Responsibilities

As a PhD Graduate Research Assistant under John Shepherd PhD, I conduct imaging research which is primarily focused on breast cancer. Our research group seeks to discover novel imaging bio-markers to detect cancer and asses risk. Machine and deep learning are big components of my research as it is a powerful tool for analyzing high dimensionality data that is images. My background in high performance computing and software engineering allows me to fully leverage our GPU clusters to rapidly build and train machine learning models.

Projects

• Breast Thickness Tomosynthesis - Breast density is a biomarker that is associated with cancer risk. Accurate point thickness measurements of the breast are necessary for calculating density. We use nine views from a sinogram to solve and inverse problem and derive point thickness measurements.

  • I developed an unsupervised machine learning model using PCA to characterize sinograms resulting from breast of different thickness, widths, and densities.
  • Work was presented in an abstract detailed HERE
• 3CB - 3 Compartment Breast or 3CB is a project that can render breast lesions specific to the three major components that make up breast tissue. These components include water, lipid, and protein and quantifying the amount of each component helps to assess cancer risk and need for biopsies.
  • For this project I refactored the existing 3D rendering code and plan to incorporate the breast thickness tomosynthesis algorithm to get accurate thicknesses and make up of breast lesions.
• Interval vs Screening Cancer Detection - Screening cancers are cancers detected from a screening mammogram. Interval cancers are a case of breast cancers that found after a previous screening mammogram reported a negative finding. We use deep learning and neural networks to search for signals in interval cancer cases to try and catch the cancers early and at the time of screening.
  • Our group is fortunate enough to work with the largest data set of it’s kind for this project. I helped with high throughput preprocessing and cleaning of our large image dataset . I also research and explore of different deep learning techniques and models to achieve high sensitivity and specificity for interval cancer cases.