Skip to main content

Together we are beating cancer

Donate now
  • For Researchers

Pictures that speak a thousand words

by Phil Prime | Interview

16 October 2024

0 comments 0 comments

Joanna Edwards

A new project will see researchers, surgeons and engineers use AI to generate incredible insights from biological scans and histological stains. We caught up with co-lead Professor Joanne Edwards to talk radiomics, stratification and potential patient impact…

Talk us through the project, and what you’ll be brining to it…

We are aiming to develop a novel method of classifying which rectal cancer patients will respond to chemoradiotherapy and which patients will gain no benefit.

The generation of a new classification system to predict chemoradiotherapy response and resistance will need a comprehensive and integrative approach – and we’ll be combining clinical, radiological, pathological and molecular features.

Advances in predictive modelling could further enhance our ability to personalise treatment plans, leading to better outcomes and less toxic side effects for patients.

Advancements in radiomics, deep learning and AI pathological approaches are now making it possible to do this. These advances in predictive modelling could further enhance our ability to personalise treatment plans, leading to better outcomes and less toxic side effects for patients. The current team includes engineering and physical science scientists, computing scientists, colorectal surgeons, bioinformaticians and translational pathology biomarker scientists to develop this novel and potentially transformative approach.

Professor Campbell Roxburgh, a colorectal surgeon, has already established a unique resource of specimens from rectal cancer patients that allow longitudinal assessment of response during, and following, neoadjuvant chemoradiotherapy. And we have already collaborated on gene expression and multiplex immunofluorescence analysis of these unique rectal cancer samples.

The project is split up into a number of packets and I will be involved in leading work stream 2. I am a translational cancer pathology scientist, and I’ll be working closely with Dr Yinyin Yuan, an expert in developing morphological classifiers. We’ll be using deep learning to identify morphological features from histological images of rectal cancer biopsies that associate with patient response or resistance to chemoradiation.

We’ll also perform immunohistochemical analysis to validate findings from the gene expression studies. This histology and protein expression data will be integrated with the genomic, transcriptomic and immune cell data we already have, and layered with the novel radiomic and AI generated morphological features. The aim is this will give us novel classifiers to define patient response or resistance to chemoradiotherapy.

Clearly radiomics is important here – but it’s a relatively new field, tell us what it is and what you think could be achieved with it

The goal of radiomics is to convert scan images into high-dimensional data that can be analysed to support clinical decision-making.

In our project we’ll extract features from scans to enable us to quantify patterns and textures and shape and sizes of different features within the image. We hope by doing this we will identify features not normally recognised by the human eye and build on the information that these images can provide.

The advancements currently being made with AI in radiomics and deep learning in histology makes this an exciting project that is highly likely to inform the management of rectal cancer.

By doing this we aim to characterise the tumour to not only predict response or resistance to chemoradiotherapy but to also monitor response of the tumour to therapy using our specimens. We are exploiting radiomics to identify patients that will respond to treatment, and monitor disease progression and therefore improve patient outcomes.

Within the current project we will be integrating radiomics with other data such as clinical, pathological and genomic, to enhance classification methods. The advancements currently being made with AI in radiomics and deep learning in histology makes this an exciting project that is highly likely to inform the management of rectal cancer.

Histopathology image

Your project is based on patient stratification. Tell us why this is important in rectal cancer patients

Patient stratification for rectal cancer is becoming increasingly necessary. Not only do we need to select the patients that will benefit from traditional chemoradiotherapy but also with the development of new drugs – such as PD-1 and PDL-1 inhibitors that harness the immune system – this work it is becoming increasingly crucial.

Only by tailoring treatment to individual patient’s clinical and tumour characteristics will we be able to improve outcomes and reduce unnecessary toxicities and interventions. In addition to selecting the best therapies for a particular patient, it might also provide information to inform dosing. By identifying patients that could benefit from a more aggressive or less aggressive approach we could maximise benefit to the patient and minimise side effects.

Identification of patients that will respond to neoadjuvant therapy will also help guide the surgeons on the type of operations that will be required, allowing the surgeon to decide if organ preservation is possible.

Do you foresee a time where your approach could be used as a strategy for population-level cancer screening for early diagnosis?

It is possible that some of the work done within this project would be able to identify features with a benign polyp that might identify a patient at high risk of developing rectal cancer. Also, it might identify patients with early rectal cancer who are less likely to progress and could be treated with surgery alone.

In an ideal world – what would you like to see develop around patient stratification and CRC over the next decade?

In the current project we will develop novel multimodal classifiers with self-supervised learning in computer vision. And we are doing this to answer a particular question: can we identify which patients will respond to chemoradiotherapy? However, as the field expands, I hope that we will be able to apply this approach to all colorectal cancers for a range of therapeutic options.

In the longer-term, patient stratification will be essential for all colorectal cancers if we are to optimise treatment effectiveness and minimise unnecessary side effects. This will improve patient outcomes and allow the most efficient use of healthcare resources.

By considering individual patient characteristics and tumour biology, clinicians can develop personalised treatment plans that offer the best possible prognosis and quality of life for their patients. With research into risk stratification there is no reason why this can’t be a reality in the next decade.


 

Joanne Edwards

Joanne Edwards is Professor of Translational Cancer Pathology, and CRC theme leader, at the CRUK Scotland Centre.

Her co-leads on the project are Campbell Roxburgh, Professor of Colorectal Surgical Oncology at the University Department of Surgery, Glasgow Royal Infirmary and Emiliano Spezi, Professor in Medical Engineering at Cardiff University.

Tell us what you think

Leave a Reply

Your email address will not be published. Required fields are marked *

Read our comment policy.

Tell us what you think

Leave a Reply

Your email address will not be published. Required fields are marked *

Read our comment policy.