Artificial intelligence and triple negative breast cancer: can pathomics help oncologists?
Abstract
Triple-negative breast cancer (TNBC) is an aggressive and heterogeneous subtype of breast cancer with limited targeted therapies and poor prognosis. Recent advances in digital pathology and artificial intelligence (AI) are transforming traditional, qualitative histopathology into quantitative, data-driven analysis. Within computational pathology, pathomics (the extraction of handcrafted, high-dimensional features from digitized histological images) and deep learning (DL)-based whole-slide image analysis have been investigated for the prediction of molecular subtypes, biomarker expression, tumor-infiltrating lymphocytes, and treatment response in TNBC. AI models, including DL and multimodal systems integrating histopathology, omics, and clinical data, are being explored for patient stratification, risk assessment, and therapy personalization, encompassing chemotherapy, immunotherapy, PARP inhibitors, and antibody-drug conjugates. By linking spatial tissue architecture with molecular profiles, these approaches may provide new insights into tumor biology and the tumor microenvironment. Most TNBC-specific studies are retrospective, small or single-center, few have been externally validated and, to our knowledge, none has yet demonstrated clinical utility in prospective trials; prediction of biomarkers such as PD-L1 or homologous recombination deficiency from hematoxylin and eosin slides remains investigational and cannot replace validated testing. Despite these challenges, AI-assisted computational pathology is a promising decision-support tool, which complements and does not replace pathologists and oncologists, for enhancing diagnosis, prognosis, and therapeutic decision-making in TNBC.
Related
- Identifying barriers to immunotherapy access for triple-negative breast cancer: A qualitative analysis of interviews with medical oncologists
- Negative Appendicectomy Rate and Surgery Despite Negative Triple-Screen Inflammatory Markers in Children: A Single-Centre Retrospective Cohort Study
- MIF-mediated intercellular crosstalk drives antigen-presenting-like cancer-associated fibroblast differentiation in triple-negative breast cancer via TGF-β signaling
- 838 ETV6 rearrangements impair T cell infiltration via activating β-Catenin signaling in triple-negative breast cancer
- The Prognostic Role of Vimentin Expression in Triple‐Negative Breast Cancer: A Retrospective Cohort Study
BibTeX
@article{faa2026artificial,
title = {{Artificial intelligence and triple negative breast cancer: can pathomics help oncologists?}},
author = {Gavino Faa and Eleonora Lai and Jasjit S. Suri and Cinzia Solinas and Mariele Dessì and Giorgio Saba and Francesco Loi and Ekta Tiwari and Flaviana Cau and Massimo Dessena and Pina Ziranu and Andrea Pretta and Claudia Codipietro and Simona Graziano and Luca Saba and Mario Scartozzi},
journal = {Frontiers in Oncology},
year = {2026},
month = nov,
doi = {10.3389/fonc.2026.1883554},
url = {https://www.frontiersin.org/articles/10.3389/fonc.2026.1883554},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Artificial intelligence and triple negative breast cancer: can pathomics help oncologists?
AU - Gavino Faa
AU - Eleonora Lai
AU - Jasjit S. Suri
AU - Cinzia Solinas
AU - Mariele Dessì
AU - Giorgio Saba
AU - Francesco Loi
AU - Ekta Tiwari
AU - Flaviana Cau
AU - Massimo Dessena
AU - Pina Ziranu
AU - Andrea Pretta
AU - Claudia Codipietro
AU - Simona Graziano
AU - Luca Saba
AU - Mario Scartozzi
PY - 2026
DA - 2026/11/05
JO - Frontiers in Oncology
DO - 10.3389/fonc.2026.1883554
UR - https://www.frontiersin.org/articles/10.3389/fonc.2026.1883554
AB - Triple-negative breast cancer (TNBC) is an aggressive and heterogeneous subtype of breast cancer with limited targeted therapies and poor prognosis. Recent advances in digital pathology and artificial intelligence (AI) are transforming traditional, qualitative histopathology into quantitative, data-driven analysis. Within computational pathology, pathomics (the extraction of handcrafted, high-dimensional features from digitized histological images) and deep learning (DL)-based whole-slide image analysis have been investigated for the prediction of molecular subtypes, biomarker expression, tumor-infiltrating lymphocytes, and treatment response in TNBC. AI models, including DL and multimodal systems integrating histopathology, omics, and clinical data, are being explored for patient stratification, risk assessment, and therapy personalization, encompassing chemotherapy, immunotherapy, PARP inhibitors, and antibody-drug conjugates. By linking spatial tissue architecture with molecular profiles, these approaches may provide new insights into tumor biology and the tumor microenvironment. Most TNBC-specific studies are retrospective, small or single-center, few have been externally validated and, to our knowledge, none has yet demonstrated clinical utility in prospective trials; prediction of biomarkers such as PD-L1 or homologous recombination deficiency from hematoxylin and eosin slides remains investigational and cannot replace validated testing. Despite these challenges, AI-assisted computational pathology is a promising decision-support tool, which complements and does not replace pathologists and oncologists, for enhancing diagnosis, prognosis, and therapeutic decision-making in TNBC.
SN - 2234-943X
ER -