A York University researcher, who led a team from U of T and Sunnybrook, undertook a study to predict whether a metastatic brain tumour would respond to radiotherapy or not. Early alterations in treatment, based on the prediction, could improve patient outcomes.
For 20 to 40 per cent of patients with cancer, the disease metastasizes to the brain, which can be deadly. A York University researcher, Professor Ali Sadeghi-Naini in the Lassonde School of Engineering, wanted to better the odds. He led a research team whose members were from the University of Toronto, Sunnybrook Research Institute, Sunnybrook Health Sciences Centre and the Medical University of Lublin (Poland).
Sadeghi-Naini’s team developed a novel methodology to predict, faster, whether or not a metastatic brain tumour would respond to radiotherapy. This new knowledge could mean that doctors would be able to intervene with adjustments in treatment and ultimately improve patient outcomes. The impact of this research could be profound.
This work was supported by the Natural Sciences and Engineering Research Council of Canada and the Canadian Institutes of Health Research. The findings were published in Scientific Reports (2019), a high-impact Nature publication.
Sadeghi-Naini, an expert in quantitative imaging and artificial intelligence (AI) for precision medicine, sits down with Brainstorm to discuss this important study.
Q: What were the objectives of your study?
A: Up to 40 per cent of all cancer patients develop brain metastasis. Radiation therapy is an established option for brain metastasis treatment. Unfortunately, up to 20 per cent of brain metastases don’t respond to radiation therapy, which means that the tumour still progresses after treatment.
The objective of this study was to develop non-invasive, quantitative MRI biomarkers that could predict, early, the outcome of local failure in brain metastasis – that is, failure to control tumour progression – after radiotherapy. Our objective was to see whether or not a metastatic brain tumour responds to radiotherapy.
Q: Tell us about the genesis of this project and how you went about undertaking this work.
A: I’m a cross-appointed scientist at Sunnybrook Health Sciences Centre. About three years ago, with my colleagues at Sunnybrook, I was discussing the clinical challenges with brain metastases treatment and how quantitative imaging and artificial intelligence (AI) techniques could potentially help therapy outcome prediction that could facilitate tailoring treatment for individual patients with brain metastasis. We were interested in personalizing treatment for patients to achieve better outcomes.
Interestingly, they mentioned that they already had clinical characteristics of more than 100 patients with brain metastasis in a clinical data set at the Odette Cancer Centre at Sunnybrook. So, after our discussion, we decided to retrieve imaging data for these patients and explore possibilities of therapy outcome prediction using these images.
It took over a year to obtain and organize those images. We developed an integrated framework for lesion delineation in these images; we could extract quantitative features describing the shape of the tumour and edema, heterogeneity within different regions of the lesion, etc. In other words, we derived various parameters that describe these tumours quantitatively.
We were looking for imaging biomarkers, the key features that could be used to predict response. We adapted methods of data analytics with AI models to develop biomarkers that can predict response of brain metastasis to radiation therapy.
Q: What were your key findings, and did anything surprise you?
A:The key finding was that non-invasive quantitative MRI biomarkers integrated with machine learning techniques can predict local control versus failure (response versus no response) in brain metastasis after radiation therapy.
The surprising observation was that the majority of the features in the developed quantitative MRI biomarkers were not derived from the tumour itself but from the surrounding regions of the tumour.
This essentially means that the region we should focus on for prediction is mainly the peritumoural area [area around a tumour]. We hypothesized that there are cancerous cells in those surrounding regions, but the number of those cells are not high enough to make an evident contrast on an MRI image. Our methodology quantifies heterogeneity in those regions, characterizing the frequency and distribution of cancerous cells, that could be linked to the outcome of the treatment.
Q: Could you elaborate on the applicability of this research and what this could mean for cancer patients?
A: The patients in this study were followed up to five years after their treatment. We analyzed survival of these patients after their radiation therapy, and found that the patients who were predicted by our model as responders demonstrated significantly better survival rates compared to those predicted as non-responders.
This means our prediction can potentially have an impact on the patient’s survival: treatment adjustments, facilitated by such an early prediction, can potentially improve survival rates and quality of life for cancer patients.
Q: How has York supported your work and, since you came here in 2018, what are your impressions of the University?
A: York provided space for my quantitative imaging and biomarker laboratory (QUANTIMB Lab) at Lassonde. York bolstered my program of research with very good administrative support.
Things that impressed me about York? The multicultural and diverse atmosphere, and the state-of-the-art facilities at Lassonde.
To learn more about Research & Innovation at York, follow us at @YUResearch; watch our new video, which profiles current research strengths and areas of opportunity, such as Artificial Intelligence and Indigenous futurities; and see the snapshot infographic, a glimpse of the year’s successes.
By Megan Mueller, senior manager, Research Communications, Office of the Vice-President Research & Innovation, York University, firstname.lastname@example.org