Research And Application Of New Method For Motion Tracking Based On The Ultrasonic RF Signals

Cardiovascular disease is an important threat to our health.It is necessary to monitor and evaluate the mechanical properties of arteries that are directly related to cardiovascular disease in the early stages of the disease.Ultrasound vascular elastography is widely used in the evaluation of mechanical properties of arterial vessels in the reason of its non-invasive,time and spatial resolution.The primary and key work of ultrasound vascular elastography is to achieve accurate motion tracking of vessel wall motion.The ultrasonic RF signal contains the amplitude and phase information of the ultrasonic signal.In the ultrasound elastography study,the speckle tracking method based on the ultrasonic RF signal is usually used to realize the motion tracking of the tissue.In the actual ultrasonic scanning,the discontinuous displacement field often appears in the RF signal,which is the challenge of the ultrasonic speckle tracking method.Therefore,it is necessary to propose a speckle tracking method with better robustness and accuracy.On the other hand,the traditional speckle tracking method can only process and analysis the current data independently,and it can not fully exploit and utilize the deep features and law of the dataset.Deep learning simulates human neural network to characterize and learn massive data.It is a data processing and analysis method for mining data abstract features.In order to achieve more accurate motion tracking and realize the full utilization of the ultrasonic RF data,and promote the ultrasonic elastography technology to better serve the clinic,this paper: 1)proposed a new two-dimensional speckle tracking method,using the dual correction mechanism that can make full use of the displacement estimation information of the seed point and its neighbor points,which ensures that the result of each seed point is the optimal estimation result of the displacement estimation of itself and its neighbors,thus improves the robustness and accuracy of the speckle tracking method.At the same time,we use the sum-table scheme to avoide the redundant judgment of the similarity in the two-dimensional normalized cross-correlation calculation,thus significantly improving the computational efficiency without sacrificing the performance of the displacement estimation.A finite element simulation model of COMSOL is used to verify the proposed motion tracking method.The result showed that the speckle tracking method proposed in this paper can achieve better displacement tracking effect.2)On the basis of the pulse wave imaging work completed by our group,the geometric correction of the pulse wave velocity estimation error caused by the irregularity of the human carotid artery is geometrically corrected,and a more accurate estimation of the mechanical properties of the carotid artery is realized.The proposed two-dimensional motion tracking method is then combined with the improved pulse wave imaging method and applied to the evaluation of carotid mechanical properties in clinical atherosclerotic patients and normal subjects.3)We proposed a new method of singleobjective displacement tracking based on deep learning that combined the Siamese network and DenseNet.Based on FieldII simulation toolkit,we designed a single vessel wall motion model.The proposed neural network was trained and tested using the simulation dataset and the clinical dataset respectively.The result showed that,compared with the single target block matching method,the motion tracking method based on deep learning that proposed in this paper can effectively mine the characteristics and laws of data,thus achieve more efficient and accurate motion tracking.