Ashley Chen
Pronouns: she/her
Research Mentor(s): Arvind Rao
Co-Presenter: Yadav, Omkar
Research Mentor School/College/Department: Computational Medicine and Bioniformatics / Medicine
Presentation Date: April 20
Presentation Type: Oral5
Session: Session 3 – 1:40pm – 2:30 pm
Room: Breakout room 4
Authors: Ashley Chen, Avery Maddox, Andi Xu, Omkar Yadav, Loria Sun, Anirudh Attaluri, Arvind Rao
Presenter: 6
Abstract
Urothelial carcinoma, more commonly known as bladder cancer, is a heterogeneous disease characterized by genomic instability and a high mutation rate at the cellular level. The presence of heterogeneous tumour populations complicates the detection of mutations since they can occur with low frequency and be indistinguishable from background noise. Current diagnostic procedures do not provide automated methods of determining tumor homogeneity and often require further manual examination which is time-consuming and costly. In recent years, digital analysis of pathology whole-slide images has been a growing field of interest for cancer diagnosis and treatment. Data obtained from real-life patient databases can be used to train and develop machine learning data models that are able to predict genetic mutations automatically. In the past 4 months, the Cancer Genome Atlas (TCGA) was conducting a system update, thus, we could not obtain a new data set for our project specifically on bladder cancer. While our team waited for image availability, our research focused on assessing the prediction accuracy and error rate of existing algorithms that permit determination of genomic/ molecular status from histology images using PyTorch and Keras. We plan to develop spatial maps to dissect H&E images into smaller-sized cells, which allows us to analyze patterns and potentially create our own genomic prediction model. Our results provide insight on the deep learning mechanisms behind automated cancer diagnosis for bladder cancer patients.
Biomedical Sciences, Engineering
