Trustworthy Multimodal Clinical AI
I evaluate calibration, uncertainty, patient-level harm, and subgroup performance across imaging and high-acuity care data.
View direction →AI scientist · Biostatistician · Applied mathematician
Habib Latifizadeh,
Ph.D.
I study when clinical AI can be trusted, where it fails, and how uncertainty should affect a decision.
My work combines biostatistics, causal inference, and mechanistic modeling across medical images, electronic health records, physiological data, and biological systems. I focus on patient-level reliability rather than average performance alone.
MIT Laboratory for Computational PhysiologyProject Co-Investigator & Visiting Scholar · 2026–2027
Duke UniversityLead Data Scientist & Biostatistician · Epilepsy Surgery Outcomes · 2026–2027
MIT Critical DataDuke UniversityNIH/NIAID R01 AI173333Editorial & Advisory ServiceAwards & Research Support
Research agenda
Reliable models begin with the decision they need to support.
Across imaging, electronic health records, and biological systems, I ask three practical questions: Where does the model fail? What does the evidence support? How should uncertainty change the decision?
Causal Inference & Patient-Specific Decisions
I use causal models and interpretable outcome analysis to study treatment response, patient selection, and prediction failure.
View direction →Mechanistic & Multiscale Scientific ML
I connect mechanistic simulation with machine learning, optimal transport, and probabilistic surrogates for biomedical systems.
View direction →Methodological coreCausal reasoningCalibrated uncertaintyMechanistic structureReproducible computation
Current research
Three programs focused on decisions, failure, and biological context.
Reliability & Harm Evaluation of Multimodal Clinical AI
At MIT Critical Data, I evaluate diagnostic-imaging model ensembles and link patient-level error patterns to ICU care phenotypes in MIMIC-CXR and MIMIC-IV.
CalibrationUncertaintySubgroup auditingMultimodal data
Surgical Decision Support for Epilepsy Outcomes
At Duke, I lead the statistical analysis of an SEEG-derived focality score and examine when it may help predict seizure freedom after epilepsy surgery.
ROC/AUCBootstrap inferenceCalibrationFailure analysis
Multiscale Scientific ML for Congenital CMV
In an NIH-supported Duke–Weill Cornell collaboration, I connect mechanistic CMV models with machine learning and cross-species immune-cell alignment.
Scientific MLOptimal transportGaussian processesMechanistic modeling
Selected publications
Published methods, open code, and work in progress.
2025Preprint
BaMANI: Bayesian Multi-Algorithm Causal Network Inference
H. Latifizadeh, A. C. Pirkey, A. Gould, D. J. Klinke II
First-author open framework for ensemble causal discovery, bootstrap stability, and edge-level uncertainty.
2022Article
Data-driven learning how oncogenic gene expression locally alters heterocellular networks
D. J. Klinke II, A. Fernandez, W. Deng, A. Razazan, H. Latifizadeh, A. C. Pirkey
Nature Communications. Computational analysis of how oncogenic gene expression changes tumor and immune-cell relationships.
2026In progress
Preoperative Focality Scoring for Predicting Seizure Freedom After Temporal-Lobe Epilepsy Surgery
H. Latifizadeh et al.
Statistical lead for a Duke study of the 5-SENSE focality score and postoperative seizure freedom.
Academic trajectory
A path from numerical analysis to clinical AI evaluation.
My training moved from nonlinear systems and numerical methods to causal network inference, computational biology, biostatistics, and patient-level model evaluation.
Research experience- MIT Laboratory for Computational PhysiologyProject Co-Investigator & Visiting Scholar
- Duke Comprehensive Epilepsy CenterLead Data Scientist & Biostatistician
- Duke Biostatistics & BioinformaticsPostdoctoral Associate
- NSF Bridges to Digital HealthResearch Fellow in AI & Machine Learning
- WVU Cancer InstituteMachine Learning Researcher
Editorial leadership
Editorial work grounded in statistical and clinical review.
I serve on boards for npj Digital Medicine, iScience, and the International Journal of Modeling, Simulation, and Scientific Computing. I have reviewed more than 50 manuscripts.
Collaboration & contact
Open to research conversations with a clear scientific question.
I welcome discussions on clinical AI evaluation, biostatistics, causal inference, computational biology, and mechanism-informed modeling.
Get in touch