Deep learning-based spatiotemporal estimation of lesion changes for patient-level assessment of breast cancer lung metastases on longitudinal CT.
AI can track how breast cancer lung tumors change over time without complex image alignment, improving how doctors assess treatment response.
A novel deep learning framework for patient-level assessment of breast cancer lung metastases on longitudinal CT avoids error-prone image registration while tracking multiple dynamic lesions over time. Multicenter validation across four centers demonstrates practical generalizability for treatment response monitoring beyond RECIST 1.1.
What the study was
- Study design
- Retrospective deep learning development and multicenter validation study
- Population
- Patients with breast cancer lung metastases; PUMCH dataset + multicenter validation across 4 medical centers
- Category
- Diagnostics
- Maturity
- Exploratory
- Journal
- NPJ Precision Oncology
Why it surfaced
NPJ Precision Oncol publication with multicenter validation (4 centers) addressing a practical limitation of current RECIST-based monitoring in metastatic breast cancer. Patient-level rather than lesion-level approach is clinically meaningful; registration-free design reduces implementation barriers.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.