Optimal sensor locations and ultrasound tomography set-up design for limited data problems

摘要

This work addresses the issue of limited data problem (of ultrasonic tomography) especially for the case when the detector-emitter pairs are only a few in numbers. It is shown that optimized locations (sensor array design with optimized non-uniform places) result in better reconstruction compared to the uniform configuration of detectors. Parallel scanning mode (4 acoustic transducer pairs for 36 views only) is used to perform a simulation study. An interface of genetic algorithm and entropy maximization process is achieved to estimate these optimal locations. Delaunay triangulation and shape function interpolation is used to discretize and connect the image space with data space. Strip integration is used to include the diverging beam width effect. Smearing is also incorporated for noise filtering during the reconstruction process. The location that provides minimum l_2 norm (in the reconstructions) is taken as the solution for optimal design. Ultrasound tomography setup is developed with uniform and optimal sensors alignment. Specimen with known locations is used to acquire the real world data. Tomographic reconstruction from optimal case shows promising results over uniform arrangement. The modified design is also has an added economical advantage since focused beam transducers are expensive and cost increases with the desired focal length.