D2IMDeformation from greyscale
Predicts bone deformation directly from X-ray tomography greyscale, without an explicit correlation step.
Extreme Mechanics Letters10.1016/j.eml.2024.102202
The work
Bone, soft tissue, bio-composites and self-assembled biomaterials — imaged under load, then modelled.
A tomogram is a measurement, not a picture. We build the models that read it.
The Predictive Imaging Lab is led by Gianluca Tozzi, Professor of Industrial Engineering at the University of Greenwich. Its research runs under Bio‑AImagiQ — Bio-Inspired Engineering with Imaging-based AI and Quantum data — an open initiative that brings together members of the lab and affiliated researchers across the Faculty and beyond.
The group advances the understanding of biological tissues and biomaterials by integrating advanced imaging (in‑situ X-ray tomography, hyperspectral imaging) with full-field measurement such as digital volume correlation, and with artificial intelligence.
The result is a pipeline that runs from the specimen on the loading rig to a trained model that predicts how a structure deforms, where a defect sits, or which formulation will assemble.
Models
Published architectures
Named models from the lab, with the papers behind them.
DiffSpectralNetDiffusion + HSI
Improves classification accuracy in hyperspectral imaging using a diffusion-based representation.
Scientific Reports10.1038/s41598-024-58125-4MedDiffHSITumour detection
Improves tumour detection from hyperspectral images of biological tissue.
Journal of Microscopy10.1111/jmi.13372
Active projects
What is running now
Six programmes, each led by a member or affiliate of the lab.
Imaging-based AI models for tissue mechanics
Fully developing D2IM by integrating additional AI strategies, advanced XCT segmentation models and the power of large language models.
LeadJon Valijonov
AI models for advanced biomaterial design
Applying advanced AI models to analyse, predict and optimise the hierarchical self-assembly of nanoparticles into macroscale soft biomaterials.
LeadMoeen Mohammady
Hyperspectral imaging and AI for biological tissues
Exploring the unique spectral signature of biological tissues and developing advanced AI models for healthcare.
LeadNeetu Sigger
Imaging-based AI models for bio-composites
Developing D2IM-based tools that give the industrial partner a technology to identify the nature of defects in bio-composites and inform manufacturing.
KTP-10108115Pinelopi Almpantaki
Quantum-AI synergy for next-generation imaging
Developing quantum-native representations of biological tissue images, enabling efficient classification, segmentation, measurement and multimodal fusion.
LeadIsabella Florez
Multi-dimensional sonification of bone XCT images
A cross-faculty collaboration investigating AI sonification to design and optimise bone biomimetic materials, blending imaging, computational modelling and additive manufacturing.
CollaborationEngineering & Science
Facility
The imaging bay
Faculty of Engineering and Science, University of Greenwich.
InCiTe 3D X-ray Microscope
KA Imaging
Sub-micron resolution with fast phase-contrast acquisition, paired with a custom-designed in‑situ loading device so specimens can be imaged while they deform.
First and only of its kind in EuropeResolution< 1 µm
ContrastPhase + absorption
In situCustom loading rig
Hyperspectral Imaging
Living Optics
Spectral signatures that reveal the biochemical state of tissue alongside its structure — the input to our HSI classification and tumour-detection models.
SignalSpectral cube
ReadsBiochemical state
News & opportunities
Latest
Call for papers · Journal of Microscopy
Special issue: AI in Imaging
The Journal of Microscopy is running a special issue on AI in imaging, guest edited by Peter Soar, Tuan Nguyen and Gianluca Tozzi at the University of Greenwich. Reviews, Methods and Protocols, and primary research articles are all welcome.
Curious about what is inside?
We take on PhD researchers, postdocs and industrial partners working at the intersection of imaging, mechanics and machine learning. Bring a specimen, a dataset or a question.