Authors: Eliza Ballantyne, Maria Lizio, Anshuman Chaturvedi
Mentors: Dustin Shipp
Insitution: Utah Valley University
We evaluate techniques for enhancing performance of Raman based classifiers of lung cancer and compare them to results from immunohistochemistry and hematoxylin and eosin (H&E) staining for fixed samples. In the United States, more patients die from lung cancer than from any other type, although it ranks as the third most common cancer. For patients with lung cancer, preserving the healthy bronchioles where cancer usually forms is vital to continued lung function. Raman spectroscopy is already a valuable asset in distinguishing between healthy tissue and many types of cancer and decreases discrepancies between diagnosis, saving medical resources and improving patient outcome. Lung cancer is especially challenging for Raman spectroscopy, in part because tar fluorescence often overpowers critical chemical features. We introduce measurement and classification approaches as the first step to overcome this challenge and create reliable Raman based classifiers for lung cancer. By working with fixed tissue sections, we avoid tar fluorescence and demonstrate the ability to detect tumors and premalignant abnormalities in lung tissue. These samples allow collaboration between adjacent sections in both H&E staining and immunohistochemistry. Furthermore, additional measurements of fixed sections can be acquired at any time. These advantages provide flexibility to acquire more detailed training sets, include more detail of any borderline cases, and compare Raman spectroscopy to more specialized histopathological techniques.