Ph.D. student in Biomedical Engineering, passionate about computational biomedicine, scientific machine learning, and biomarker discovery. Experienced in designing foundation-guided, physics-informed latent diffusion architectures; high-throughput IHC cellular quantification (Roche); anatomy-aware self-supervised frameworks (VSI Lab); and machine learning-based neural biomarker predictive workflows (Telkes Lab). Demonstrated track record of leading clinical engineering initiatives and collaborating with the WHO; proficient in English, Turkish, Azerbaijani, and Persian.
Ph.D., Biomedical Engineering, The University of Arizona, Tucson, AZ, USA | 2024 – Present
MorphoStain: Multiplex Stain Decomposition in Alzheimer's Neuropathology [Repo]
Lead Researcher, The University of Arizona, Tucson, AZ, USA | Fall 2026
To resolve multiplex IHC stain unmixing and dual-DAB Amyloid-β/Tau disentanglement in Alzheimer’s neuropathology, I developed a physics-guided teacher–student generative framework. Coupling a Phikon-conditioned Beer–Lambert Latent Diffusion Teacher with Monte Carlo sampling achieved calibrated uncertainty (ECE = 0.0125), while symmetric morphological routing distilled the 169.67M teacher (>1,400×) into a real-time 0.12M-parameter FastStudentUNet (140.1 FPS, 40.56 dB PSNR) with strong zero-shot HER2 IHC generalization (38.45 dB) and illumination drift resilience (+11.64 dB over SOTA).
HemQ: Automated Cellular Quantification Tool for Dialable LOTUS Hematoxylin
Computational Pathology Intern: Pathology Lab Research & Early Development, Roche, Tucson, USA | Summer 2026
To support the Roche agile team’s development of a customizable Hematoxylin counterstain reagent for the LOTUS platform, I engineered the core computational pipeline for automated tissue segmentation, stain unmixing, and the extraction of nuclear and non-nuclear features to drive pathologist-aligned scoring and reduce diagnostic variability. Building upon this automated quantification tool designed to optimize laboratory efficiency and lower operational costs, I further integrated CPU-bound performance enhancements—including parallel computing, partial patch processing, and coarse-to-fine analysis—to ensure efficient, high-throughput segmentation across standard laboratory workstations without requiring dedicated GPU infrastructure.
ViT-2SPN: Dual-Stream Self-Supervised Pretraining Tool for Retinal OCT Classification
Graduate Research Assistant: VSI Lab, ECE Department, University of Arizona, Tucson, USA | 2024 – 2025
My role designed and developed ViT-2SPN, an anatomy-aware self-supervised framework for retinal OCT classification. By integrating ophthalmic foundation models, asymmetric knowledge distillation, dual-stream cross-attention, and axial-aware positional encoding with a triple-objective loss (global contrastive, local patch alignment, and frequency reconstruction).
Machine Learning-Based Neural Biomarker Predictive Workflow for Spinal Cord Stimulation
Graduate Research Assistant: Telkes Lab, BME & Neurosurgery Department, University of Arizona, Tucson, USA | 2025 – Present
My role architected explainable machine learning and signal processing workflows to uncover objective neural biomarkers of spinal cord stimulation response, incorporating graph neural networks for spectral connectivity analysis and phase-amplitude coupling.
Medical Device Lifecycle Management and Clinical Data Analysis
Supervisor, Clinical Engineer: Tabriz University of Medical Sciences, East Azerbaijan, Tabriz, Iran | 2014 – 2024
My role involved implementing machine learning frameworks incorporating feature extraction and ensemble methods for multimodal clinical data analysis, while leading clinical engineering teams in overseeing capital and disposable medical device management, operational training, preventive maintenance, statistical analysis, clinical readiness, and collaboration with the WHO on COVID-19 response to ensure safe, efficient, and effective healthcare operations.
Undergraduate Teaching Assistant: East Azerbaijan, Tabriz, Iran
- [BME 090] Introduction to Clinical Engineering, Tabriz University – Dr. Sebelan Danishvar | 2016 – 2017
- [BME 020] Equipment of Hospitals & Medical Centers, Islamic Azad University of Tabriz – Dr. Hashemiaghdam | 2013 – 2014
- [BME 006-8] Computer Programming & Algorithm Calculus, Islamic Azad University of Tabriz – Dr. Rajabioun | 2012 – 2013
- Saraei, M., et al. (2026). MorphoStain: Physics-Guided Latent Diffusion and Morphological Distillation for Multiplex Stain Decomposition in Alzheimer's Neuropathology. Under review in IEEE J. Biomed. Health Inform.
- Saraei, M., Lee, E.J., & Lalinia, M. (2025). Deep Learning-Based Medical Object Detection: A Survey. IEEE Access (EMBS), 13, 53019–53038. 🔗 DOI
- Saraei, M., & Liu, S. (2023). Attention-Based Deep Learning Approaches in Brain Tumor Image Analysis: A Mini Review. Front. Health Inform., 12, 164. 🔗 DOI
- Saraei, M., Rahmani, S., Rajebi, S., & Danishvar, S. (2023). A Different Traditional Approach for Automatic Comparative Machine Learning in Multimodal COVID-19 Severity Recognition. Int. J. Innov. Eng., 3(1), 1–12. 🔗 DOI
Peer-Reviewer:
- IEEE Access, New York, NY, USA | 2025 – Present
- CaPTion @ MICCAI, Strasboug, France | 2026
- MIUA, Dublin, Ireland | 2026
[Research Assistantship]: Supported by the Telkes Lab, University of Arizona ($71,158) | 2025 - Present
[Herbold Fellowship]: Awarded by the College of Engineering, University of Arizona ($58,470) | 2024 - 2025
© 2024-2026 Reza · All rights reserved!
📧 mrsaraei@arizona.edu | mrsaraei@yahoo.com