Clinical Scorecard: Utilizing deep learning for the identification and measurement of brain metastases in black-blood imaging to inform treatment options: a clinical cohort analysis

At a Glance

Key Highlights

• Deep learning systems incorporating black-blood CE-T1WI improve sensitivity (82.4%) for detecting small (<3 mm) brain metastases with low false-positive rates (0.59 per patient).
• Volumetric and numeric assessment of brain metastases by DLS can inform treatment decisions aligned with ASCO-SNO-ASTRO and EANO-ESMO guidelines.
• Treatment options suggested based on DLS output include follow-up imaging, stereotactic radiosurgery (SRS), surgery, whole-brain radiotherapy (WBRT), and systemic chemotherapy.

Guideline-Based Recommendations

Diagnosis

• Use combined gradient-echo and black-blood CE-T1WI MRI protocols for improved detection and segmentation of brain metastases.
• Apply voxel-wise segmentation via deep learning for accurate volumetry of brain metastases.

Management

• Consider number and volume of brain metastases and patient condition when deciding treatment.
• Patients with ≤2 BMs ≤5 mm diameter (volume ≤65 mm3) are recommended for short-term imaging follow-up without immediate treatment.
• Patients with up to 10 brain metastases may be candidates for stereotactic radiosurgery (SRS).
• Other treatment options include surgery, whole-brain radiotherapy (WBRT), and systemic chemotherapy based on clinical guidelines.

Monitoring & Follow-up

• Perform follow-up MRI scans for small lesions (<5 mm) to confirm true metastases before treatment initiation.

Risks

• False positives in BM detection can occur; combining imaging modalities reduces this risk.
• Treatment decisions should consider patient-specific factors and tumor characteristics to avoid overtreatment.

Patient & Prescribing Data

112 patients with newly diagnosed brain metastases from solid tumors excluding small cell lung cancer and prior WBRT

Deep learning-based detection and volumetric quantification of brain metastases can guide personalized treatment planning according to established oncology guidelines.

Clinical Best Practices

• Incorporate black-blood CE-T1WI in MRI protocols to reduce false positives from enhancing vessels.
• Use deep learning systems for automated segmentation and volumetric analysis to assist radiologists and oncologists.
• Apply guideline-based algorithms integrating number and volume of metastases to recommend appropriate treatment or follow-up.
• Exclude patients with small cell lung cancer or prior whole-brain radiotherapy from standard BM treatment algorithms due to differing management.
• Ensure multidisciplinary review of imaging and clinical data for optimal treatment planning.

References

• ASCO-SNO-ASTRO Practice Guidelines for Brain Metastases
• EANO-ESMO Clinical Practice Guidelines for Brain Metastases
• RANO-BM Response Assessment Criteria