Artificial intelligence for triple-negative breast cancer from imaging to multi-omics

Area of Science:

• Oncology
• Medical Imaging
• Computational Pathology
• Genomics
• Artificial Intelligence

Background:

• Triple-negative breast cancer (TNBC) is an aggressive subtype with limited diagnostic and prognostic biomarkers.
• Artificial intelligence (AI) is emerging as a tool to analyze radiology, digital pathology, and molecular data in TNBC.

Purpose of the Study:

• To synthesize existing literature on AI applications in TNBC.
• To evaluate AI's use across various data modalities and clinical tasks.
• To identify challenges and future directions for AI in TNBC research and clinical practice.

Main Methods:

• A structured narrative synthesis of published studies on AI for TNBC.
• Literature organized by data modality (imaging, pathology, genomics, multi-omics), clinical task, and validation strategy.
• Emphasis on external validation, calibration, interpretability, and handling of missing data.

Main Results:

• AI is applied to lesion segmentation, subtype classification, treatment response prediction, recurrence risk stratification, and survival modeling.
• Multimodal fusion and radiogenomic approaches show potential for capturing TNBC heterogeneity.
• Current evidence is limited by small cohorts, inconsistent definitions, inadequate external testing, and domain shift.

Conclusions:

• Clinically credible AI studies in TNBC require alignment with actionable decisions, robust validation, transparent reporting, and biological interpretability.
• Future progress necessitates multi-institutional data curation, advanced pretraining methods, privacy-preserving collaboration, and robust multimodal designs.
• Addressing real-world distribution shifts and missing data is critical for reliable AI deployment in TNBC.