Ultrasonic Embolus Signal Detection And Classification

Emboli in the cerebral artery may block the blood vessel and thus leads to cerebrovascular disease such as the ischemic stroke. Meanwhile, the generation mechanism and the clinical treatment for the solid emboli and the gas emboli are different. So it is valuable to detect and classify the embolic signals in clinic. With the fast development of the medical Ultrasound technique, Transcranial Doppler (TCD) ultrasound equipment is now widely used to detect emboli non-destructively. However, when the size of the emboli is little or the sample volume is small, the power increased by the emboli may be rather limited. It is necessary to apply advanced signal processing methods to analyze emboli Doppler signals. Aiming at limitations of traditional emboli detection and classification methods, this dissertation aims to give some improvements in both the emboli detection area and the emboli classification area.In the emboli detection area, an innovative method is firstly proposed to detect emboli by combining the ultrasound color flow imaging and the ultrasound power imaging. This method aims to implement the diversification of the emboli detection methods and monitor the trajectory of the emboli on-line. Then, in order to solve the conflict between the size of the sample volume and the emboli detection probability in the TCD, an optimal sample volume selection method is proposed to achieve an ideal balance between the emboli-blood ratio (EBR) and the emboli detection probability.In the emboli classification area, the simulation model of the gas emboli will be established by analyzing the features and reasons of the gas emboli Doppler ultrasound signals. It is useful for the further classification of the solid emboli and gas emboli. Then, with application of the adaptive Gaussian-Chirplet expansion, ak which represents the distribution of velocity of the emboli in the vessel, βk which represents the frequency modulation of the Doppler ultrasound signals and a which represents the frequency change of the Doppler ultrasound signals are extracted. Finally, the Support vectors machine (SVM) is used to classify the gas emboli and solid emboli Doppler signals.Relevant studies and comparisons are performed for computer simulated ultrasound signals and clinical collected ultrasound signals. Experimental results indicate the following facts.1. The ultrasound flow power imaging can show emboli clearly, which may be useful to detect emboli accurately.2. The better sample volume may be selected to achieve an ideal balance between the EBR and emboli detection probability.3. It is shown that the features of the simulated gas emboli signals match with those of the gas emboli signals collected from clinic. It demonstrates that the simulation method of the gas emboli is reasonable, and can be useful for the further classification as the simulation sources.4. The classifications are performed for80cases of computer simulated Doppler ultrasound signals and111cases of clinical collected Doppler ultrasound signals. After extracting the features (ak,βk, a) of these signals by the adaptive Gaussian-Chirplet expansion, the SVM is used to classify the gas and solid emboli with these features as inputs. Experimental results indicate that these features can achieve a better emboli classification performance. Therefore, this method may be useful in the emboli classification in clinic.