(Coming Soonish)

(Coming Soon)

A trajectory-constrained control optimization framework for MRI acquisition. Given a target multidimensional Fourier-space trajectory (a standard MRI approach), the method formulates and solves an optimization problem based in a phase-plane representation. OPTIKS facilitates optimization of objectives dependent on the time domain gradient waveform and the arc-length domain k-space speed. We apply OPTIKS to meet both patient and system safety standards while performing fast imaging. The design method implementation is made available as an open source Python package through GitHub. 

A collection of work designing controls for MRI imaging tasks which minimize system mechanical resonance. This includes design of diffusion lobes and gradient rewinders as convex optimization problems, and a heuristic approach to safe spiral waveform design.

A recurrent neural network for automatic polyphonic music transcription trained using the MusicNet dataset. Special interest is given to the learned temporal structure of music. Produced in fulfillment of the Stanford CS 229: Machine Learning final project.

A novel method for single-shot high-dynamic-range imaging is presented. This method is demonstrated to produce hdr images from a single shot without explicit use of deep learning, relying on spatially varying exposure and compressed sensing techniques. Produced in fulfillment of the Stanford EE 367: Computational Imaging final project.

A fully autonomous "search and rescue" mission was carried out in simulation.  This work incorporates techniques such as autonomous frontier exploration, A-star path planning coupled with real trajectory controllers, and EKF SLAM. Produced in fulfillment of the Stanford AA 274: Principles of Robot Autonomy final project.

A fast-field cycling method for molecular imaging using MRI. This MSc thesis work focused on the effects of field-shift inhomogeneities in dreMR imaging and minimization of such inhomogeneities through an improved design method.