|M.Sc Student||Drozdov Gilad|
|Subject||Optoacoustic Tomography with Negatively Focused Detectors|
|Department||Department of Electrical Engineering||Supervisor||Professor Amir Rosenthal|
|Full Thesis text|
Optical imaging is an essential tool for biological discovery with growing clinical applications. One of the fundamental challenges in performing optical imaging at depths beyond 1 mm is that light scattering by tissue heterogeneity leads to loss of spatial coherence. Accordingly, deep-tissue optical imaging is characterized by diffusive light propagation and low spatial resolution of the recovered image, limiting its usefulness in biomedical applications. Optoacoustic tomography (OAT) is a hybrid non-ionizing imaging modality that enables visualizing optical absorption with typical ultrasound resolutions in the diffusive regime of light. In OAT, the excitation is performed by light pulses that generate acoustic sources within the tissue whose strength is proportional to the optical absorption. By measuring the acoustic waves, whose frequencies are in the ultrasound regime, on the tissue surface, an image of the acoustic sources may be formed via tomographic algorithms. Because the optoacoustic effect is weak, deep-tissue imaging often requires using detectors significantly larger than the acoustic wavelength, often at the price of reduced tangential resolution. Negatively focused (NF) detectors have been shown to mitigate the loss of resolution while achieving high sensitivity and broad acceptance angles. Commonly, image reconstruction with NF detectors is performed by using the virtual-detector approach, which does not accurately account for the detector geometry, leading to image artifacts.
In this work we present a theoretical study of the effect negatively focused detectors have on OAT image reconstruction. As part of our study, we developed analytical time-domain formulae for the spatially dependent impulse response of NF detectors in two and three dimensions, supplemented by asymptotic expressions in the frequency domain. Our analysis elucidates the working principle behind the virtual-detector approximation and the origin of image artifacts when this approximation is used. While the principle behind the virtual detector approach is valid for both two- and three-dimensional imaging scenarios, we show that the artifacts associated with it are of different nature. Based on our analysis, we introduce a simple correction to the virtual-detector approximation that significantly enhances image contrast and reduces artifacts. Our developments are also applicable in superior image reconstruction under the model-based approach for image recovery, where the spatial-impulse response of the transducer can be implicitly integrated into the inversion scheme.