for Administrative Supplement: Current methods identified at late to identify pancreatic cancer are suboptimal stage resulting in a substantially unmet clinical with the majority of cases need. The current imaging approaches are often limited in spatiotemporal resolution and specificity with high inter- and intra-reader variability in radiological exams that often result in flawed evaluation in identifying pancreatic cancer. Our original grant R01EB020125 aimed to utilize UPRT nanoparticles containing IR780 dye to detect pancreatic cancer using Multispectral optoacoustic tomography (MSOT) imaging. The objective of our administrative supplement is to develop machine learning algorithms to accurately, objectively and consistently assess and distinguish pancreatic cancer versus normal pancreas as utilizing MSOT images. Building upon our experience in theranostic nanoparticles, MSOT imaging, and machine and deep learning, the focus of this supplement is to identify molecular features of pancreatic cancer using MSOT. As MSOT is a new imaging modality, interpreting its images will be challenging for medical professionals. Therefore, we will develop a computer-assisted image analysis (CAIA) system which will help physicians to interpret these images accurately and consistently, minimizing inter-reader variability. Similarly, we will develop and evaluate a machine-learning classifier to quantitatively identify pancreatic cancer. Together, these studies aim to optimize and validate our novel MSOT imaging combined with machine learning to identify pancreatic cancer.
This project extends our existing grant which develops a pancreatic tumor targeted nanoparticle detectable using multispectral optoacoustic tomography to include evaluation of the images using deep learning. The deep learning algorithms will segment possible pancreatic cancer regions and identify features to characterize possible pancreatic cancer regions.
