PREDICT

Page 1

Techniques that point the way towards personalised treatment Radiomics technologies can provide deeper insights into a patient’s condition, which then opens up new possibilities in terms of personalised treatment. We spoke to Dr Anke Wind, Professor Michael Brady and Dr Mathieu Hatt about the work of the Predict project in training the next generation of radiomics researchers. A number of different kinds of image are taken for a patient suspected of having cancer, each of which provides complementary pieces of information, which together gives clinicians an insight into their condition. The full nature of an individual’s condition may not be apparent to the naked eye however. “It can be difficult for radiologists to grasp the whole picture from an image. It’s difficult to delineate and segment a tumour for example,” explains Dr Anke Wind, a researcher in the Precision Medicine department at Maastricht University. This is where the emerging field of radiomics, which involves using technology to analyse radiographic images and enhance the quality of information they provide, comes in. “Radiomics could help save time. It would be much faster if we could analyse images in an automatic way rather than going through them by hand,” outlines Dr Wind. “We could also do it in a more precise way. A computer may be better at finding different lesions on an image.”

that brings together academic and industrial partners from across Europe, including several SMEs, to provide training to 14 early stage researchers (ESRs). The objective here is to both develop new radiomics methods and also to equip ESRs with a broad range of skills. “On the one hand we want to train the ESRs as radiomics researchers. For that they need to know about things like machine learning, artificial intelligence, and data processing, that’s the technical part,” says Dr Wind.

The ESRs also receive training in areas like project management, patient engagement and cost analysis, with the aim of giving them a rounded perspective, rather than focusing solely on the technical work. “Not only will the ESRs know a lot about radiomics, they’ll also know about project management, patient involvement and how to write grants,” says Dr Wind. “If they want to then pursue a career in industry, they will have a broader range of skills to draw on.”

Predict project This issue is central to the work of the Predict project, an Innovative Training Network (ITN)

www.euresearcher.com

19


The aim in the technical side of the project is to develop underpinning technologies that apply to a range of cancers and which detect true positives as early as possible, while at the same time also minimising the numbers of false positives. With radiomics technologies, large numbers of features are extracted from medical images, including MRI, PET and CT images. “The features that we extract are those that can be identified by image analysis methods – both classical methods of estimating textures, and also more novel methods fuelled for example by machine learning,” says Professor Michael Brady, Chairman of Perspectum Diagnostics in the UK, one of the project partners. These features may give information about the tumour microenvironment; but that information is determined by the spatial and temporal - resolution of the imaging technique. “MRI and CT generally image a tumour at mm scale, whereas various forms of microscopy can image at micron scale, and sometimes more finely than that,” continues Professor Brady. Many tumours are highly heterogenous, and this can affect how a patient will respond to treatment. The aim with radiomics techniques is to characterize numerous features of a given tumour in a single analysis from routine clinical images, in order to more completely characterize its phenotype, which can then inform clinical decisions. “The features usually considered include volume, size and shape. Intensity features are also considered as well as textures, which are aimed at quantifying the heterogeneity of intensities and their spatial relationships. The challenge is to identify and select the most relevant features amongst the dozens - or even hundreds that are typically extracted,” explains Dr Mathieu Hatt, a researcher in the Laboratory of Medical Information Processing at Inserm, another project partner. In a typical radiomics workflow, these features are extracted from each available image, then a selection process is applied to identify which are useful for a specific task. “This task might be predicting a patient’s response to therapy for example,” says Dr Hatt. A model capable of accurately predicting a patient’s response to therapy would attract a great deal of interest, as it could help clinicians tailor treatment management strategies to the specific needs of the individual patient. With radiomics, machine learning techniques are typically applied on data relating to certain cohorts of patients to identify the important features in terms of their response to treatment, as Dr Hatt explains. “Algorithms

20

PREDICT A new era in personalised medicine: Radiomics as decision support tool for diagnostics and theragnostics in oncology

Project Objectives

PREDICT is a MarieCurie innovative training network that aims at training a new generation of early stage researchers to become leaders in the field of Radiomics and personalised medicine, ultimately aimed at improving the diagnosis and treatment of cancer.

The concept of Radiomics.

recurrence of a condition. “Ultimately we could go more towards personalised followups, not every patient needs to be followed that intensely for that period of time. But that’s more of a goal for the future,” says Dr Wind.

Personalised medicine The wider context here is the goal of developing more personalised treatments tailored to the needs of individual patients. Alongside more tailored treatment options, many patients today are also involved in identifying what course of treatment is

Algorithms mine the available data and try to identify which features are the most important in order to combine them into a meaningful model. Such models are sometimes called a ‘radiomic signature’.

Inter/multi-disciplinary and intersectoral interactions of PREDICT.

mine the available data and try to identify which features are the most important in order to combine them into a meaningful model. Such models are sometimes called a ‘radiomic signature’,” he outlines. While the project is primarily focused on detecting tumour heterogeneity, Dr Hatt says this research holds wider relevance. “Any clinical task where imaging is relevant could potentially benefit from radiomics. This includes screening, diagnosis, staging, treatment assessment and follow up,” he says. There is interest in using radiomics methods to monitor a patient’s response to treatment

“With radiotherapy, a patient may have to go into hospital 4-5 days a week, whereas with surgery there’s the operation then a subsequent recovery period.” There are also a number of other projects within Predict, with researchers developing their technical skills, while also gaining experience in industrial settings. “A big part of this project is about giving students the chance to do part of their research at another company or university, so that all of them have the experience of conducting research at both a company and also a

for example. This could involve combining baseline information with data from a few within-treatment datapoints, in order to identify which patients are responding well to treatment, which are responding partially, and which are not responding. “Once that is in hand we can begin to make predictions on ‘what if’ scenarios, such as dose boosting. Such modifications to therapy are familiar from radiation oncology,” outlines Professor Brady. Effective and reliable radiomics techniques could also give clinicians the ability to identify which patients will require a follow-up, and which are at lower risk of experiencing a

EU Research

right for them, another topic which is being addressed in the Predict project. “One of the students in our project is working on a patient decision aid. She is using the available information and looking into the side-effects associated with different treatments. With that information, she’s then trying to make an individualised patient decision aid,” says Dr Wind. This aid would help a patient assess different treatment options, for example surgery or radiotherapy to treat lung cancer, which may have similar outcomes but different side-effects. “What would the patient actually prefer?” continues Dr Wind.

www.euresearcher.com

university environment,” stresses Dr Wind. This has been unavoidably disrupted by the Covid-19 pandemic, yet Dr Wind is hopeful that the students will have the opportunity to experience a different setting. “We give our students the chance to do their PhD at a university, and they will all get their doctorates at a university. However, some of them are doing a lot of their research within a company, which is a different environment,” she says. “All of the students will have the chance to experience working at a company as well as a big university. And I think that’s very valuable.”

Project Funding

This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 766276.

Project Consortium

• Maastricht University • Fondazione IRCCS Istituto Nazionale dei Tumori • University Liege • Deutsches Krebsforchungszentrum • Institut National de la Sante et de la Recherche Medicale • Oncoradiomics • Mirada Medical • Perspectum Diagnostics • Health Innovation Ventures

Contact Details

Principal investigator: Prof. Philippe Lambin Senior Scientist: Dr. Henry Woodruff Faculty of Health Medicine and Life Science Department Precision Medicine - The D-Lab Universiteitssingel 40, 6229 ER Maastricht, The Netherlands Dr Anke Wind Maastricht University Department of Precision Medicine E: a.wind@maastrichtuniversity.nl W: https://www.predict-itn.eu/home W: https://www.predict-itn.eu/news W: https://www.predict-itn.eu/archive Dr Anke Wind

Dr Anke Wind is a project manager at Maastricht University, while she is also involved in teaching and student supervision. She previously worked as a post-doctoral researcher at Rijnstate Hospital, while she has also conducted research into benchmarking comprehensive cancer care.

21


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.