Supervisors: Dr. Sergio Sanabria, Prof. Dr. Orcun Goksel
Breast cancer has threatened to be one of the deadliest diseases to affect mankind in the coming decades. There are various methods and techniques that have been invented to diagnose breast cancer early. Ultrasound imaging stands out as one of the promising techniques in terms of lower risk and cost-effectiveness and faster turnaround time. However accuracy of various conventional ultrasound imaging methods and techniques in predicting harmful cancer tumors is somewhat less-satisfactory. The Computer-assisted Applications in Medicine (CAiM) group at ETH Zurich developed a speed of sound ultrasound imaging method that addresses accuracy concerns of conventional ultrasound method. The new algorithmic approach was initially tested offline on a research ultrasound device and a separate personal computer system. This thesis work carried out at CAiM in collaboration with ultrasound device manufacturer Fukuda Denshi. The thesis contributes to further advancement of the project with integration of an automatic SoS imaging pipeline as an add-on to the ultrasound machine UF-760AG. The project first addresses the hardware interface between the UF-760AG and a dedicated computing Platform (RADA), which captures real time synthetic aperture data from the UF-760AG. First task was the research and procurement of appropriate hardware components of the computer system, including dimensioning DMA and storage components so that the frames output by UF-760AG can be both directly mapped to the RAM memory of the RADA system and simultaneously permanently stored in a hard disk grid for posterior follow-up. Apart from standard components, a hardware connector module was developed to link the data transfer between UF-760AG and RADA systems. The second task was to define a software architecture to pipeline the data processing of the incoming synthetic data frames from the UF-760AG in the RADA system. Synchronization aspects between the incoming data flow and the processing algorithms were solved, allowing to drop frames when necessary and avoid visualization glitches. Finally, a speed of sound software algorithm in Matlab was integrated into the support environment that manages input and output data, and acceleration mechanisms, such as code optimization and GPUs processing were incorporated to the new architecture.