline-formula>\n                        <mml:math xmlns:mml=\"http:\/\/www.w3.org\/1998\/Math\/MathML\" overflow=\"scroll\">\n                          <mml:mrow>\n                            <mml:mn>98.05<\/mml:mn>\n                            <mml:mo>%<\/mml:mo>\n                            <mml:mo>\u2009<\/mml:mo>\n                            <mml:mo>\u00b1<\/mml:mo>\n                            <mml:mo>\u2009<\/mml:mo>\n                            <mml:mn>0.10<\/mml:mn>\n                            <mml:mo>%<\/mml:mo>\n                          <\/mml:mrow>\n                        <\/mml:math>\n                      <\/jats:inline-formula>\n                      ) and Balanced Accuracy (\n                      <jats:inline-formula>\n                        <mml:math xmlns:mml=\"http:\/\/www.w3.org\/1998\/Math\/MathML\" overflow=\"scroll\">\n                          <mml:mrow>\n                            <mml:mn>98.25<\/mml:mn>\n                            <mml:mo>%<\/mml:mo>\n                            <mml:mo>\u2009<\/mml:mo>\n                            <mml:mo>\u00b1<\/mml:mo>\n                            <mml:mo>\u2009<\/mml:mo>\n                            <mml:mn>0.11<\/mml:mn>\n                            <mml:mo>%<\/mml:mo>\n                          <\/mml:mrow>\n                        <\/mml:math>\n                      <\/jats:inline-formula>\n                      ), outperforming the strongest single backbone by 0.45%. Misclassification rates for equivocal classes decreased by &gt;18%. The framework demonstrated high resilience to structural perturbations (\n                      <jats:inline-formula>\n                        <mml:math xmlns:mml=\"http:\/\/www.w3.org\/1998\/Math\/MathML\" overflow=\"scroll\">\n                          <mml:mrow>\n                            <mml:mi>m<\/mml:mi>\n                            <mml:mi>C<\/mml:mi>\n                            <mml:mi>E<\/mml:mi>\n                            <mml:mo>\u2009<\/mml:mo>\n                            <mml:mo>&lt;<\/mml:mo>\n                            <mml:mo>\u2009<\/mml:mo>\n                            <mml:mn>0.05<\/mml:mn>\n                          <\/mml:mrow>\n                        <\/mml:math>\n                      <\/jats:inline-formula>\n                      ) while exhibiting expected sensitivity to extreme photometric variations. XAI analysis confirmed that model attention consistently prioritizes clinically relevant membrane staining patterns, mirroring expert diagnostic reasoning.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>This study establishes a highly accurate and interpretable pipeline for automated HER2-IHC scoring, demonstrating that hybrid transformer-CNN ensembles can effectively resolve diagnostically challenging cases. By combining superior predictive performance, structural robustness, and transparent decision-making, the proposed framework offers a highly promising foundational framework that, following prospective multi-reader clinical validation, could serve as a robust decision-support tool to standardize HER2 assessment, reduce inter-observer variability, and accelerate precision oncology workflows.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1177\/15330338261465522","type":"journal-article","created":{"date-parts":[[2026,7,4]],"date-time":"2026-07-04T10:22:06Z","timestamp":1783160526000},"update-policy":"https:\/\/doi.org\/10.1177\/sage-journals-update-policy","source":"Crossref","is-referenced-by-count":0,"title":["Bridging Global Attention and Local Hierarchies: A Robust Hybrid Ensemble Framework With Multi-Perspective Explainability for Automated HER2-IHC Scoring"],"prefix":"10.1177","volume":"25","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-0686-4411","authenticated-orcid":false,"given":"Hossam Magdy","family":"Balaha","sequence":"first","affiliation":[{"name":"University of Louisville"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Norah Saleh","family":"Alghamdi","sequence":"additional","affiliation":[{"name":"Princess Nourah Bint Abdulrahman University (PNU)"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nithya Rekha","family":"Sivakumar","sequence":"additional","affiliation":[{"name":"Princess Nourah Bint Abdulrahman University (PNU)"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shakila","family":"Basheer","sequence":"additional","affiliation":[{"name":"Princess Nourah Bint Abdulrahman University (PNU)"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Khadiga M.","family":"Ali","sequence":"additional","affiliation":[{"name":"Mansoura University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7264-1323","authenticated-orcid":false,"given":"Ayman","family":"El-Baz","sequence":"additional","affiliation":[{"name":"University of Louisville"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"179","published-online":{"date-parts":[[2026,7,4]]},"reference":[{"key":"e_1_3_8_2_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.jncc.2024.05.001"},{"key":"e_1_3_8_3_2","doi-asserted-by":"publisher","DOI":"10.5858\/arpa.2018-0902-SA"},{"key":"e_1_3_8_4_2","doi-asserted-by":"publisher","DOI":"10.1111\/his.15213"},{"key":"e_1_3_8_5_2","doi-asserted-by":"publisher","DOI":"10.1200\/JCO.19.00443"},{"key":"e_1_3_8_6_2","doi-asserted-by":"publisher","DOI":"10.1200\/JCO.20.01204"},{"key":"e_1_3_8_7_2","doi-asserted-by":"crossref","unstructured":"Kabir S Vranic S Al Saady RM et al. The utility of a deep learning-based approach in Her-2\/neu assessment in breast cancer. 2024;238:122051.","DOI":"10.1016\/j.eswa.2023.122051"},{"key":"e_1_3_8_8_2","doi-asserted-by":"publisher","DOI":"10.1200\/JCO.22.02864"},{"key":"e_1_3_8_9_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.prp.2018.10.015"},{"key":"e_1_3_8_10_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.prp.2024.155472"},{"key":"e_1_3_8_11_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41591-018-0177-5"},{"key":"e_1_3_8_12_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41467-020-20030-5"},{"key":"e_1_3_8_13_2","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.1717139115"},{"key":"e_1_3_8_14_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.csbj.2021.12.028"},{"key":"e_1_3_8_15_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41379-021-00911-w"},{"key":"e_1_3_8_16_2","doi-asserted-by":"publisher","DOI":"10.1109\/ACCESS.2024.3374890"},{"key":"e_1_3_8_17_2","doi-asserted-by":"publisher","DOI":"10.3389\/fonc.2022.829041"},{"key":"e_1_3_8_18_2","doi-asserted-by":"publisher","DOI":"10.1109\/InCIT66780.2025.11276061"},{"key":"e_1_3_8_19_2","first-page":"111922","volume-title":"Her2-ihc-40x: A high-resolution histopathology dataset for her2 ihc scoring in breast cancer","author":"Nabi MS","year":"2025","unstructured":"Nabi MS, Fauzi MFA, Rehman ZU, et al. Her2-ihc-40x: A high-resolution histopathology dataset for her2 ihc scoring in breast cancer; 2025.111922."},{"key":"e_1_3_8_20_2","doi-asserted-by":"publisher","DOI":"10.1136\/bmj-2023-078378"},{"key":"e_1_3_8_21_2","volume-title":"Beyond empirical risk minimization","author":"Zhang H","year":"2017","unstructured":"Zhang H, Cisse M, Dauphin YN, Lopez-Paz D. Beyond empirical risk minimization; 2017."},{"key":"e_1_3_8_22_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.imavis.2024.105171"},{"key":"e_1_3_8_23_2","volume-title":"An image is worth 16x16 words: Transformers for image recognition at scale","author":"Dosovitskiy A","year":"2020","unstructured":"Dosovitskiy A, Beyer L, Kolesnikov A, et al. An image is worth 16x16 words: Transformers for image recognition at scale; 2020."},{"key":"e_1_3_8_24_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01548"},{"key":"e_1_3_8_25_2","volume-title":"Decoupled weight decay regularization","author":"Loshchilov I","year":"2017","unstructured":"Loshchilov I, Hutter F. Decoupled weight decay regularization; 2017."},{"key":"e_1_3_8_26_2","volume-title":"Proceedings of the 2021 AIChE Annual Meeting","author":"Bhakte A","year":"2021","unstructured":"Bhakte A, Vasista BS, Srinivasan R. Gradient-weighted class activation mapping (Grad-CAM) based explanations for process monitoring results from deep neural networks. In: Proceedings of the 2021 AIChE Annual Meeting. AIChE; 2021."},{"key":"e_1_3_8_27_2","doi-asserted-by":"publisher","DOI":"10.1109\/WACV.2018.00097"},{"key":"e_1_3_8_28_2","first-page":"4846","article-title":"Augmented grad-cam++","volume":"12","author":"Gao Y","year":"2023","unstructured":"Gao Y, Liu J, Li W, Hou M, Li Y, Zhao H. Augmented grad-cam++. Super-resolution saliency maps for visual interpretation of deep neural network. 2023;12:4846.","journal-title":"Super-resolution saliency maps for visual interpretation of deep neural network"},{"key":"e_1_3_8_29_2","volume-title":"Rethinking class activation maps for segmentation: revealing semantic information in shallow layers by reducing noise","author":"Dong HC","year":"2023","unstructured":"Dong HC, Jiang Y, Huang Y, et al. Rethinking class activation maps for segmentation: revealing semantic information in shallow layers by reducing noise; 2023."},{"key":"e_1_3_8_30_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2019.05.010"},{"key":"e_1_3_8_31_2","doi-asserted-by":"publisher","DOI":"10.1155\/2022\/9910471"},{"key":"e_1_3_8_32_2","doi-asserted-by":"publisher","DOI":"10.1007\/s00521-025-11750-0"},{"key":"e_1_3_8_33_2","doi-asserted-by":"publisher","DOI":"10.1177\/09287329241296438"},{"key":"e_1_3_8_34_2","doi-asserted-by":"publisher","DOI":"10.1007\/s12672-026-04491-6"},{"key":"e_1_3_8_35_2","doi-asserted-by":"publisher","DOI":"10.3390\/diagnostics12081812"},{"key":"e_1_3_8_36_2","doi-asserted-by":"publisher","DOI":"10.3390\/curroncol31110486"}],"container-title":["Technology in Cancer Research &amp; Treatment"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/journals.sagepub.com\/doi\/pdf\/10.1177\/15330338261465522","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/journals.sagepub.com\/doi\/full-xml\/10.1177\/15330338261465522","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/journals.sagepub.com\/doi\/pdf\/10.1177\/15330338261465522","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,6]],"date-time":"2026-07-06T11:46:22Z","timestamp":1783338382000},"score":1,"resource":{"primary":{"URL":"https:\/\/journals.sagepub.com\/doi\/10.1177\/15330338261465522"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7]]},"references-count":35,"alternative-id":["10.1177\/15330338261465522"],"URL":"https:\/\/doi.org\/10.1177\/15330338261465522","relation":{},"ISSN":["1533-0346","1533-0338"],"issn-type":[{"value":"1533-0346","type":"print"},{"value":"1533-0338","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7]]},"article-number":"15330338261465522"}}