Summary
This technology research and development (TRD) project focuses on developing advanced computational and machine learning methods for image formation and analysis in high-resolution label-free optical microscopy, aiming to improve computational specificity, semantic segmentation, spatial resolution, and 3D imaging.
What they want
The project will develop advanced computational and machine learning methods to address needs in image formation and analysis for high-resolution label-free optical microscopy. It integrates imaging science, physics-based, and deep learning (DL)-based approaches to overcome limitations of label-free imaging, using objective image quality measures for validation. Three classes of methods will be investigated: (1) image-to-image mapping for computational specificity, semantic segmentation, and enhanced spatial resolution; (2) improved reconstruction for 3D cellular imaging using diffraction tomography and inverse scattering; and (3) extraction of biologically relevant information from multi-modality label-free image data, including biomarker discovery and multi-modal DL methods. The developed methods will serve as enabling technologies for projects within a proposed P41 center and will be jointly developed and evaluated with TRD and driving biological projects.
Deliverables
- Computational methods for label-free imaging technologies
- Deep learning methods for label-free imaging technologies
- Improved computational staining methods
- Methods for enhanced spatial resolution
- Methods for semantic segmentation
- Methods for 3D image formation
- Methods for analysis of multi-modality label-free image data
- All source code (open-source)
- Trained models (open-source)
- Documentation (open-source)
Technical requirements
- Advanced computational methods
- Machine learning methods
- Imaging science integration
- Physics-based approaches
- Deep learning (DL)-based approaches
- Objective image quality measures for validation
