Analysis Of Radial Arterial Ultrasonic Blood Flow Signal And Its Application In Medical Diagnosis

Biometric technologies grew rapidly during the last 20 years of the century, and medical biometrics, one of its branches in medical informatics, has received lots of attention from researchers around the world. In traditional Chinese pulse diagnosis, doctors put their fingers on human wrist to feel vascular pulsation, so as to collect the pathological and physiological information of a patinet's overall body condition, which is a naive means of medical biometrics. In contrast, although human pulse is influenced by the circulatory system and organs blood goes through, western researchers often focus on its application in diagnosing and preventing cardiovascular diseases, yet neglect that pulses or blood flow status reflect on the overall body condition of a person.So to make efficient integration of Chinese medicine's perspective of holism and western medicine's modern diagnostic methods should be a very promising topic. This thesis studied how to extract features on Doppler ultrasound blood flow signals of wrist radial arterial and make analysis of its reactions to certain pathological changes of human body, for the purpose of computer-aided diagnosis. At the same time we stepped out a small stepAnalysis of ultrasound blood flow signals was mainly based on the envelope of Doppler sonogram. So first we recovered the sonograms from sampled original data. Sonogram is a kind of approximate visualization of original data which is for ease of observation. Then we extracted the maximum frequency waveform, i.e. the sonogram's envelope. The envelope of a sonogram was taken as one-dimensional time series and it reflects on the information of blood flow velocity. Pathological changes of certain organs will influence the movement of blood, thereby influence the shape of envelope. Theories of traditional Chinese medicine insist that pulses of wrist radial arterial include the most pathological information in aid of diagnosis. So in this thesis, sonograms were recorded at wrist radial arterials of both hands. After that start points of periods of each envelope were determined and the work next was based on the results of envelope extraction and period determination.Biomedical signals are usually non-stationary. Amongst non-stationary signal processing technologies, wavelet transform is a newly developed signal analyzing technology which has already been widely used in the area of biomedical signal analysis. Through wavelet and wavelet package decomposition of sonogram envelope and taking wavelet and wavelet package powers as features for classification, we used a support vector classifier to perform the classification experiment on 3 groups of data, which were sonograms sampled from gastritis suffers, cholecystitis suffers and healthy persons. Experimental results showed that wavelet and wavelet package powers provide a wonderful discriminating capability between two groups of patients and a fairly good distinguishablity between groups of health persons and patients with a certain disease.The fifth chapter did bispectral estimation on sonogram envelopes of healthy persons and patients with gastritis and cholecystitis. Observation of bispectral magnitude maps and contour maps revealed that although magnitude provided not enough information for correct classification of different groups of data, however, there existed distinct differences in where the bispectral magnitude maps peak. The frequencies at which the bispetral magnitude maps of patients peak were obviously lower than that of health persons, and location information of magnitude peaks could be used for discriminating data of two different types. We got slides of different frequencies of f₁, divided each slide into segments with equal frequency band of f₂, and computed the"axially integrated bispectra"of each segment as features for classification. Experimental results showed that, this method could efficiently discriminate gastritis and cholecystitis sufferers from healthy persons, and the two groups of patients with different diseases could also be distinguished with a high accuracy.