서강대학교

초록/요약

Medical Ultrasound imaging has several advantages over other modalities: it displays images in real-time, emits no ionizing radiation, costs an affordable price and is physically portable. This shows that ultrasound systems are suitable for this current transition from the central, provider-centric and hospital-based healthcare to distributed, patient-centered and home-based one. With the increased interest in wearable devices, wearable ultrasound devices have been previously proposed for various monitoring applications and ultrasound therapies. These all require smaller in size and cheaper in cost. Xilinx, Inc. recently introduced Zynq-7000 family which is the industry’s first extensible processing platform: dual ARM Cortex-A9 MPcore processing system tightly integrated with 28nm low-power programmable logic. In this paper, a low-cost wearable ultrasound system working as a standalone device is presented, in which a single Zynq chip is employed. Also, four possible vectorization methods for 1.5-D FIR filtering are presented, and the combination of the two methods are proposed which is the most suitable for implementation on SIMD architecture. ARM Cortex A-9 which is built in Zynq 7000 is used to confirm that the proposed implementation strategies and the NEON code vectorized by using the VSC and VT methods shorten the execution time up to 5.95 and 7.11 times compared to the conventional C program. Using the proposed methods, it is possible to operate the maximum input size of 1024x64 in real time, and of 4096x64 at 10 fps. The quantitative evaluation shows that there is no significant difference in the images reconstructed by the reference C program with serial computing of double-precision type data and the proposed method. It is anticipated that it will be possible to use in various wearable applications, emergency medical services, animal scans, etc. where the small size and mobility with a decent quality image are required.