Clinical Scorecard: A Deep Learning Approach for Classifying Grade 2 Nottingham Histologic Breast Tumors Using Dynamic Contrast-Enhanced MRI

At a Glance

Key Highlights

• NHG2 tumors represent a heterogeneous group with variable prognosis, complicating treatment decisions.
• Deep learning applied to DCE MRI enables non-invasive, cost-effective risk stratification of NHG2 tumors into prognostically distinct subgroups.
• DL-based classification provides additional prognostic value beyond conventional clinical predictors such as age, lymph node status, tumor stage, and molecular subtype.

Guideline-Based Recommendations

Diagnosis

• Use Nottingham Histologic Grade (NHG) to assess breast tumor differentiation based on tubule formation, nuclear pleomorphism, and mitotic rate.
• Incorporate DCE MRI with deep learning models to refine risk stratification of NHG2 tumors into NHG1-like or NHG3-like subgroups.

Management

• Tailor treatment intensity based on DL-based risk stratification: intensify therapy for higher-risk NHG3-like tumors and consider de-escalation for lower-risk NHG1-like tumors to avoid overtreatment.
• Integrate DL MRI assessment with existing clinical and molecular data for personalized treatment planning.

Monitoring & Follow-up

• Follow recurrence-free survival (RFS) outcomes to validate prognostic stratification.
• Utilize imaging and clinical follow-up data to monitor treatment response and disease progression.

Risks

• Potential for overtreatment or undertreatment due to heterogeneity within NHG2 tumors without refined risk stratification.
• Limitations of molecular profiling due to cost and time may restrict routine use; DL MRI offers a practical alternative.

Patient & Prescribing Data

Patients with biopsy-confirmed invasive breast cancer classified as NHG2

DL-based DCE MRI classification supports personalized treatment decisions by identifying NHG2 patients with higher or lower recurrence risk, guiding therapy intensity to optimize outcomes and minimize toxicity.

Clinical Best Practices

• Combine histopathological NHG assessment with DL-based DCE MRI analysis for comprehensive tumor grading.
• Use standardized MRI acquisition protocols (e.g., pre-contrast and multiple post-contrast T1-weighted sequences) to ensure data quality for DL modeling.
• Apply rigorous image preprocessing including normalization and tumor-focused cropping guided by expert radiologists.
• Validate DL models externally to confirm generalizability across different patient cohorts and imaging platforms.
• Incorporate DL risk stratification results alongside traditional clinical factors to inform multidisciplinary treatment planning.

References

• Nottingham Histologic Grade and Breast Cancer Prognosis
• Deep Learning in Breast MRI for Risk Stratification
• Duke Breast Cancer MRI Dataset
• Advanced MRI Breast Lesions Dataset