Computational Analysis and Classification of SHG Images of Cancerous Pancreatic Tissue Based on Collagen Fiber Alignment
Author:
Gerren Welch
Name Change:
Major:
Physics
Graduation Year:
2021
Thesis Advisor:
Karissa Tilbury
Description of Publication:
Pancreatic cancer is a deadly disease, with a low five-year survival rate partly due to the difficulty in diagnosing the cancer early in its development, as it shares symptoms with more common and less lethal conditions. Using Second Harmonic Generation (SHG) microscopy and computer analysis, our knowledge of the biophysics of the pancreatic tumor microenvironment increases which may lead to the development of more effective therapies. In collaboration with Maine Medical Center Research Institute (MMCRI), we have identified 20 pancreatic cancer patients. In these 20 pancreatic cancer patients, Dr. Jones, a pathologist at MMCRI has identified normal adjacent pancreas, fibrotic pancreas tissue, and tumorous pancreas tissue. Using SHG imaging microscopy with an 890 nm excitation laser and collection via a 445/20 bandpass filter in the forward direction, we imaged the collagen primarily around pancreatic ductal structures, which studies have shown to be common tumor origin sites1. OrientationJ, a Java plugin in ImageJ, uses a structure tensor to quantify morphological changes in the collagen fibers within the extracellular matrix. In this analysis, the quantitative orientation measurement function built into OrientationJ provides orientation and average coherency values for regions of interest selected by the user. Collagen within the ECM generally has a random, basketweave pattern, but in other cancerous tissues there are studies which show a correlation between progression of cancer and the increased aligning of the collagen fibers1–3. We found that quantitative biophysical alterations that are distinct between cancerous, fibrotic, and normal adjacent tissues, but the results were inconclusive. With a larger sample size, the results might demonstrate the pattern in the data more conclusively.
Location of Publication:
URL to Thesis:
https://digitalcommons.library.umaine.edu/honors/693/