Clinical Scorecard: Radiomics Utilizing Deep Learning Fails to Enhance Prediction of Residual Cancer Burden Following Chemotherapy in the LIMA Breast MRI Study

At a Glance

Key Highlights

• Deep learning-based radiomics did not improve prediction of residual cancer burden compared to standard clinical predictors such as tumor volume and subtype.
• Dynamic contrast-enhanced MRI is sensitive for breast lesion visualization but lacks specificity to reliably identify pathological complete response.
• Current imaging and biopsy techniques are insufficiently accurate to safely omit surgery based on predicted tumor response.

Guideline-Based Recommendations

Diagnosis

• Use DCE-MRI at baseline and post-neoadjuvant chemotherapy to assess tumor response.
• Classify residual cancer burden using histopathological evaluation post-surgery as the reference standard.

Management

• Continue standard surgical resection after neoadjuvant chemotherapy regardless of imaging-based predictions due to insufficient specificity.
• Consider tumor subtype and volume as key clinical predictors in treatment planning.

Monitoring & Follow-up

• Perform imaging before and after neoadjuvant chemotherapy to monitor tumor response.
• Use pathological assessment of resection specimens to confirm residual disease.

Risks

• Avoid omitting surgery based solely on imaging or biopsy predictions due to risk of sampling error and inaccurate response assessment.

Patient & Prescribing Data

Patients with invasive breast cancer undergoing neoadjuvant chemotherapy

No evidence that deep learning radiomics improves prediction of residual cancer burden to guide treatment modifications or surgery omission.

Clinical Best Practices

• Employ standardized DCE-MRI protocols with fat suppression and multiple post-contrast acquisitions for imaging consistency.
• Use centralized, expert pathological review to classify residual cancer burden accurately.
• Integrate clinical tumor subtype and volume data with imaging findings for comprehensive response assessment.
• Maintain surgical resection as standard care post-neoadjuvant chemotherapy until reliable non-invasive predictors are validated.

References

• Symmans et al. Residual Cancer Burden Classification
• LIMA Study Protocol
• Breast Imaging Reporting and Data System (BI-RADS)