Novel multi-parametric, multi-modality imaging for the assessment of tumour biology in renal cell carcinoma

Renal cancer (RCC) is the 7th commonest cancer in the UK and clinically, morphologically, genetically and metabolically heterogeneous. Inaccurate patient stratification limits the treatment of RCC. This thesis investigates imaging techniques and image-analysis methods ranging from easily transferrable to highly specialised and dependent on dedicated infrastructure. Artificial intelligence (AI) for diagnosis, prediction and prognostication of RCC experiences increasing scientific interest. However, methodological challenges prevent the validation and qualification of algorithms as imaging biomarkers. Texture features, quantitative descriptors of image composition, are sensitive to minor variations in the tumour segmentation, which future modelling approaches should consider. Twenty-one and 47% of features were poorly reproducible after manual and algorithmic re-segmentation. Measuring response to systemic treatment becomes increasingly important with more therapeutic options available. The poor sensitivity of size-based criteria in targeted and immuno-oncology treatment is well-documented. Three clinical trials investigated morphological and physiological 1H-MRI to detect early response and predict outcome. Multiparametric MRI demonstrated physiological changes after 12 days of anti-vascular therapy, which were compatible with its mechanism of action. The reduction in tumour diffusivity correlated with the long-term volumetric response and progression-free survival. Two trials demonstrated the ability of neoadjuvant antiangiogenic treatment to reduce the extent of venous tumour thrombi in 10/35 patients and enable a less morbid surgical approach in 41%. The most common genetic alteration in RCC impacts metabolism including glycolysis. Hyperpolarised [1-13C]pyruvate MRI (hpMRI) probed the spatially heterogeneous lactic acid fermentation in nine patients with renal tumours. Paired multi-regional tissue samples and imaging afforded a biological understanding of the hpMRI signal. The apparent reaction rate constant (kPL) correlated with the tumour grade, expression of MCT1, the membrane transporter taking up pyruvate, and total LDH, catalysing pyruvate-to-lactate conversion. Improved image analysis tools, physiological and novel metabolic MRI can support precision oncology through better patient stratification and early treatment response detection. The WIRE window-of-opportunity trial of novel treatments in RCC has adopted computational and imaging techniques investigated here as primary, secondary and exploratory endpoints.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2021

Author dc:creator

• Ursprung, Stephan

Advisors dc:contributor.advisor

• Gallagher, Ferdia A
• Stewart, Grant D

• Renal Cell Carcinoma
• Magnetic Resonance Imaging
• Computed Tomography
• Hyperpolarised MRI
• Radiomics
• Treatment Response
• Artificial Intelligence
• Dynamic Contrast-Enhanced MRI
• Diffusion-Weighted MRI
• 3D Printing

Licence

• https://www.rioxx.net/licenses/all-rights-reserved/

Language dc:language

eng

Author Identifier

0000-0003-2476-178X

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/332679