Research background
Ultrasound examination has become a key tool in clinical diagnosis due to its safety, non-invasiveness, lack of ionizing radiation, and ease of use. However, traditional ultrasound imaging faces a significant limitation. Due to the physical diffraction limit, its spatial resolution is inadequate for clearly visualizing the tiny microvascular networks within organs and tissues. Often just one-tenth the diameter of a human hair, these microvessels are critically associated with major diseases in organs like the heart, brain, liver, and kidneys, as changes in their microcirculatory blood flow are critical indicators of these conditions.
Super-resolution ultrasound localization microscopy (ULM) technology makes it possible to “see microvessels” by tracking micron-sized ultrasound contrast agent microbubbles and accurately mapping their movement paths, successfully producing high-definition “microvascular pathway maps”. This breakthrough has, for the first time, allowed researchers to clearly observe detailed blood flow patterns in critical regions such as brain tissue and tumor microenvironments using ultrasound imaging. However, practical application of this technology faces several major challenges:
These technical bottlenecks significantly hinder the broader clinical application of ULM. As a result, the development of a next-generation, smarter, and faster ULM technology has become a critical challenge that must be addressed.
Research progress
To address the above challenges, the research team at Nanjing University has made a key technological breakthrough by developing a velocity-constrained Kalman filtering ultrasound localization microscopy (vc-Kalman ULM) algorithm that integrates motion compensation. The main innovations of this method include:
(1) Velocity-constrained Kalman filtering tracking algorithm: The algorithm first constructs a state vector that includes the spatial position and brightness characteristics of the microbubbles. Using a recursive optimization framework of “prediction-matching-correction” and applying velocity constraint rules, it automatically filters out false signals, allowing for precise tracking of the microbubble trajectories. Imaging results from rat brain microvasculature demonstrate that the vc-Kalman ULM significantly improves microbubble tracking accuracy, clearly maps the spatial distribution and movement direction of microvascular networks, and accurately quantifies changes in microcirculatory blood flow speed.
(2) Adaptive motion compensation mechanism: This mechanism combines dynamic programming with cross-correlation search to effectively eliminate image distortions caused by physiological motions such as breathing and heartbeat. It enables the reconstruction of high-precision super-resolution blood flow pathway images even in highly motile organs s like the heart and kidneys. Microcirculatory imaging experiments in the rat renal cortex show that the dynamic programming-based cross-correlation search provides effective motion compensation, successfully reducing noise and artifacts due to normal breathing, heartbeat, and slight tissue movements. As a result, the edges of small vessels are clearly defined, with image contrast-to-noise ratio (CNR) and error metrics (nRMSE) significantly superior to those of traditional algorithms.
(3) Maintaining stability, accuracy, and clarity at low frame rates: The proposed vc-Kalman algorithm effectively overcomes the drop in trajectory matching accuracy that traditional ULM experiences at low frame rates. By combining multiple types of information—such as microbubble brightness and past positions—it reliably localizes microbubbles and images the microvascular flow even at frame rates as low as 146 Hz. This advancement greatly reduces data acquisition and transmission demands, opening the door for portable devices and real-time intraoperative imaging. The figure below compares imaging results from traditional ULM and the vc-Kalman ULM at various sampling frame rates, clearly showing the superior robustness of the vc-Kalman algorithm.
Future prospects
Looking ahead, the research team will focus on optimizing GPU acceleration using more powerful parallel computing architectures, and further exploring 3D and multi-channel imaging techniques. Leveraging these innovations, our goal is to achieve ultra-high-speed 3D ULM imaging at 2000 fps and create a portable, high-throughput, high-resolution “next-generation ultrasound microscopy platform”. This breakthrough will accelerate the clinical adoption of the technology in vital areas such as early stroke detection, tumor screening, and intraoperative navigation.
Sources: https://spj.science.org/doi/10.34133/research.0725
Research
News article
Not applicable
Velocity-Constraint Kalman Filtering for Enhanced Bubble Tracking in Motion-Compensated Ultrasound Localization Microscopy
4-Jun-2025