Dr. Calum MacAulay is a distinguished scientist at the BC Cancer Research Institute. With a background in physics and engineering, his work on lung cancer is predominantly focused on new and innovative ways of early cancer detection methods.
A prime example of this is Dr. MacAulay's work on tissue sections, which includes identifying various biomarkers and cell composition. Cell composition can highlight how a type of immune cell, called 'killer T-cells', recognize foreign cells like cancer within a person's body and work to destroy the foreign cells.
"My team and I have been evaluating how to quantify the interaction of specific immune cells with specific tumour cells at a cell by cell level. In particular we have been reviewing the frequency that a killer or helper T-cell is located next to a tumour cell across a whole section," said Dr. MacAulay.
"In order to do that, we have to find every individual cell, decide what type it is and then look at all of its neighbours and quantify how many times this specific cell has that specific neighbour. We have discovered that placement of neighbouring cells (killer T-cell next to tumour cell for example) offers a better indication of what's likely to happen to the patient than the frequency of helper or killer T-cells. The placement of the cells is a better predictor of how aggressive a tumour is or whether the tumour is going to respond to immunotherapy treatment or have recurrence than just the frequency of the different types of cell within a sample."
According to MacAulay, the way many T-cells work is that the cells must have surface contact with tumour cells and recognize it as a foreign invader in order for it to work against a cancer. If a patient has a lot of killer T-cells but they are not adjacent to a tumour cell that may not help a patient's prognosis. Dr. MacAulay and team are working on tools to help analyze staging and methods that allow for more efficient and holistic analysis of many cell types in tissue conducting these cell by cell evaluations on thousands of people.
To accomplish this, it is necessary to identify nuclei and their boundaries (segmenting nuclei) in tissue images. This has historically been extremely challenging but is very needed to enable individual cell centric molecular analysis of cancer tissues. As artificial intelligence (AI) becomes more refined, Deep Learning is being applied to performing cell segmentation with results close to those achievable by humans' hand tracing the boundaries of every nucleus (something that is too laborious to consider given that a tissue section could have 100,000s of cells).
"If you'd asked me five years ago if we could do that, I'd have said no. But now we can. We've got some really recent data that suggests we can train a machine to segment out the nuclei of cells even from a cluster of overlapping cells as accurately as a human. I'm excited about that because now we're able to build multiple molecular dimensions for every cell with a tissue."
What comes next will be how to manage the significant amount of data that comes possible from this advancement.
"How to analyze this massive data meaningfully – that's going to be an interesting challenge. A challenge we at BC Cancer and other around the world are working to meet."