Multi-modal molecular imaging of radiation therapy response in the tumour vascular microenvironment

Radiation therapy is recognised globally as a mainstay of treatment in most solid tumours and is essential in both curative and palliative settings. The use of ionising radiation is frequently combined with surgery, either preoperatively or postoperatively, as well as with systemic chemotherapy. Recent advances in imaging have enabled precise targeting of solid lesions, yet substantial intratumoural heterogeneity means that treatment planning and monitoring remains a clinical challenge as changes in tumour size can take weeks to manifest and may be misleading. Photoacoustic imaging is an emerging modality for molecular imaging of cancer, enabling noninvasive assessment of endogenous tissue chromophores with optical contrast at unprecedented spatio-temporal resolution. Preclinical studies have shown that PAI could be used to assess response to radiation therapy and chemo-radiation therapy based on changes in the tumour vascular architecture and blood oxygen saturation, which are closely linked to tumour hypoxia. Given the strong relationship between hypoxia and radioresistance, PAI assessment of the tumour vascular microenvironment has the potential to detect radioresistance and response at much earlier time-points than currently achieved by shape measurements with conventional imaging modalities. In return, this may provide the opportunity to tailor treatments based on noninvasively assessed tumour oxygen availability and heterogeneity. Here, after comprehensively reviewing the multi-faceted roles of molecular imaging in radiation oncology, we develop methodologies for assessing the potential of preclinical photoacoustics to interrogate the pathophysiological response of the tumour microenvironment to radiation therapy validated ex vivo: i) high-resolution longitudinal vascular imaging is performed in physiological and pathological tissues for the comparison of co-registration methods applied to vascular networks imaged with photoacoustic mesoscopy; ii) a novel spatial biology modality, hyperplex sequential immunofluorescence, is applied and validated for breast cancer models vascular phenotyping, focusing on the development of a deep learning-enhanced semi-quantitative image analysis framework; iii) a preclinical radiation therapy trial is conducted in breast cancer models leveraging photoacoustic imaging biomarkers to measure early response, validated ex vivo with immunohistochemistry and hyperplex immunofluorescence. First, the comprehensive co-registration performance comparison revealed the feasibility of aligning vascular networks imaged with photoacoustic mesoscopy. Intensity-based approaches provided accurate alignments when the vasculature was not altered significantly between scans, while a deep learning model provided improved performance when the imaged vasculature was altered between time-points. Second, the hyperplexing of vascular protein expression markers in sequential immunofluorescence could map differences in breast cancer subtypes, leveraging deep learning segmentation models for capturing the vasculature and single cell nuclei. The quantification was validated and enhanced with aligned gold standard serial haematoxilin and eosin sections, for spatially investigating machine learning-derived subregions, enabling multi-modal ex vivo spatial phenotyping of cancer vascular composition. Integrative analyses in viable tumour subregions highlighted the capabilities of novel hyperplex immunofluorescence in capturing phenotypic characteristics of the breast cancer-recruited vasculature, underpinning features of radiosensitivity. Third, in the same preclinical models, features of radioresistance in the most aggressive triple negative breast cancer model could be identified at baseline on photoacoustic imaging, providing predictive biomarkers of response. Early response could be identified with intratumoural blood oxygenation and vascular morphology as early as 24h following radiation delivery in the most radiosensitive model, which was validated at endpoint using immunohistochemistry and the hyperplex immunofluorescence framework, revealing mechanisms of vascular response to radiation therapy. Taken together, a comprehensive multi-modal molecular imaging framework was developed to assess hypoxia and radiation therapy response in the tumour vascular microenvironment. Future plans are developed to further investigate the use of photoacoustics preclinically and clinically in radiation oncology, focussing on the evaluation of early response to radiation therapy and radiation-induced side effects.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2025

Author dc:creator

• Lefebvre, Thierry Luc

Advisor dc:contributor.advisor

• Bohndiek, Sarah Elizabeth

• Biomedical imaging
• Breast cancer
• Deep learning
• Digital pathology
• Medical image analysis
• Medical physics
• Oncology
• Photoacoustic imaging
• Radiation therapy
• Treatment response
• Vascular biology

Licence

• https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/9cda194b-3290-4f8a-b58a-ebd5f4761ee5/download
• http://purl.org/NET/rdflicense/allrightsreserved

DOI dc:identifier.doi

https://doi.org/10.17863/CAM.119110

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/385533