Dissertation Title:  Seeing and Shaping: Advanced Theranostic Ultrasound Modalities for Imaging, Neural Modulation, and Drug Delivery

Abstract:
Ultrasound can both see within tissue in real time and shape physiology through safely delivered mechanical energy. Directed ultrasound energy can target soft tissues to illuminate and capture reflections but also to modulate through heating and mechanical stimulation. Within the brain, ultrasound has unlocked new therapies in neuromodulation, blood-brain barrier opening for drug delivery, and functional 3D imaging. 

The core device stack of console-style electronics, rigid transducers, and visualization bound to screen displays has changed little in decades. This stagnation limits clinical impact in two distinct ways: i.) an underutilization of volumetric imaging due to bandwidth constraints, hardware complexity, and interface challenges ii.) and a lack of miniaturized, biocompatible transducer interfaces for chronic therapeutic intervention. The field thus faces a critical gap: we possess the physics and engineering capability to modulate biology, but lack the system architectures to bridge the spatiotemporal mismatch between diagnostic intent and therapeutic delivery.

This thesis poses three core research questions in the themes of seeing and shaping: Seeing: (1) Can sparse array architectures combined with compressed sensing reduce bandwidth sufficiently for real-time volumetric imaging on portable hardware, and can an augmented reality pipeline render this data with latency low enough to be functional and augmentative during live examinations? Shaping: (2) Can flexible, multi-element pMUT arrays produce significant long-term streaming and convection forces with low power consumption, and can phased-array beamforming be implemented at implantable scale to steer stimulation targets dynamically? (3) Can ultrasound-enhanced convective delivery significantly increase the diffusion radius of therapeutics across the blood-brain barrier? 

Answering these questions demands new device systems co-designed across hardware, packaging, algorithms, and user interfaces. Three integrated platforms anchor this theses: (i) a long-term ultrasound-enhanced convective injection device (LUCID), a drug-delivery implant that uses acoustic streaming and radiation forces to convectively drive large biologics and small molecules into neural tissue, offering a new tool for neuro-oncology; (ii) conformable ultrasonic femoral stimulation (CUFs), implantable pMUT neuromodulators for the PNS that deliver focal and steerable stimulation to peripheral nerves; and (iii) AR-VIU, a mixed-reality volumetric imaging system using sparse CODA-box arrays and compressed sensing to register true-scale 3D ultrasound in situ, reducing cognitive load by eliminating the screen-displacement problem. 

By treating ultrasound as a general-purpose interface for perception and intervention, this thesis aims to shift the field from screen-bound diagnostics toward embedded, closed loop theranostics.

Committee members:
Canan Dagdeviren, Associate Professor of Media Arts and Sciences, LG Career Development Professor of Media Arts and Sciences, MIT

Alan Jasanoff, Eugene McDermott Professor in the Brain Sciences and Human Behavior, Professor of Brain and Cognitive Sciences, Associate Investigator, McGovern Institute, MIT

Brian Anthony, Principal Research Scientist, Department of Mechanical Engineering, Associate Principal Research Scientist, MIT IMES, Associate Director, MIT.nano, MIT