Doctoral research · Medical AI
Multimodal Survival Prediction in Multiple Myeloma
Developing an end-to-end survival modeling pipeline that connects bone-marrow single-cell morphology with clinical variables through self-supervised vision encoders, multiple-instance learning, Cox models, and late fusion.
Role: doctoral researcher and first author; responsible for study framing, modeling, evaluation design, and research communication.
First-author doctoral research integrating bone-marrow single-cell morphology with clinical variables to predict overall survival in newly diagnosed multiple myeloma.
- 01Bone-marrow cell morphology images and clinical variables
- 02Self-supervised vision transformer embedding pipeline
- 03Plasma-cell / plasmablast classifier and malignancy index
- 04Multiple-instance-learning survival model over patient cell bags
- 05Late fusion with clinical Cox model for final risk prediction
Problem
Cancer prognosis depends on both microscopic morphology and patient-level clinical variables. The technical challenge is to convert millions of single-cell images into patient-level survival signals without losing clinically relevant heterogeneity.
Contribution
Designed an end-to-end multimodal pipeline that uses self-supervised cell embeddings, a malignancy index, multiple-instance survival modeling, and late fusion with clinical Cox models.
Engineering
Built Python/PyTorch feature pipelines with HDF5 storage, 384-dimensional image embeddings, patient-level bag construction, survival metrics, and interpretable risk stratification.