Medical Ultrasound Device

High-Frequency Vector Doppler Imaging (HFVDI)

High-frequency ultrasound imaging has been used widely in many preclinical and clinical applications. For example, the superficial tissue imaging such like the eye or skin, intravascular ultrasound or small animal imaging. Traditionally, power and color Doppler are utilized for blood flow imaging. Recently, high-frequency vector Doppler imaging (HFVDI) based on multi-beams was realized by using ultrafast system. With this technique, we can measure the blood flow from all the directions with high spatial resolution. For now, we have successfully applied it on the superficial human venous blood flow (upper video) and the mice brain imaging (below figure).

High Frequency Elastography and Pulse Wave Imaging

The mechanical properties of soft tissue are considered effective biomarkers for the diagnosis of various diseases such as ophthalmic diseases, tendon elasticity after damage and arterial viscoelasticity of multiple comorbidities. In our lab, we are devoted to do high-frequency elastography image. These pictures show our recent research.

Artificial Intelligence for IVUS Image

In clinic, physicians believe analyzing distribution and morphology of plaques can improve treatment strategies on cardiovascular diseases. Gray-scale intravascular ultrasound (IVUS) is a common technique for visualization of intravascular structures and plaque distribution. However, there are some existing limitations to automatic results, for instance, RF attenuation and shifting problems cause conflicts between VH-IVUS and physicians. This condition prompts our team to develop a new image-based approach to automatically characterize plaque components from IVUS images with artificial intelligence techniques.

The proposed method is shown in Fig.1(c). Fig. 1(a) shows image overlapping degree between ground truth and prediction with dice score. Fig. 1(b) is the visualized results, and each column means gray-scale IVUS image overlapping with ground truth, difference between ground truth and prediction of media, lumen and calcification. Yellow curves are borders of ground truth and red regions mean predicted output. For media and lumen regions, their dice scores are 0.931 and 0.947 which are better than analyzing software, and standard deviation are both below 0.05. Compared to commercial methods, our approach could provide better and precise information on characterizing plaque components.