Home / Health / AI Detects Breast Cancer Risk from Images
AI Detects Breast Cancer Risk from Images
26 Oct
Summary
- AI assay showed high agreement with MammaPrint genomic test.
- AI risk classification independently predicted survival.
- AI can analyze routine pathology slides for tumor biology.
An innovative AI-driven digital pathology approach for assessing the MammaPrint 70-gene recurrence signature in early-stage breast cancer was presented. This deep learning pipeline analyzes standard hematoxylin and eosin-stained pathology slides to predict a patient's risk of recurrence. The investigational assay demonstrated a high concordance rate of 0.954 when compared to the MammaPrint genomic test.
In a validation study of 372 patients with a median follow-up of 10.3 years, the AI classified 49% as high-risk and 51% as low-risk. Patients identified as low-risk by the AI exhibited a remarkable 10-year breast cancer-specific survival rate of 98.1%, significantly higher than the 86.7% survival rate for high-risk patients. The AI's digital risk classification proved to be the only independent predictor of breast cancer-specific survival in multivariable analysis, outperforming traditional clinical factors.
This AI-driven digital pathology has the potential to broaden access to crucial genomic information for breast cancer patients worldwide, especially in areas with limited access to tissue-based genomic testing. Agendia, a leader in precision oncology, developed this model, integrating decades of genomic science with advanced digital pathology capabilities to better understand tumor biology and inform personalized treatment strategies.
