Abstract
Medical ultrasound has many benefits over other imaging modalities, including lack of ionizing radiation, relative portability and low cost. However, the majority of ultrasound scanners image only a 2D plane of tissue, which moves with the probe. This arrangement, along with the remote display screen, requires a high level of skill and experience to operate. Some recently introduced scanners have more intuitive 3D operating modes, but are often bulky and expensive. In this dissertation, several of the fundamental chal- lenges of handheld, intuitive 3D ultrasound imaging systems are addressed, including energy efficient 2D beamforming, enhanced motion tracking using sector-scan probes, and optimal combination of motion estimates from various sensor modalities. In the field of vascular ultrasound imaging, the size and portability of ultrasound devices is limited by the energy cost of beamforming, particularly when transducers with many thousands of elements are involved. A separable approach to 2D beamforming, optimized for complex short-time-sequence signals is developed that reduces the energy cost of beamforming by a factor of 20, enabling real-time imaging in a 170 g device with multi-hour battery life. Imaging of spinal bone anatomy has poor performance for most ultrasound systems due to extremely bright bone reflections and systems optimized for tissue. Recognizing that a mechanically-scanned single piston transducer has intrinsic contrast advantages when imaging bone, techniques are developed to improve motion estimation using sector-scan ultrasound data, so that handheld freehand 3D spinal bone imaging is enabled. Finally, to address anisotropic motion estimation resolution using ultrasound alone, other sensor modalities (camera, accelerometer) are optimally combined to produce a handheld 3D imaging system capable of real-time guidance of epidural anesthesia procedures, with an RMS bone surface localization error of only 2.2 mm. These capabilities are demonstrated in a handheld battery-powered prototype with real-time 3D bone surface display. Initial in-vivo 2D and 3D images demonstrate feasibility of the device and imaging methodology.
