A new segmental pulse wave velocity imaging based on ultrafast sliding angle compound technique

서현아 서강대학교 대학원 2018 국내석사

A study on programmable system-on-chip implementation of a wearable ultrasound bladder monitoring system

한혜윤 서강대학교 대학원 2021 국내석사

Highly increasing focus on quality of life, the needs of connected healthcare is vastly emerging. This trend has brought a number of new applications of monitoring a patient’s health condition in real time. Medical ultrasound imaging has several advantages over other imaging modalities: it displays images in real time without ionizing radiation, costs an affordable price and is physically portable. This shows that ultrasound systems are suitable for this current trend of patient-centered wearable health care system. For managing urination disorders, it is important to routinely measure the bladder volume. Recently developed portable ultrasound devices can periodically measure the bladder volume. However, they are not suitable for bladder monitoring in point-of-care settings due to their size, price and usability. For bladder monitoring, wearable ultrasound devices can be attached to the lower part of the patient’s abdomen to continuously measure the bladder volume. In this paper, a programmable system-on-chip (SoC) solution to minimize the size and cost of a wearable bladder ultrasound imaging system is proposed. In the proposed programmable SoC solution, the recently introduced processing platform (Zynq-7000 family, Xilinx Inc.) where, in addition to a 28-nm-flexible programmable gate array (FPGA) module, a dual ARM Cortex-A9 MPcore processing module is integrated. The core ultrasound signal and image processing blocks for a wearable bladder monitoring system were implemented on the dual ARM Cortex-A0 Mpcore processor by using SIMD parallel processing with NEON intrinsic and dual-core processing. For generating an ultrasound bladder image, with parallel programming with NEON intrinsic on a dual-core processor, it takes less than 98.21 ms, resulting in >10 frames/s. The results showed that the proposed programmable SoC solution can reconstruct a real-time bladder ultrasound image with the minimum hardware and software resources.

다층 매질 초음파 영상에 대한 초음파 속도 추정에 관한 연구

진성민 서강대학교 대학원 2013 국내석사

일반적인 초음파 의료 영상 장치는 수신 빔 집속 시 초음파의 속도를 1540m/s의 고정된 상수로 가정하여 전체 영상에 적용한다. 하지만 다양한 매질로 이루어진 인체에서는 각각의 매질을 진행하는 과정에서 초음파 진행 속도가 변화하게 되며, 이는 고정된 속도로 수신 빔 집속을 수행하는 경우 영상에서의 화질 저하를 야기한다. 이러한 영상의 해상도 저하를 제거하기 위하여 영상에서 특정 깊이 관심 영역을 설정하여 최적의 초음파 속도를 추정하는 다양한 기법들이 제안되고 있다. 하지만, 이러한 방식들은 관심 영역의 영상 질만을 개선하여 영상의 전반적인 성능의 향상을 얻을 수 없는 문제가 있으며, 또한 모든 영상 집속점에 대하여 최적 초음파 속도를 추정하기에는 많은 연산 양과 후처리 과정에서의 복잡도가 증가된다. 본 논문에서는 상대적으로 균일한 조직 특성을 갖는 매질이 상하층부로 구분되어 있는 복부 초음파에 대하여, 각각의 매질 (e.g., 지방, 간 등) 에 대한 절대 초음파 진행 속도를 각기 추정하고, 이를 통하여 추정된 모든 깊이의 집속점에 대해 최적의 누적 초음파 속도로 수신 빔 집속에 적용하는 새로운 방법을 제안하고자 한다. 제안하는 속도 추정 기법은 먼저 조직간의 경계 깊이를 추정하고, 상, 하층 매질에서 단일 및 다중 관심 영역을 설정하여 각 최적 누적 초음파 속도를 추정한다. 각각 추정된 하층부에서의 다중 관심영역 최적 누적 초음파 속도는 진행한 매질의 기하학적 비율에 따라 결정되므로 이들의 선형 관계식을 통해 균일하다고 가정된 각 매질의 절대 초음파 속도를 추정할 수 있다. 결과적으로 이렇게 얻어진 각 매질의 절대 초음파 속도와 조직의 기하학적인 모델링을 통해 모든 깊이 집속점에 대한 최적 누적 초음파 속도를 추정할 수 있다. 제안한 속도 추정 기법은 각 선형 및 곡형 프로브에 대하여 인체를 모사한 팬텀 실험 및 인체 실험을 통해 검증되었다. 제안된 적응 초음파 속도 추정 및 적용 알고리즘은 수신 빔 집속을 수행한 영상은 고정된 1540m/s로 수신 빔 집속된 영상과, 기존 초음파 추정 및 적용 알고리즘을 통해 추정된 최적 초음파 속도로 수신 빔 집속된 영상과 비교하였다. 영상의 화질에 대한 정량적인 평가를 위하여 측 방향 해상도와 영상의 평균 대조도를 측정하여 제안하는 속도 추정 기법을 검증하였다. Ultrasonography has conducted a critical role in diagnosing abdominal disorders due to its noninvasive, real-time, and low cost properties. However, for the obese patients with thick fat layer, sufficient image quality cannot be acquired with conventional beamforming due to phase aberration caused by the sound speed difference between tissues, e.g., 1580 and 1450m/s for liver and fat, respectively. For this, various sound speed correction methods (SSC) that estimate the accumulated sound speed for a region-of-interest (ROI) were previously proposed. However, the image quality improvement is limited only for the ROI depth. In this paper, we present the adaptive sound speed correction (ASSC) method which can enhance the image quality for whole depths by estimating two accumulated sound speeds of different depths in lower layer. Since the accumulated sound speeds contain the respective contributions of layers, optimal sound speed for each depth can be estimated by solving contribution equations. To evaluate the proposed method, phantom study was conducted with pre-beamformed RF data acquired with a SonixTouch research package (Ultrasonix Corp., Canada) with a linear and convex probe for the In-Vitro data from a gel pad-stacked tissue mimicking phantom (Parker Lab. Inc., USA and 040GSE, CIRS, USA), whose sound speeds are 1610 and 1450 m/s, respectively and In-Vivo data. The ASSC method indicates more uniform and improved mean spatial resolution and its standard deviation among axial depths than the SSC method. The result indicates that the proposed ASSC method can provide better spatial resolution than conventional SSC method.

Design and implementation of a wearable ultrasound bladder monitoring system with single-chip solution

윤종민 서강대학교 대학원 2020 국내석사

배뇨활동은 삶의 질에 크게 관여한다. 그러나 배뇨장애를 겪게 되면 일상 생활에 많은 제약이 생길 뿐만 아니라 사회적인 인식이나 정신적인 스트레스로 인해서 삶의 질이 크게 저하된다. 배뇨 장애로 인한 삶의 질의 저하를 막는 방법으로는 방광에 있는 오줌의 양을 모니터링하여 배뇨장애를 겪고 있는 환자들이 정상적인 배뇨 활동을 할 수 있도록 유도하는 것이다. 대표적인 모니터링 시스템은 초음파 시스템이다. 초음파는 비침습적이고 방사능에 대한 위험이 없고, 다른 의료용 시스템과 충돌할 위험이 없다. 또한 초음파 영상을 이용하여 방광의 부피를 측정하는 방법은 이전부터 연구되어 왔다. 그러나 기존의 초음파 방광 모니터링 시스템은 센서의 초기 위치 선정의 어려움이나 정확한 부피 측정이 불가능한 한계점이 있었으며, 초음파 영상 시스템의 경우는 크기와 비용적 측면에서 모니터링 시스템에 적합하지 않았다. 이 논문에서는 기존의 초음파 모니터링 시스템과 초음파 영상 시스템의 한계를 극복하기 위한 초음파 B-mode 이미지를 기반한 웨어러블 방광 모니터링 시스템을 설계하였다. 시스템의 크기와 가격, 그리고 전력 소모를 줄이기 위해서 이미징 시스템이 필요한 처리과정을 Zynq-SoC을 활용하여 one-chip 솔루션을 설계하였다.

혈관 내 초음파 영상을 위한 합성 구경 집속과 Coherence Factor Weighting 적용 조건에 대한 연구

강성우 서강대학교 대학원 2020 국내석사

합성 구경 집속(Synthetic Aperture Focusing, SAF)과 coherence factor weight(CFW)은 초음파 이미지의 측 방향 해상도를 개선하는데 사용되었다. 두 방법이 다중 소자 기반의 초음파 영상에 효과적이지만, 많은 연구자들은 혈관 내 초음파(IVUS) 영상을 포함하여, 단일 소자 기반의 영상을 위한 방법에 적용하였다. 단일 소자 기반의 영상의 경우, CFW는 일반적으로 SAF에 의해 획득되고, SAF를 수행한 이후에 적용되는 주사선으로부터 계산되며, 이를 SAF-CFW 방법이라고 한다. 본 논문에서는 단일 소자 기반 영상에 대한 SAF 및 CFW의 효과를 이론적 모델을 도출하여 사용하였으며, 이 모델은 SAF가 IVUS 영상에 효과적이지 않다는 설명을 위해 사용되었다. 또한 IVUS 영상의 경우 SAF 지연시간 계산 없이 인접한 주사선에서 CFW를 계산하여 적용해야한다. 이는 SAF-CFW 방법이 IVUS 영상에 효과적이지 않음을 보였다. 시뮬레이션 및 실험에서 각 기법은 집속 되지 않은 변환기와 집속 된 변환기에 적용되었다. 결과적으로 두 변환기는 각각의 방법에 대해 유사한 결과를 나타냈고, 기법을 적용하지 않은 집속 된 변환기가 시스템 구현에 대한 복잡성을 고려하였을 때, IVUS 영상에 유리하다는 결론을 내렸다. Synthetic aperture focusing (SAF) and coherence factor weighting (CFW) have been used to improve the lateral resolution of ultrasound images. Although the two methods are effective for array-based ultrasound imaging, many researchers have also employed the methods for single-element-based imaging including intravascular ultrasound (IVUS) imaging. For single-element-based imaging, CFW is generally calculated from the scanlines obtained by SAF and applied after carrying out SAF, which is called a SAF-CFW method. In the paper, a theoretical model was derived to explore the effectiveness of SAF and CFW for single-element-based imaging, and the model was used to explain that SAF is not effective for IVUS imaging. Also, it was found that, for IVUS imaging, CFW should be calculated from adjacent scanlines of a target scanline without SAF delay and applied to the target scanline. This means that the SAF-CWF method is also not effective for IVUS imaging. In simulation and experiments to verify the findings, the SAF, SAF-CFW, and CFW methods were applied to both conventional flat aperture and focused IVUS transducers. As a result, both transducers exhibited similar trends for each method, but it could be concluded that focused IVUS transducers without the application of those methods are more advantageous, considering the system complexity in the implementation of such methods.

Ultrafast ultrasound imaging for assessment of arterial wall stiffness and blood flow

강진범 서강대학교 대학원 2019 국내박사

In the past decades, manifold medical imaging modalities have been developed to evaluate cardiovascular disease (CVD), e.g., X-rays, magnetic resonance imaging (MRI), Doppler ultrasonography, intravascular ultrasound (IVUS) and computed tomography (CT) angiography. Among these techniques, real-time ultrasound imaging with many advantages such as non-invasive, cost effective, portability and no-ionizing radiation has rapidly evolved as screening tools that are useful for major or peripheral vasculature assessment. The ultrasound imaging can provide morphological information for the vessel lumen (stenosis or occlusion or aneurysm), for the vessel wall in terms of calcification, intima-media thickness (IMT), atherosclerotic plaques and wall edema. It also offers physiological information such as blood flow, volume and velocity, thereby differentiating hemodynamically significant from non-significant stenosis. Recently, more useful ultrasound cardiovascular imaging techniques have been introduced. 3-D ultrasound imaging can visualize the vascular geometry and allow to measure plaque volume. Vessel wall motion imaging, which estimates the distension waveform during diastolic and systolic cardiac phase, evaluates vessel wall stiffness indirectly. Vector Doppler imaging assesses the full span directions of complex blood flow in stenosis or branch vessels. In addition, shear-wave elastography, which enables the estimation of local stiffness, may be useful for detection of unstable plaques. However, the current ultrasound cardiovascular imaging techniques are typically based on the focused imaging method that each image scanlines of 2-D image are sequentially reconstructed by individual transmission and reception. Therefore, the frame rate is significantly reduced by a factor of a number of transmit events since the conventional imaging method produces high-quality images by a few tens of focused beam transmissions. This lower frame rate occurs important limitations for cardiovascular applications because the transient motion of organs, intrinsic mechanical waves from the heart and diverse blood flow velocities along the vascular tree require real-time tracking with very high spatiotemporal resolution. In this thesis, we have developed new three ultrasound imaging techniques, which are focused on evaluating of vessel wall stiffness and blood flow, using a very high spatio-temporal ultrafast data acquisition (i.e., a few kilohertz). First, a non-invasive arterial wall stiffness assessment method based on a pulse wave velocity (PWV) measurement, which has the potential to directly evaluate the segmental stiffness changes in arterial vessel walls, was developed. Second, a high pulse-repetition frequency (PRF) ultrafast blood flow imaging method was developed to enable a fully qualitative, retrospective flow assessment. Third, a wide field-of-view (FOV) microvascular imaging method based on ultrafast diverging-wave transmission, which is useful for the depiction of abdominal microvasculature, was developed. The developed techniques were assessed with in vitro phantom and in vivo data acquired using commercial ultrasound research platform. These methods demonstrated its potential as useful diagnostic tools to evaluate CVDs.

반사 스펙트럼 추정을 통한 혈관 영상화 기법에 관한 연구

김초예 서강대학교 대학원 2013 국내석사

Multispectral imaging technique has been applied to various fields and still has great potential as it contains larger information than images with visible ray. However, in order to acquire multispectral data, long acquisition time is required and the mobility of the equipment is poor when compared to acquiring conventional RGB(red-green-blue) data. To overcome these drawbacks, techniques for estimating reflectance spectrum with RGB data have been proposed. Nonetheless, the estimation techniques have been used only in few applications due to the fact that the range of wavelengths which can be estimated is limited and the accuracy of the estimation falls behind the measurement values. Among countless applications, visualizing veins underneath the skin is achievable by multispectral imaging technique due to the special absorption coefficient of hemoglobin. Vein images using multispectral imaging technique illuminates light with specific wavelength which is appropriate for acquiring vein information and detect light of that wavelength. However, in this case, special equipment is required for acquiring vein images which causes inconvenience in use (e.g., cost, usage location, possible applications). On the contrary to this, if vein images can be acquired with RGB data, these inconveniences can be minimized.Therefore, in this thesis, vein visualization method is proposed using RGB data which can be acquired with conventional camera. Veins located underneath the skin are visualized with the estimated reflectance spectrum from conventional RGB data. For the estimation of reflectance spectrum, Wiener estimation method was used. In addition, histogram stretching was performed for adjusting the contrast and the brightness of the image. To evaluate the performance of the proposed method, experiments were conducted with color checker and vein images. In the color checker experiment, the average error was 12.0%. For the vein image experiment, images were taken from four body sites (i.e., inside elbow, wrist, palm, back of hand). With all four sites, the proposed method could visualize more veins with higher conspicuity when compared to naked eyesight. In addition, the accuracy of the proposed method was proven by comparing captured images with medical ultrasound system. 다중 스펙트럼 영상(multispectral image)은 가시광선 영상에 비해 훨씬 많은 정보를 포함하고 있어 농업, 광물학, 물리학 등 여러 분야에서 응용되고 있다. 하지만 많은 정보를 가지고 있는 다중 스펙트럼 영상은 일반 RGB(red-green-blue) 영상에 비해 데이터 획득 시간이 길며 장비의 이동성이 떨어진다. 이를 극복하기 위해 RGB 영상으로부터 반사 스펙트럼(reflectance spectrum)을 추정하는 연구가 이루어져왔다. 반면 추정된 반사 스펙트럼의 경우 추정할 수 있는 파장이 제한되거나 정확성이 떨어지는 등의 단점이 있어 제한된 분야에서만 사용되어왔다. 다중 스펙트럼 영상의 다양한 응용분야 중 헤모글로빈의 특수한 흡수 스펙트럼을 이용하여 피부 아래 혈관을 가시화 하는 기술이 있다. 혈관영상화에 적합한 파장의 빛을 조사하고 그 파장 대역의 신호를 감지하는 센서를 사용하여 혈관정보를 획득한다. 하지만 이 경우 혈관영상을 획득하기 위해 특수한 장비가 필요하기 때문에 비용, 사용 장소, 응용 분야 등 사용이 제한적이다. 반면 RGB 영상을 이용한 반사 스펙트럼을 추정으로 혈관영상을 획득 할 경우 이러한 제한점을 극복할 수 있다. 본 논문에서는 일반 카메라를 이용하여 획득한 RGB 영상으로부터 반사 스펙트럼을 추정하고 얕은 깊이의 혈관을 가시화 하는 방법을 제안한다. RGB 데이터로부터 반사 스펙트럼 데이터를 추정하기 위해 비교적 간단하면서도 좋은 성능을 내는 위너 추정(Wiener estimation) 기법을 이용하였으며 획득한 혈관영상의 대조도와 밝기를 조절하기 위하여 히스토그램 확장(histogram stretching)을 수행하였다. 제안한 혈관영상 기법의 성능을 확인하기 위해 컬러 체커(color checker)실험과 인체 혈관영상 실험을 진행했다. 컬러 체커 실험에서는 측정 스펙트럼 데이터와 추정 스펙트럼 데이터의 오차가 평균 12.0%로 나타났다. 혈관영상 실험에서는 팔꿈치 안쪽, 손목, 손바닥, 손등을 촬영하여 실험하였으며, 모든 실험영상에서 더 많은 혈관이 더 선명하게 보임을 확인하였다. 또한 초음파 영상 장치와의 비교를 통하여 제안하는 혈관영상화 방법의 정확성을 검증했다.

A Study on an Unsupervised Learning-Based Reverberation Artifact Reduction Method in Medical Ultrasound B-mode Imaging

김상민 서강대학교 일반대학원 2024 국내석사

Reverberation artifacts in medical ultrasound B-mode images, caused by multiple repetitive reflections of the echo signal, significantly degrade image quality and hinder accurate diagnoses. Various studies, particularly those focusing on deep learning, have proposed techniques to mitigate these artifacts. Deep learning-based methods face a primary challenge in training strategy, which is categorized into supervised, semi-supervised, and unsupervised approaches. While supervised learning is simple and effective when input and ground-truth data are available, it is often impractical for artifact reduction due to the difficulty of obtaining suitable data. Conversely, unsupervised learning presents a promising alternative to overcome these data acquisition challenges. A recent advancement is deep coherence learning (DCL), an unsupervised technique specifically for enhancing ultrasound imaging quality. In this thesis, a custom phantom is designed to leverage DCL for suppressing reverberation artifacts. The effectiveness of the developed deep coherence learning with reverberation dataset (DCL-Reverb) was assessed using real-world experimental data. Quantitatively, DCL-Reverb demonstrated higher contrast-to-noise ratio (CNR) and generalized contrast-to-noise ratio (gCNR) compared to conventional methods. Qualitatively, it also achieved clearer B-mode images and superior artifact suppression in axial profiles.

Efficient embedded segmentation network for a wearable ultrasound bladder monitoring system

김규태 서강대학교 대학원 2021 국내석사

Medical ultrasound imaging system is safe and simple to implement on a single chip so that has a great potential for being used as the next generation connected healthcare device. However, there is an obstacle that a doctor's diagnosis is required to provide clinically useful information in ultrasound images. To resolve this problem, various studies using deep learning for diagnosis are actively being conducted. However, deep learning is difficult to use in edge devices due to its high computational complexity. Deep learning tasks are mainly performed based on general-purpose computing on the graphics processing unit (GPGPU) of the graphics processing unit (GPU), which can handle parallel tasks. However, it is not a proper architecture for deep learning tasks since it is inefficient for use in edge devices. In this paper, the architecture of a new AI accelerator for a system on chip (SoC), which can perform efficiently deep learning tasks without the GPU, is proposed. The proposed AI accelerator uses a FPGA module and an ARM core processor, and quickly computes separable convolutions with low hardware utilization to efficiently perform deep learning inference tasks. In addition, deep learning inference tasks are implemented using lightweight segmentation networks based on CNN at AI accelerators. This network has a U-Net-based deep learning architecture, and it is designed for AI accelerators. Finally, ultrasound bladder image segmentation is implemented using the AI accelerator and the lightweight segmentation network in a wearable ultrasound bladder system, which is as a connected healthcare device application, and showing about 10% reduction in performance compared to that from commercial CPUs with much less resource utilization.

Study on a New Real-Time Spatial Compounding Method with Elevational Synthetic Aperture Focusing for an Automated Breast Ultrasound System

김종석 서강대학교 대학원 2023 국내석사

X-ray mammography has been mainly used for early diagnosis of breast cancer, which has the world's number one cancer incidence and mortality rate. However, for women under 40 years of age or with dense breasts, the diagnosis accuracy of X-ray mammography becomes substantially low. The recently developed automated breast ultrasound system (ABUS) is used as a complementary diagnostic tool. In ABUS, multi-angle spatial compounding, in which multiple transmit beams are steered with pre-determined angles and the receive beams are coherently summed together, is applied for enhancing image quality while reducing shadow artifacts caused by absorption of sound waves from objects, such as nipple and mass. However, this method suffers from the reduction in the intensity from the target lesions and the blurring since the transducer mechanically moves and compounds between frames in different positions. Therefore, in this paper, we propose a real-time spatial compounding method with elevational synthetic aperture focusing, which can compensate the unwanted reduction in target intensities by performing synthetic aperture focusing in the elevation direction. In addition, beamformed frames in the lateral direction are subjected to bidirectional dynamic focusing by calculating transmit and receive delays in the elevation direction. Finally, spatial compounding is applied to those frames. With the proposed method, spatial resolution, signal-to-noise ratio, and contrast can be improved by compensating intensity loss of the target yielded from the ABUS. Moreover, the proposed method can facilitate the diagnosis improvement in early detection of breast cancers.