Lung cancer is accountable for more deaths than breast, colon and prostate cancer combined.
As lung cancer screening advances, it is anticipated that more patients might be diagnosed at earlier stages, enabling them to undergo surgery, the first treatment for early-stage patients.
However, in a major proportion of patients after resection (surgery to remove the tumor), the cancer will come back. Unfortunately, current clinical guidelines cannot help predict which patients are at risk. Better knowledge of who’s at risk has essential implications for choosing systemic therapies, corresponding to chemotherapy, for early-stage lung cancer patients after surgery.
To find an answer to this problem, our research group at McGill University launched a project in collaboration with Université Laval. Preliminary results were published in Nature. In our work, we found that the use of latest imaging technology combined with artificial intelligence can improve treatment outcomes for cancer patients.
Too much or too little intervention
This clinical dilemma has essential implications for the alternative of treatment corresponding to chemotherapy. For example, lung cancer patients who were cured by surgery could avoid the toxicity of chemotherapy. Patients at risk of cancer recurrence may gain advantage from additional therapeutic interventions.
The challenge of predicting recurrence in patients with early-stage lung cancer has essential implications for the 31,000 Canadians who are diagnosed with this terrible disease every year.
Mass cytometry imaging
To address this clinical problem, we used imaging mass cytometry (IMC), a brand new technology enabling comprehensive characterization of the tumor microenvironment.
The tumor microenvironment is a posh ecosystem composed of interactions between tumor cells, immune cells, and various structural cells. IMC could be used to visualize up to 50 markers on the cell surface, which is way more than was previously possible.
This technology makes it possible to discover differing types of cells and determine their spatial organization, i.e. the way in which they interact with one another. IMC generates images that could be analyzed to determine the frequency of cell subpopulations, their activation states, the opposite cell types with which they interact, and their organization in cell communities.
The results of our study, published in Naturereveal that different cell types can interact in cell communities and that communities composed of B cells are strongly related to prolonged survival in lung cancer patients. Our study highlights that beyond cellular frequency, cellular interactions and spatial organization also strongly correlate with essential clinical outcomes corresponding to survival.
Using artificial intelligence to create higher forecasts
Based on our preliminary results, we hypothesized that essential spatial features embedded in IMC images, corresponding to cellular interactions, could also be essential in predicting clinical outcomes.
Our dataset of 416 patients and over 1.6 million cells provided enough power to make predictions using AI. We sought to predict which patients with early-stage lung cancer would experience tumor recurrence after surgery.
Using 1 mm2 tumor samples, material available from surgical resections or biopsies, we used artificial intelligence algorithms together with IMC images to make our predictions. Our algorithm was able to predict with 95% accuracy which patient would experience a cancer recurrence using the spatial information contained in the pictures.
Six markers could make a difference
One of the challenges of applying our results to clinical settings is that IMC just isn’t available. Clinical pathologists typically use less complex technologies corresponding to immunofluorescence, which are often limited to three or fewer markers.
(Shutterstock)
To address this challenge, we sought to determine the minimum number of markers needed to obtain meaningful predictions of disease recurrence in lung cancer patients after surgery. Using six markers, we achieved an accuracy of 93%, which is close to the 95% accuracy achieved using 35 markers.
These results suggest that by harnessing the facility of artificial intelligence with existing technologies available in hospitals, we may give you the chance to improve postoperative clinical management of patients with early-stage lung cancer. Our ultimate goal is to increase cure rates in people at high risk of cancer recurrence while minimizing toxicity in individuals who could be cured surgically.
