Location: ICR / Royal Marsden

Supervisory Team: Mr. Myles Smith, Dr. Stamatia Giannarou and Professor Andrew Hayes 

Clinical Specialities: General surgery, surgical oncology, breast surgery, plastic surgery, dermatology

Cutaneous angiosarcoma (cAS) is a rare, aggressive malignancy with poor prognosis, challenging diagnosis, and high recurrence rates. In addition, rare skin malignancies—including in-transit melanoma metastases, radiotherapy-associated sarcomas, and dermal involvement in advanced breast cancer—exhibit clinical heterogeneity and are often difficult to diagnose and monitor.

This PhD project will develop novel artificial intelligence (AI) and computer vision tools to improve detection, delineation, and monitoring of such malignancies. Using a rich dataset from the Royal Marsden Hospital, comprising annotated clinical photographs of angiosarcoma and associated metadata from these and other rare skin cancer patients, the project will construct high-fidelity segmentation models based on diffusion and transformer architectures. These models will be designed to function across primary, recurrent, and therapy-altered disease states.

Optimised imaging protocols and viewpoint-aware diagnostic frameworks will be explored to standardise image acquisition. Synthetic data generation will be employed to overcome the inherent rarity and diversity of disease appearances. Furthermore, 2D imaging will be integrated into 3D reconstruction workflows to visualise disease extent and therapeutic response, enabling augmented reality-based visual tools for surgical planning and patient education.

The project will leverage established industry collaborations with Holocare for augmented and mixed reality integration, and incorporate existing proprietary algorithms developed at RMH/ICR. It will also evaluate novel biosensor platforms in collaboration with engineering partners (e.g. Ladame Laboratory in Imperial College) to support multimodal diagnostics and assessment.

Grounded in patient and public involvement (PPIE), this work will contribute open-source tools and datasets. It aims to deliver clinically integrated diagnostic solutions, improve personalised treatment planning, and lay the foundation for AI-driven stratification in clinical trials targeting rare and complex cutaneous malignancies

Location: Imperial

Supervisory Team: Dr Reza Skandari

Clinical Specialties: Oncology, public health or epidemiology, data science with clinical training.

Early detection and post-treatment monitoring are cornerstones of effective cancer control, yet both screening and recurrence surveillance strategies are often suboptimally designed. This project proposes the development of a data-driven, AI-enhanced framework for optimizing cancer screening and post-treatment surveillance protocols.

The project will use large-scale, routinely collected cancer data—such as those available through the National Cancer Registration and Analysis Service (NCRAS)—to model the natural history of cancer and the timing of disease progression or recurrence. Access to NCRAS or similar datasets must be applied for by the fellow or institutional collaborator via standard procedures.

Multistate models will be constructed to simulate transition dynamics across preclinical, clinical, and post-treatment phases. These models will be integrated with statistical learning and machine learning (ML) techniques to predict individual and population-level risk patterns. Mathematical optimization methods will then be applied to identify screening and surveillance strategies that maximize clinical benefit (e.g., early detection, survival gains) while minimizing unnecessary interventions and resource use.

This work directly supports CRUK’s strategic priorities:

• Advancing early detection through risk-stratified screening and monitoring strategies,
• Promoting convergence science by integrating cancer epidemiology, AI, and health policy,
• Driving innovation in the use of real-world data to inform personalized and sustainable cancer care

Supervisory Team:

The lead supervisor brings expertise in cancer natural history modeling, statistics, machine learning, and optimization. This will be complemented by a co-supervisor with clinical oncology expertise, who will ensure the clinical relevance and feasibility of the proposed strategies. The fellow will benefit from comprehensive interdisciplinary training, equipping them for a leadership role in translational cancer research.

Location: ICR

Supervisory Team: Professor Manuel Salto-Tellez and Dr Tom Lund

Clinical Specialities: Pathology, oncology

Breast cancer therapy is not tailored to the profile of individual tumours. As a result, many patients receive ineffective or excessive treatment. This is particularly true in triple-negative breast cancer (TNBC) where response to neoadjuvant chemo-immunotherapy is highly variable and cannot be predicted.

In this project, we will define histopathological spatial determinants of therapeutic response in TNBC using samples acquired from the BELIEVE (NCT06681064) translational programme. In this study, serial tumour and blood samples are acquired during neoadjuvant therapy to understand how tumours evolve and change during treatment and how these evolutionary trajectories associate with response. By facilitating real-time treatment adaptation and identifying novel therapeutic targets, the BELIEVE study aims to transition TNBC management from a ‘one-size-fits many’ approach to one that is truly personalised, improving patient outcomes.

We will apply established computer vision techniques to extract human interpretable features from H&E and multiplex immunofluorescence (mIF) slides to characterise the tumour microenvironment in TNBC. Spatial metrics, such as immune cell identity, activation state, proximity to tumour cells, and tissue neighbourhood structure, will be derived from mIF data to investigate histopathological spatial determinants of response to neoadjuvant chemo-immunotherapy in TNBC.

To enhance biological resolution, we will integrate spatial imaging features with single-cell RNA sequencing data using advanced machine learning approaches. This integrative analysis aims to uncover robust, multimodal predictors of response, with the potential to inform stratified therapeutic strategies and clinical trial design.

This project offers training in spatial biology, computational pathology, machine learning, multi-modal data integration and translational oncology

Location: ICR

Supervisory Team: Professor Victoria Sanz-Moreno, Professor Andrew Tutt and Dr Stephen-John Sammut

Clinical Specialties: Surgery, Oncology

The spreading of cancer cells from one part of the body to another, called metastasis, is one of the main causes of cancer death. To metastasise, tumour cells must move through tissues, cross tissue boundaries, survive in circulation and later at secondary sites. In our lab we are studying how the cytoskeleton can control all these processes that are crucial for metastatic cells to succeed at growing in a distant tissue, with a focus on these processes in the context of breast cancer.

On the other hand, circulating tumour cells (CTCs) are cancer cells that detach from a primary tumour and enter the bloodstream. They are a critical component of the metastatic cascade, playing a role in the spread of cancer to other parts of the body. While most CTCs die in circulation, a small fraction can survive and seed distant metastatic disease.

In this project, we have access to CTCs from patients enrolled in different studies and clinical trials. This will allow us to profile CTC cytoskeletal dynamics and cell survival in different conditions.

In this project, the student will:

1. Isolate CTCs from patients with advanced metastatic disease. Different breast cancer subtypes will be studied and compared.
2. Using state of the art imaging, the cytoskeleton of CTCs of different breast cancer subtypes will be characterised at high resolution. We will pay special attention to cell blebbing and actomyosin dynamics, as key features of metastatic amoeboid cancer cells.
3. CTCs will be cultured in organ-on-a-chip platforms to study their survival capacity.

Since CTCs are responsible for metastatic seeding, we will investigate if manipulating the cytoskeleton of CTCs has an impact in their survival. The overall goal is to kill these dangerous cells as a strategy to prevent metastasis.

Primary Supervisor: Prof. Maria Kyrgiou 

Primary Division: Metabolism, Digestion and Reproduction, Faculty of Medicine, Imperial   

Endometrial cancer is the commonest gynaecological cancer. Its incidence has risen by 10% over the past decade due to the obesity epidemic. Traditional histopathological classification of endometrial cancer has been insufficient for explaining the molecular heterogeneity of the disease and variation in clinical outcomes. The TCGA endometrial collaborative project described four distinct prognostic endometrial cancer subtypes based on genomic abnormalities (‘POLEmut’, mismatch repair-deficient (MMRd), p53abn, no specific molecular profile (NSMP) endometrial cancer). Molecular profiling using the Proactive Molecular Risk Classifier for Endometrial Cancer (ProMisE) algorithm has shown that molecular subtypes are strongly associated with prognostic significance in endometrial cancer and may better predict outcome than traditional histopathological classification alone. Furthermore, two clinical trials recently reported significantly improved progression-free survival from the universal use of immune checkpoint inhibitors as an adjunct to chemotherapy (in advanced/recurrent endometrial cancer) independent of molecular subtype and are likely to revolutionise management for endometrial cancer. However, currently transference of these two major advances to fertility-sparing management in young women remains inadequately explored, as does the potential for non-surgical therapies in women with early-stage endometrial cancer. Patient-derived organoids have been shown to represent an enhanced model of cancer biology compared to 2D cell culture and animal studies. Our team have been working with the ICR towards the development of benign and malignant endometrial (and cervical) organoids. We have successfully initiated work to demonstrate whether benign glandular and endometrial cancer organoids recapitulate the phenotypic features, epigenetic and metabonomic signatures of primary tissue. Organoids offer aunique opportunity to not only study key carcinogenic processes but also to further explore the potential of alternative and novel therapies During the fellowship we plan to investigate the effects of different therapies, such as checkpoint inhibitors,stratified by molecular profile and receptor status, for precision treatment of early endometrial cancer in fertility-sparing management and beyond.