Leveraging AI and image analysis technology to improve prognostication in colorectal cancer

Korsuk Sirinukunwattana completed a PhD degree in Computer Science from the University of Warwick, focusing on computational pathology. He then became a postdoctoral research fellow at Beth Israel Deaconess Hospital, Harvard Medical School and is currently a postdoctoral research assistant at the University of Oxford, based at the Big Data Institute.

Korsuk’s research concerns leveraging artificial intelligence and image analysis technologies for the development of novel biomarkers extracted from histological slides with molecular and biological interpretability has remarkable potential for clinical translation. Using deep learning, he has predicted consensus molecular subtypes (CMS) of colorectal cancer (CRC) from standard histology sections.

The current standard for prognostication of CRC patients is based on the assessment of histologic materials and tumour progression as defined by the anatomical criteria (TNM staging system). This information supports the definition of broad prognostic risk groups but has no predictive value. The integration of genomic technologies in the clinical care of CRC patients has immense potential to drive personalised treatment but requires substantial financial, personnel and infrastructural resources. On the other hand, histology slides are generated as part of the standard work-up of any CRC treated by surgical resection. Combining morphological information derived from histology slides with molecular profiles to identify genotype-phenotype correlations is a promising and cost-effective approach to extend the amount of clinically relevant information that can be extracted from standard histologic slides.

In Oxford Korsuk collaborates with Jens Rittscher (IBME), and Enric Domingo and Tim Maughan (Oncology) as part of the S:CORT consortium. Internationally, he works with Viktor Koezler at the University of Zurich. He is funded by the NIHR Oxford Biomedical Research Centre.

Find out more about our research below

The search for pancreatic cancer biomarkers

Nuffield Department of Population Health researchers are using the China Kadoorie Biobank to identify potential protein biomarkers in the blood that could be used to predict the presence or development of pancreatic cancer

Early stage ‘red flag’ symptoms for pancreatic cancer

The ADEPTS study uses the QResearch database to better understand what ‘red flag’ symptoms may be associated with pancreatic cancer, in the hopes of promoting earlier diagnosis from primary care.

Detecting pancreatic cancer through blood tests

The Song Lab recently developed an effective and accurate way of detecting cancer biomarkers in the blood. Now, they are looking at the application of TAPS technology in pancreatic cancer

Understanding breast cancer risk in Chinese populations

Researchers from the Clinical Trial Service Unit and Epidemiological Studies Unit (CTSU) are utilising the China Kadoorie Biobank to better understand how breast cancer risk factors may differ between individuals from Western and Eastern populations
Women

Using big data in breast cancer research

The Cancer Epidemiology Unit has been using the largest epidemiological data set of its kind to unlock the secrets of breast cancer, what can be done to prevent it, and which women are most likely to develop it

A new FRONTIER for breast cancer

Latest news from FRONTIER, the trial investigating the potential of the radiotracer Fluciclovine in the subtyping and staging of breast cancers

New AI technology to help research into cancer metastasis

DeepScratch is a new AI technology that can be used to analyse how cells move in response to wound, building on the latest advances in deep learning

IL-22 pathway linked to poor prognosis in colon cancer

Research shows how IL-22 interacts with KRAS mutant tumours to promote excessive growth in colorectal cancer

Using machine-learning approaches to identify blood cancer types

Oxford researchers have outlined the applications of AI in the classification of myeloproliferative neoplasm cancers in a new study