At my current role, I translate business challenges in the ultrasound domain to a scientific problem. After prototyping an algorithm, I implement the solution as part of our application. I also help out with software development tickets. Below are some of my projects:
I designed and implemented a machine learning framework for an image search service. The company and technology was later acquired by Xperi. Below are some of my projects:
My project is under a non-dislosure agreement. The company was disolved due to lack of funding.
I was doing research on image-based techniques for prostate interventions. As part of my PhD, I wrote a clinical software to collect B-mode, RF time-series and elastography images during a prostate biopsy. Using this dataset, I developed image registration methods to compensate for motion and deformations during prostate biopsies. I was also involved in machine learning projects for detection of tumors in ultrasound data. My Google Scholar includes a list of published papers from my PhD.
My thesis was on image registration methods for spinal interventions. I developed a generative model for spinal anatomy and by fitting and overlaying it on ultrasound images, showed that it could be used to interpret ultrasound images during spinal injections.
PyCPD is a pure numpy implementation of the coherent point drift (CPD) algorithm. I developed it as part of my invited talk at the University of British Columbia.
This is a gym environment for training a reinforcement learning to find matching patches in two identical translated MNIST pairs. The environment consists of a pair of mnist digits against a zero background. The dimension of each image is 40-by-40, mnist digits are standard 28-by-28 pixels. The location of each mnist digit is uniformly random and independent.
GMM-FEM is a point cloud registration technique that I developed as part of my PhD thesis. It is implemented in Matlab, with computationally intensive parts being developed in C++.