Research Engineer | PhD Applicant in AI for Healthcare
Multimodal AI for healthcare, medical image analysis, computational pathology, and spatial transcriptomics.
I am a Research Engineer at Institut Curie - PSL Research University in Paris, where I develop computational pipelines using deep learning and statistical methods to study the tumor microenvironment from histopathology and spatial transcriptomics data. My research focuses on developing multimodal AI, integrating radiology, digital pathology, multi-omics, and clinical information to advance personalized healthcare.
I hold two european master's degrees: an Erasmus+ Joint Master's in Medical Imaging and AI (MaIA), and a second Master's degree in Bioinformatics from Sorbonne University. My graduate training spans at top institutions in France, Italy, and Spain, and I have contributed to impactful research at Institut Curie, CNRS, and RadboudUMC. I completed a Bachelor of Science degree in Electrical Engineering from UIU in Bangladesh, graduating Summa Cum Laude and receiving the Vice-Chancellor's Award.
My academic and research training has been supported by competitive funding, including the EU-funded Erasmus Mundus scholarship, the Acquisis International Mobility Grant from the RBFC, and research internships at CNRS and Institut Curie. My work has been recognized with distinctions, including the Best Thesis Award. Beyond research, I am actively involved in academic and professional communities through the EMA and IEEE.
Research
My current research focuses on multimodal AI for healthcare, aiming to advance diagnosis, prognosis, and precision medicine. I develop deep learning methods that integrate radiomic features, tissue morphology, molecular information, and clinical records from real patient samples to support clinical decision-making. I am particularly interested in key challenges in multimodal learning, including fusion strategies, missing modalities, uncertainty estimation, and robustness in real-world clinical settings. More broadly, my work explores weakly supervised and self-supervised learning under limited-label conditions, alongside core medical image analysis tasks such as segmentation, registration, reconstruction, and AI-assisted diagnosis.
Deep learning for medical image reconstruction, registration, segmentation, quantitative imaging, and AI-assisted methods for image-guided intervention and diagnosis.
AI-driven analysis of whole-slide histopathology images using weakly supervised learning, multiple instance learning, and foundation models to identify biomarkers, characterize the tumor microenvironment, and support precision oncology.
Development of computational pipelines for Visium HD and other spatial omics platforms, including cell-type deconvolution, spatial tissue profiling, and integration of histology with transcriptomic and multi-omics data for molecularly informed tissue mapping.
Experience
Institut Curie - PSL Research University, Paris, France
Leading development of multimodal computational pipelines integrating histopathology and spatial transcriptomics to study tumor microenvironment dynamics. Collaborating with clinical partners and preparing manuscripts.
Institut Curie / Sanofi collaboration, Paris, France
Developed analysis pipelines for Visium V2 and Visium HD spatial transcriptomics data. Identified eight candidate genes potentially mediating CAF-driven T cell exclusion in lung squamous cell carcinoma. Manuscript in preparation.
National Center for Scientific Research (CNRS), France
Benchmarked fully supervised, weakly supervised, and self-supervised approaches for HRD detection in breast and ovarian cancer whole-slide images. Best model achieved AUC 0.78 on breast cancer and 0.68 on ovarian cancer.
Diagnostic Image Analysis Group, Radboud University Medical Center, Netherlands
Developed preprocessing pipelines for k-space undersampling and evaluated deep learning models for real-time MRI reconstruction in interventional radiology.
Education
Sorbonne University, France | Grade: 14.85/20 | Rank: 1st of 8 | Best Thesis Award
Thesis: Identification of Tumor Gene Signatures Underlying Fibroblast-Mediated T Cell Exclusion in Lung Cancer Using Spatial Transcriptomics.
ETH Zurich & University of Zurich, Switzerland | 4 ECTS
University of Girona, University of Cassino, and University of Burgundy | Grade: 8.30/10
Fully funded by the European Union (EUR 42,000).
United International University, Bangladesh | GPA: 3.97/4.00 | Summa Cum Laude | Rank: 1st of 120
Preprints & Manuscripts
Hossain M.I. et al. Institut Curie / Sanofi collaboration. Manuscript in preparation, 2026.
Hossain M.I., Amin M.Z., et al. arXiv:2411.05456
Tweneboah A.D., Hossain M.I. (co-author), et al. arXiv:2502.19258
Projects
Visium HD pipeline for identifying CAF-mediated T cell exclusion signatures in lung squamous cell carcinoma. Identified 8 candidate genes. Manuscript in preparation.
Benchmarked AB-MIL, CLAM, Trans-MIL, and foundation models for homologous recombination deficiency detection. Achieved AUC 0.78 on breast cancer whole-slide images.
ArXiv preprint
nnU-Net pipeline for automated echocardiogram segmentation. Dice scores: LV endocardium 0.94, LV epicardium 0.91, left atrium 0.93.
XGBoost pipeline integrating DNA methylation, CNV, mRNA, and miRNA data for breast cancer subtyping. Balanced multiclass accuracy: 0.90.
