Hello, I am

Mazharul Hossain

AI Researcher in Computational Imaging & Scientific Machine Learning

I develop robust, generalizable, physics-informed machine learning methods for scientific and biomedical imaging.

Explore my research
Portrait of Mazharul Hossain

Computational imaging · Robust ML · Scientific AI

My Research

I develop machine learning methods for scientific images, with a focus on reliability when data or measurement conditions change.

Robust & Generalizable ML

Learning methods that account for shifts across sensors, scenes, and data distributions.

Computational Imaging

Reconstruction methods informed by how images are measured and formed.

Biomedical & Scientific AI

Machine learning for scientific and biomedical imaging questions.

Selected Research

Conceptual illustration of continuous pose refinement in SIMPLE; not an experimental result from this work. Example reference volume: human 80S ribosome, EMDB EMD-2938, Khatter et al. (2015), CC0. Source: https://www.ebi.ac.uk/emdb/EMD-2938.

Continuous Pose Refinement for Cryo-EM

Cryo-EM reconstruction depends on estimating the orientations and image-plane shifts of noisy particle images relative to a reference volume.
Continuous in-plane polishing jointly refines two image-plane shifts and one rotation after discrete matching. Cartesian pose refinement extends this to five parameters—two shifts and three rotational degrees of freedom—using Levenberg–Marquardt optimization. A dedicated local 3D workflow refines existing particle poses against saved reference maps, with optional Gaussian shift and Matrix–Fisher rotation priors to regularize translational and rotational updates.
Public source code and numerical and workflow regression tests document these research-software contributions. Experimental reconstruction-quality results are not reported here.
UPIGAN discriminator and generator architecture, including the microscope point spread function. Dajkhosh, Hossain, and Preza, Computational Imaging VII, SPIE, 2023, Figure 3, doi:10.1117/12.2663268.

Physics-Informed Reconstruction for 3D Microscopy

Reconstructing 3D structured-illumination microscopy images from noisy measurements requires both fine detail and fidelity to how the microscope forms an image.
USR combines a learned image prior with repeated corrections based on the microscope point spread function. UPIGAN extends this physics-informed, unrolled approach with adversarial learning to improve reconstruction.
Both studies used 18 fixed-cell mitochondrial samples. Their visual comparisons and reported PSNR and SSIM plots favor physics-informed unrolling over the respective non-physics baselines, but neither paper reports an independent external test.
Map901 project website banner. Source: map901.memphis.edu.

Map901: Indoor Hazard Mapping for Public Safety

Emergency responders need more than floor plans: they need to know where safety-critical objects are inside complex buildings.
The Map901 team combined backpack LiDAR and 360° video, segmented safety objects in images, transferred their labels into 3D point clouds, and stitched and georeferenced the building scans.
The studies surveyed seven Memphis buildings totaling 1.86 million square feet and addressed 30 classes of public-safety objects. The expanded study used 2,964 labeled images; its 3D maps still required manual cleanup of some labels.

Selected Publications

Representative work in computational imaging, robust machine learning, and scientific AI.

  1. 2026
  2. 2025
  3. 2024

    Greedy Ensemble Hyperspectral Anomaly Detection

    M. Hossain et al.

    Journal of Imaging

  4. 2023
  5. 2022

    Building Rich Interior Hazard Maps for Public Safety

    M. Hossain et al.

    Springer Communications in Computer and Information Science

Experience

  1. National Cancer Institute, National Institutes of Health

    Visiting Fellow

    I develop and validate computational methods and software for cryo-electron microscopy.

  2. University of Memphis

    Graduate Assistant

    I researched hyperspectral imaging, LiDAR indoor mapping, microscopy reconstruction, and responsible AI.

  3. Neuron23

    Data Scientist Intern

    I developed compound-modeling software and interpretable molecular-property models.

About Me

I am Mazharul Hossain. My background brings together computer science, scientific imaging, and research software development. I earned my Ph.D. in Computer Science at the University of Memphis, where my dissertation investigated target detection in hyperspectral images for aided target recognition.

My master’s work at the University of Memphis focused on indoor mapping for public safety. It brought together LiDAR, 360° imagery, and object detection to describe hazards inside buildings. My subsequent doctoral work explored hyperspectral image analysis, alongside work on physics-guided microscopy reconstruction.

I have also studied explainable AI, with an emphasis on making model requirements, assumptions, and evaluation understandable to the people who use and assess learning systems. My technical background includes machine-learning frameworks, scientific computing on HPC systems, and software services; the skills listed here summarize those tools and methods.

Outside of research, I enjoy hiking, exploring nature, and cycling through cities. Biking lets me experience everyday life at a closer pace and notice things I might miss when traveling by car.

Notes on research, engineering, and learning

Loading recent writing…

View all writing

Contact

Let’s Connect

I welcome conversations about computational imaging, scientific machine learning, and research collaboration.