Memorial Sloan Kettering Cancer Center
Research Assistant, Advanced Computing and Oncology Laboratory · PI: Tafadzwa L. Chaunzwa, MD
Aug 2026–Present
Graduate Research Assistant, Mar 2025–Jul 2026
- Conceived and engineered HostBio (Comp-MRF), an interpretable, scan-adaptive probabilistic framework for recovering host-tissue composition from routine CT by translating mixture-deconvolution concepts from spatial transcriptomics into quantitative imaging; developed the framework into a tool adopted for institution-wide clinical research use at MSK via MIM Development.
- Designed multimodal translational studies integrating quantitative imaging with targeted tumor and cfDNA sequencing, bulk and single-cell transcriptomics, flow cytometry, cytokine profiling, treatment exposure, and longitudinal outcomes; validated HostBio-derived thymic phenotypes across 1,078 patients in four independent NSCLC cohorts, including randomized RTOG-0617, identifying treatment-context-dependent associations with survival and toxicity and linking lower thymic reserve to persistent suppressive/exhaustion immune programs and impaired treatment-emergent CD8+ proliferation.
- Applied deep-learning models to thoracic imaging, supporting training and deployment of an nnU-Net v2 model for automated thymus segmentation and evaluating the pretrained Sybil model for CT-based lung-cancer risk prediction in 77 women; predicted risk significantly separated cases from controls across all six prediction horizons.
- Developed longitudinal quantitative-imaging approaches for serial cone-beam CT and tumor trajectories using radiomics, longitudinal-survival modeling, and dynamic prediction; contributed the MSK clinical-imaging arm to a collaborative mechanistic-modeling and machine-learning study of radiation-induced tumor migration and treatment scheduling.