الفهرس 
INTRODUCTIONOnly 14 pages are availabe for public view
 |  | from | 80 |  |  |
| | | from | 80 | | |
Abstract
Blood pressure is important vital signs of cardiovascular of the human health. When the measurements of blood pressure are accurate, these accurate measurements are necessary in the diagnosis and hypertension management and also related risk for dependent blood pressure. The auscultatory method is done via using mercury sphygmomanometer that is considered as the most accurate measurement technique for non-invasive blood pressure (NIBP), this is complex and only used for clinical evaluation. In these days, devices of blood pressure measurement that are digital self-monitoring are popular and spread in the market and also can be used at home. A lot of these devices are dependent on oscillometric method, such that it needs less occupational performer and is less sensitive to the external noise. The techniques of blood pressure measurements are dependent on measuring the pressure of the cuff and on palpating the variety of the pulse amplitude. These measurements are affected by the movement of patient sensitively. The unexpected movements that are slightest may recompense the reading of the meter of the blood pressure by a huge amount or recognize the reading totally meaningless. In this thesis, Windkessel models (two, three and four) were
applied for healthy subjects whose blood pressure vary between 120mmHg/80mmHg and unhealthy subjects whose blood pressure don’t vary between 120mmHg/80mmHg. We decrease the influence of noise from blood pressure measurements for unhealthy subjects whose arterial blood pressure (ABP) don’t vary between 120mmHg/80mmHg via stratifying a technique (comb filtering). The ABP waveform has been digitized and using digital signal processing techniques to process the signal waveform that is noisy (for unhealthy subjects). The noise spectrum is not always defined well, since it can contain different components frequencies depending on the movement type. The ABP signal waveform is more or less a rhythmic signal waveform. That translates to periodicity in the frequency time domain. A digital filter is designed that have an advantage of the periodic nature of the blood pressure signal waveform for unhealthy subjects. The filter is looked like a comb with periodic peaks around the signal waveform frequency components.
Fast artificial neural network (FANN) is significant tool for pattern classification computational that have been the matter of research that is renewed. FANN uses several learning algorithms and formats are being used in clinical applications, industrial, and academic research FANN were used in past studies to evaluate the measurement of the blood pressure. In this thesis, FANN designs have been used for determining mean square of the error (MSE) and standard deviation of the error (SD) for blood pressure (BP) measurements between inputs and outputs using back propagation training algorithm, where the input values in the neural network are the values estimated from Windkessel models (two, three and four). The result showed that the FANN was more accurate method. Also large databases are needed for this method. The over training and under training affect the accuracy of the measurement.