Ben Daya, Ibrahim2020-03-122020-03-122020-03-122020-03-11http://hdl.handle.net/10012/15694Ultrasound imaging is a valuable tool in many applications ranging from material science to medical imaging. While 2-D ultrasound imaging is more commonly used, 3-D ultrasound imaging offers unique opportunities that can only be found with the help of the extra dimension. Acquiring a 3-D ultrasound image can be done in two main ways: mechanically moving a transducer over a region of interest and using a fixed 2-D transducer. Mechanical motion introduces unwanted artifacts and increases image acquisition time, so a fixed 2-D is usually preferred. However, a fully addressed 2-D array will require a significant amount of connections and data to handle. This motivated the exploration of different simplification schemes to make 2-D arrays for 3-D ultrasound imaging feasible. A method that received a lot of attention for making real-time volumetric ultrasound imaging possible is the row-column method. The row-column method simplifies the fully addressed 2-D array by utilizing a set of 1-D arrays arranged in rows and another set in columns, one set will be responsible for transmit beamforming, while the other for receive beamforming. Using this setup, only $N+N$ connections are needed instead of $N\times N$. This simplification comes at the cost of image quality. Recent advances in row-column ultrasound imaging systems were largely focused on transducer design. However, these imaging systems face a few intrinsic challenges which cannot be addressed through transducer design alone: the issues of sparsity, speckle noise inherent to ultrasound, the spatially varying point spread function, and the ghosting artifacts inherent to the row-column method must all be taken into account. As such, strategies for tackling these intrinsic challenges in row-column imaging would be highly desired to improve imaging quality. In this thesis, we propose a novel compensated row-column ultrasound imaging system where the intrinsic characteristics of the transducer and other aspects of the physical row-column imaging apparatus are leveraged to computationally produce high quality ultrasound imagery. More specifically, the proposed system incorporates a novel conditional random field-driven computational image reconstruction component consisting of two phases: i) characterization and ii) compensation. In the characterization phase, a joint statistical image formation and noise model is introduced for characterizing the intrinsic properties of the physical row-column ultrasound imaging system. In the compensation phase, the developed joint image formation and noise model is incorporated alongside a conditional random field model within an energy minimization framework to reconstruct the compensated row-column ultrasound imagery. To explore the efficacy of the proposed concept, we introduced three different realizations of the proposed compensated row-column ultrasound imaging system. First, we introduce a compensated row-column imaging system based on a novel multilayered conditional random field driven framework to better account for local spatial relationships in the captured data. Second, we incorporated more global relationships by introducing a compensated row-column imaging system based around a novel edge-guided stochastically fully connected random field framework. Third, accounting for the case where the analytical image formation model may not optimally reflect the real-world physical system, we introduce a compensated row-column imaging system based around a data-driven spatially varying point-spread-function learning framework to better characterize the true physical image formation characteristics. While these different realizations of the compensated row-column system have their advantages and disadvantages, which will be discussed throughout this thesis, they all manage to boost the performance of the row-column method to comparable and often higher levels than the fully addressed 2-D array.enConditional Random FieldsRow-columnultrasound imagingcompensated imagingpoint spread function modellingpoint spread function learningUltrasonic imagingDoctoral Thesis