Specialist Training at EMBL Hamburg - Nimeet Desai

At PETRA III, DESY, during my specialist training visit to EMBL Hamburg.

From 5 to 10 July 2026, I visited the Biological X-ray Imaging team at EMBL Hamburg/DESY for specialist training in synchrotron-based phase-contrast X-ray tomography. My PhD combines the development of a mitomycin C (MMC)-releasing ocular composite with a chick chorioallantoic membrane (CAM) xenograft model of conjunctival melanoma. Although conventional histology provides detailed cellular information, physical sectioning removes much of the three-dimensional context. The DiMeN Flexible Funding award enabled me to investigate intact tumours volumetrically and develop the computational skills needed to analyse these complex datasets.

At the P14 experimental setup during phase-contrast X-ray tomography data collection.

I brought pre-processed, paraffin-embedded CAM tumour xenografts representing different growth stages. These included untreated tumours and tumours treated with either standard MMC solution or MMC delivered using my ocular composite. During allocated beamtime at the P14 beamline of the PETRA III synchrotron, we used the EMBL High-Throughput Tomography workflow to perform high-resolution propagation-based phase-contrast imaging. Unlike conventional absorption imaging, this approach exploits small differences in refractive index, providing strong structural contrast within soft biological tissues without additional contrast staining. Because each high-resolution acquisition covered only part of the sample, adjacent tomographic volumes were reconstructed and stitched to generate continuous three-dimensional representations of the xenografts.

The scans were successful and resolved tumour boundaries, internal structural heterogeneity and treatment-associated changes that were not fully apparent from the external appearance of the samples. The reconstructed volumes also allowed the vascular network to be followed through the tumour rather than assessed from isolated two-dimensional sections. This provides a basis for comparing the growth pattern, vascular organisation and tissue architecture of the CAM xenografts with recognised clinical features of conjunctival melanoma.

Importantly, the datasets showed the spatial relationship between the externally applied composite and the underlying tumour. This indicated that the position and degree of contact between the biomaterial and xenograft may influence the direction and shape of tumour growth, identifying an important variable to control when refining the implantation and treatment procedures.

Representative stitched X-ray reconstruction (left), Dragonfly 3D rendering (centre) and segmented tumour vasculature shown in green (right).

The most valuable component of the visit was the hands-on training in Dragonfly with Dr Ksenia Denisova and Dr Fabio De Marco [Duke Team, Biological X-ray imaging]. I learned how to import, orient and navigate large stitched volumes; optimise intensity windowing; and define focused volumes of interest to make the analysis computationally manageable. For segmentation, we combined intensity- and region-based selection with manual correction across consecutive slices. A major challenge was distinguishing genuine vessel lumens from processing voids, cracks, paraffin boundaries and phase-contrast edge-enhancement artefacts. This required examining each candidate structure in axial, coronal and sagittal planes and checking its continuity and branching morphology in the three-dimensional render.

I also learned that a convincing 3D visualisation is not necessarily an accurate segmentation. Each labelled structure had to be repeatedly checked against the underlying greyscale data, particularly at tumour margins and around small vessels where partial-volume effects reduced boundary clarity. This iterative quality-control process was one of the most important insights from the training. It provided a reproducible route towards extracting quantitative endpoints such as tumour volume, vessel volume fraction, vascular distribution and the spatial relationship between the composite and tumour.

The visit has moved my analysis beyond descriptive imaging towards quantitative 3D assessment. I can now integrate volumetric measurements with H&E and immunohistochemistry to interpret tumour viability, vascularisation and treatment response while retaining spatial context. The workflow will also help me select representative regions for subsequent histology, refine composite placement and adapt similar segmentation approaches to laboratory microCT datasets. By linking 3D tumour architecture and vascular organisation with histological and treatment-response endpoints, the data generated during this visit will provide a strong foundation for a rigorous and impactful publication on the CAM conjunctival melanoma model and the therapeutic performance of the MMC-releasing composite.

I also plan to transfer the practical learning, quality-control principles and analytical insights gained at EMBL to the Egg Facility Shared Research Facility (SRF) at the University of Liverpool. Sharing this workflow with other researchers will strengthen local CAM imaging and analysis capabilities, support future collaborations and promote wider scientific advances in quantitative preclinical research. I am grateful to Dr Ksenia Denisova and Dr Fabio De Marco for their experimental support and detailed Dragonfly training, and to the wider Biological X-ray Imaging team at EMBL Hamburg. I also thank the MRC DiMeN DTP for making this technically valuable visit possible.