Predicting Breast Cancer Therapy Response with Longitudinal MRI and Immune Data

A study from 2026 by Lan Yan developed a multimodal model for predicting pathological complete response to neoadjuvant therapy in breast cancer. The model integrates deep learning features from longitudinal DCE-MRI, peripheral blood inflammatory indices, and baseline tumor-infiltrating lymphocytes. It was trained and validated on data from 262 retrospectively enrolled patients.

Use Cases

• Training predictive models for pathological complete response based on integrated deep learning and immune-inflammation features.
• Benchmarking single-modality models (e.g., imaging-only or blood-index-only) against a multimodal fusion approach.
• Investigating the predictive value of early-treatment DCE-MRI features compared to pre-treatment baselines.
• Developing tools for early assessment of neoadjuvant therapy efficacy to aid clinical decision-making.

Strengths

• Model validation on a cohort of 79 patients, achieving an AUC of 0.90 for the combined model.
• Integration of three distinct data modalities: longitudinal MRI, blood indices, and tumor microenvironment data.
• Clear performance comparison between models, with the Post-2nd-NAT DL model (AUC=0.85) outperforming the Pre-NAT model (AUC=0.75).

Limitations

• The dataset is described in a 74.6 KB PDF document; the underlying raw data files are not directly available.
• Row count and column-level documentation for the underlying patient data are unknown.
• Data is from a retrospective study, which may introduce selection bias.

Provenance

Source

Lan Yan via figshare.

Collection Method

Retrospective collection from 262 breast cancer patients receiving neoadjuvant therapy.

Freshness

Last updated 2026-04-28 05:27:32.