Meet the Experts
Content provided by Dr Stephanie Craig and Natalie Fisher.
Within the PGJCCR, several projects implement digital pathology to answer biological problems across a variety of cancer types.
The Dunne Lab focus on identifying early-stage colorectal cancers which are at a high risk of metastasizing, one biomarker of these aggressive early-stage tumours is tumour budding. Tumour buds are small clusters of tumour cells which break away from the main tumour body. It is widely accepted that these are the tumour cells which enter the bloodstream and seed metastasis at distant sites, indicative of a worse prognosis. Manual assessment of tumour budding is time-consuming and difficult to reproduce, therefore, in a study lead by Natalie Fisher, we have developed a digital pathology solution to enable individual tumour buds to be annotated (red). This means quantitative analysis can quickly and accurately score across a tumour, allowing reproducible assessment of this important prognostic factor.
Image credit: Natalie Fisher
H&E-stained tissue sections are generated for routine diagnosis and grading of tumours in pathology but are also an untapped resource for gleaning information on tumour composition and infiltration of immune cells. The Allott lab have digitized H&E-stained diagnostic biopsies to quantify neutrophils and tumour infiltrating lymphocytes in the prostate tumour microenvironment, exploring their relationship with lifestyle factors and with clinical outcomes for these men. We found that physically active men had lower tumour inflammation while those experiencing weight gain had greater tumour inflammation. Prostate tumour inflammation, in turn, was associated with more aggressive tumour features. Our findings highlight how digital pathology approaches can be leveraged to identify possible prevention strategies for aggressive prostate cancer.
Illustration of digital analyses of H&E stained tissue. Image courtesy of Allott Lab
Radiotherapy plays a key role in cancer treatment for more than half of all patients, including people with breast, lung and oesophageal cancers, and lymphoma. Heart disease is common after treatment, however, and has recently been discovered to be related to the radiation dose received by specific substructures within the heart. Identifying these substructures during radiotherapy planning is difficult and time-consuming (~2 hrs/case).
As part of his PhD, and in collaboration with Memorial Sloan Kettering Cancer Centre, Dr Gerard Walls recently tested the application of a deep-learning-based auto-segmentation tool. The algorithm, which was previously trained on a historical group of patients with outdated imaging, was evaluated in a modern cohort of patients from Northern Ireland that underwent contemporary 4D-CT radiotherapy planning scans. Dr Walls and team found the tool to have a high performance in the current standard-of-care imaging, and plan to integrate it in future radiotherapy research in Belfast, with the ultimate aim of reducing heart disease after cancer treatment.
Three-dimensional reconstruction of automated cardiac substructures from the 4D radiotherapy planning CT scan of a representative patient, from the anterior (1), posterior (2), left (3) and right (4) perspectives
(cyan = right atrium; orange = left atrium; blue = right ventricle; red = left ventricle; green pulmonary artery; magenta = aorta; yellow =superior vena cava; brown = inferior vena cava)
Image Credit: Dr Gerard Walls
Work is also being done across multiple labs within the centre to correlate the findings from animal studies into the human setting. This is an example from the Small lab where biomarker X is seen to be more enriched in lung tumours compared to normal lung tissue in the mouse setting. Further analysis has discovered biomarker X is degrading the extracellular matrix, enabling tumour growth. Tissue from normal human lung and human lung tumours was stained for biomarker X, and an enrichment was also observed in the tumour setting. This analysis shows how a biomarker discovered in the animal setting can be explored, and then validated in the human setting, to enable the identification of human lung tumour tissue.
Image credit: Small lab, courtesy of Roisin Morelli
Also, beyond the PGJCCR, the Tissue Hybridisation and Digital pathology section of the Precision Medicine Centre, lead by Professors Salto-Tellez and James focuses on translating novel biomarkers into tools that can be utilised for clinical benefit. One of the many projects currently underway is using digital pathology and artificial intelligence approaches to develop novel algorithms for the clinical assessment of PD-L1 in Lung Cancer. Other studies are focusing on determining HPV status in Head and Neck Cancers, and this has been recently implemented in changes to clinical practice.
