In the battle against breast cancer, identifying the disease's molecular sub-type as part of a patient's diagnosis is key to proper treatment and long-term prognosis. However, effective sub-typing requires not only an examination of the cancerous tissue but the surrounding microenvironment, as well. `
“Though it has been known to be important for over a hundred years, imaging the microenviroment is quite complicated today and it's not used for clinical diagnoses in most cancers," said Illinois Bioengineering Professor Rohit Bhargava. "Further, using the full tumor and microenvironment information in manual diagnoses is too complicated for any practical use.”
Bhargava and members of his Illinois research group recently unveiled a new imaging method for simultaneously sub-typing cancer cells and the tumor microenvironment. Their approach used modern artificial intelligence algorithms to relate the information to disease to provide a powerful new all-digital, automated capability.
The group also developed a novel infrared (IR) microscope and combined it with machine-learning models. This enabled them to quickly and accurately sub-type breast cancer samples with greater accuracy compared to conventional analysis by a pathologist.
Lead author Shachi Mittal, a Bioengineering doctoral student, developed two algorithms that diagnose cell-level tumors and discover tumor-associated microenvironments. One model, known as 6E, digitally examines breast cancer samples and classifies them according to epithelial sub-types—normal, hyperplasia, atypical hyperplasia, and invasive—which are the same categories a pathologist uses.
"The model is basically detecting the hyperplasia [samples] and separating them from the normal or cancer samples," said Mittal. "The second algorithm provides unique insight by identifying the reactions occurring around the tumor in the surrounding microenvironment."
This microenvironment, which consists of many types of cells, including normal tissue, blood vessels, stromal cells, and immune cells, plays a critical role in the progression and metastasis of cancer.
In her experiments to develop and test the models, Mittal initially imaged breast cancer tissue samples using a commercial Fourier-transform infrared microscope (FTIR). Although the FTIR technique yielded rich information content, it took an extremely long time to complete the image processing, which made the method impractical for research or clinical use.
In subsequent work, Mittal tested her models on a novel confocal IR microscope built by her lab colleague and fellow Bioengineering doctoral candidate and co-lead author, Kevin Yeh.
"The image quality, noise performance, and resolution of our microscope exceeds all other instruments of this class,” said Yeh, noting that the instrument also cuts imaging time drastically from days to hours when compared to standard technologies. “These advantages give us the means to provide machine learning algorithms with the cleanest raw data available to enable accurate cancer classification results."
As a result of this work, the team has advanced IR imaging as a viable clinical tool for clinical breast cancer diagnoses.
According to Mittal, IR technology could go beyond just diagnosing breast cancer—it may someday help physicians determine the best course of treatment for cancer patients, determine if a tumor has been completely removed during surgery, and predict the chances of the cancer recurring.
“As opposed to shape-based pathology, this is a practical approach to molecular pathology, said Bhargava, a Founder Professor at Illinois and director of the Cancer Center on campus. "IR imaging offers an opportunity for cancer analysis to be truly all digital, in recognition of disease and in quantification of disease severity."
Added Bhargava: "While useful for visualizing routine samples, this is also an opportunity to leverage the recognition capabilities of computer algorithms for including the microenvironment, difficult cases or those of unclear pathology in limited tissue that is often available in modern biopsies.”
Their work, which was funded by the National Institutes of Health, was published June 5, 2018, in the Proceedings of the National Academy of Sciences: "Simultaneous cancer and tumor microenvironment subtyping using confocal infrared microscopy for all-digital molecular histopathology." Other members of the research team include Dr. L. Suzanne Leslie, Mechanical Science and Engineering graduate student Seth Kenkel, and University of Illinois at Chicago Physician and Professor Andre Kadjacsy-Balla.