AmirHessam Zare
Medical AI & Biostatistics
I work on clinical AI and biostatistical methods for oncology and neuroimaging data. My research focuses on developing interpretable machine learning approaches for survival prediction and treatment response modeling in neuro-oncology, with particular emphasis on radiomics and multi-center MRI data integration. I have also contributed review articles in neuro-oncology and neurosurgery.
Research Interests
Clinical Prediction Models
Survival analysis and prognostic modeling for cancer patients using radiomics, genomics, and clinical data. Focus on handling competing risks and time-varying covariates in real-world oncology settings.
Genomics & Multi-Omics
Bulk and single-cell RNA-seq analysis for tumor characterization and treatment prediction. Deconvolution methods, generative models for class imbalance, and integrative multi-omics approaches for oncology.
Statistical Methods
Sparse regression techniques, penalized survival models, and methods for high-dimensional biomedical data. Particular interest in handling missing data and measurement error in clinical datasets.
Recent Writing
Uncertainty-Aware Generative Oversampling: Introducing LEO-CVAE
Local Entropy-Guided Oversampling with Conditional Variational Autoencoders for addressing class imbalance in clinical genomics. Combining information theory with deep learning for improved synthetic data generation.
BayesPrismExt: Enhanced Chain Storage and CV Computation
Extended version of BayesPrism with practical improvements for MCMC chain storage and coefficient-of-variation calculations in bulk RNA-seq deconvolution workflows. HDF5-backed chain management.
What tumor purity does to your differential expression results
A practical look at how sample contamination affects RNA-seq analysis in TCGA data, with code for adjusting differential expression models.