Many medical algorithm development works are reported without implementation details. That makes it difficult to estimate the effort requires to transition research knowledge to commercial realization. In this paper, we will make an effort to trace the lineage of the open source codes, describe the modifications in sufficient detail to aid the readers in reproducing results and duplicating the prototype. While the sensor we built is not optimum for mass production, there will be sufficient technical specifications for anyone interested in such an endeavor.
The detection of the heart sounds S1 and S2 is accomplished with a beat finding technique developed for the music industry. The specific beat tracking technique is based on dynamic programming. In the first step of the detection algorithm, audio signal is converted to its onset strength envelope (ose). The ose is calculated as the sum of the difference between the spectra of the current and the previous waveform segments. The ose therefore represents the instantaneous overall change in spectral content (distribution of energy at different frequencies). To calculate the ose, a window of N data points is advanced in equal steps until the window reaches the end of the waveform. The number of data points N in each window corresponds to 1/8 s for the selected audio sampling frequency. The step is only half the size of the window so there is overlap between consecutive windows. The window is analyzed to calculate the spectral content or the energy contained in 20 frequency bins.
A comparison of wired and wireless amplitudes shows that the voltage of the wireless signal is lower but the signal-to-noise ratios (quality) are comparable.
Data collection starts first with strapping the microphone over the heart of the examinee, secondly the examiner putting on the headphones to monitor the recording and to ensure that the signal strength is sufficiently high but not too close to saturation level, and thirdly the examiner commanding the MATLAB program to record heart sounds and display the PCG signal. A frequently used record length of 50 s, recording 55 to 100 heartbeats, is sufficiently long to warrant that the timings of the first and second heart sounds are statistically significant for a relatively constant heart rate or when the subject is at rest.
Sometimes, records of 200 s or longer are collected to study the change of heart rate in the recovery phase after physical exercise. In those cases, the objective is to monitor the gradual decrease of heart rate in the recovery phase. In this proof- of-concept study, the PCG signal was recorded to show that useful physiological indicators can be acquired. The study is not intended to validate the tool’s clinical readiness. With the intended scope, the numbers of subjects (five) and samples
36
(26) are deemed sufficient. Since the objective is only to capture the timing of the S1 and S2 sequences and not to diagnose particular aspects of the hemodynamic response, auscultation placement is straightforward and doesn’t require cardiologist’s expertise. For our purpose, placing the stethoscope near the heart’s apex typically results in a strong signal to noise ratio which is the most important factor in capturing the heartbeat sequence timings. The stethoscope microphone is connected to the 3transmitter unit and the receiver is connected to the laptop to record heart sounds. A pair of headphones is also connected to another port in the laptop configured to monitor the audio. Ideally, the microphone only senses the heart sounds of the subject and not ambient noise. Thus, data collection is best in a quiet room, with the subject sitting completely still, and the chest strap adjusted so that the microphone is directly over the heart. However, the processing techniques we use are effective in alleviating the effects of extraneous noises.
When needed, the subject may wear the wireless microphone and jog on a treadmill while data is being collected. The data taker, listening through the headphones, can help with the adjustment of the microphone gain and placement of the sensor over the heart.
In a typical data collections of audio data are collected using the MATLAB audio recorder built-in function, at a rate 32.000 samples per second. The entire record consists of 1.600.000 values. Since the sampling rate is much higher that the highest frequency found in actual heart sounds, signal with frequency higher than 1000 Hz is filtered out. The beat tracking script made available at the LabROSA internet site was designed to extract a single dominant beat, not two beat sequences as in the case of heart sounds. We modified the codes to extract both heart sounds by running the algorithm in two passes. After the first pass, the signal that corresponds to the first detected sequence of heart sounds is removed and the pruned signal is processed again to detect the second sequence, as described in «Segmentation Techniques».
Using the timing relationship between the S1 and S2 sounds, we proceeded to identify S1. The S1 and S2 beats are subsequently paired up and the beat intervals (T11) and the systolic intervals (T12) are calculated. The beats which are not detected because of noise and their potentially unpaired beats are not analyzed. Note that the instantaneous heart rate can be estimated in real time by calculating the inverse of T11. Two additional diagnostic parameters, heart sound temporal width T1 and T2, are calculated directly from the Shannon energy envelope (see). Note that they are not derived from the ose. The heart sound is composed of several fre-
37
quencies, all measurable by the PCG and should be included in the see though not all are within the human audio spectrum. The see which is calculated from acoustic energy in all frequencies may be different from the humanly perceived heart sound. We would like to hypothesize that the see is an unbiased representation of the mechanical sound. Therefore, T1 and T2 extracted from the see envelope are representative of the mechanical sound made by the heart. The program displays the four diagnostic parameters and indicates the range of nominal values. The physiological parameters are useful for primary care physicians in screening referable patients and for specialists to infer preliminary diagnosis. It’s conceivable that the primary care physician may select to send forward the information generated by this system.
38
Тема 5. Реографический метод
Прочитайте текст, посвященный реографии. Переведите его и ответьте на следующие вопросы.
1.В чем заключается основной принцип работы реографа?
2.Что вы знаете об использовании реографа?
3.В чем преимущества и недостатки его использования?
4.Подумайте, как можно было бы усовершенствовать прибор?
5.Что подразумевается под реографическим методом исследования?
Rheography
Rheography is a method used to study the filling of a part of the body with blood by graphically recording the fluctuations in the resistance of that part of the body. It is used in physiology and medicine.
Rheography is based on the fact that when an alternating current of sonic or ultrasonic frequency (16 – 300 kilohertz) passes through a part of the body, the organic fluids (chiefly blood in the large blood vessels) act as conductors. This makes it possible to determine the condition of the blood circulation in a particular region of the body or in an organ, for example, in an extremity or in the brain, heart, liver, or lungs.
Among all the structures of the body blood has the highest electrical conductivity. This means that during systolic contraction of the heart when the blood flows into nearby bodies electrical conductivity of these parts of the body will be high and the moment of cardiac muscle relaxation (diastole), opposite – low. Based on the testimony rheograph output curve of pulse oscillation, called rheogram.
Rheogram looks sinusoid with a steeper rise characterizing arterial blood flow and smooth descent, which, in turn, is a reflection of venous flow. To thoroughly analyze the state of blood flow during reography necessary to remove a lot of curves. An experienced diagnostician will pay attention to the regularity of the curve and its shape, presence and amount of additional curves in a downward phase. In addition to the external characteristics of the curves, the doctor decides to several mathematical problems: special formulas calculated rheographic index, for which a certain interval, when going beyond which one can judge the presence of pathology, and a few other parameters (amplitude-frequency rate, the rate of venous outflow, pulse wave propagation time).
39
Rheography – non-invasive method that is harmless to the body. The skin and tissues are not damaged, as transmitted through them, an electric current is so small magnitude and frequency that is just not able to do any significant damage.
Besides this method has a high sensitivity.
Rheogram analysis and the Stroke Volume
During the isometric phase, the beginning of which coincides with the Q wave of the electrocardiogram, at the time of displacement of the valvular plane of the heart, there occurs a drop in the curve corresponding to sucking in of blood into the heart from the great veins. At the beginning of the ejection period, that is with the opening of the semilunar valves, a steep ascent appears, which breaks off sharply from the rest. The steep ascent corresponds to the rapid filling of the arteries near the heart. In the further course of this phase of systole there occurs a preponderance of outflow of blood into the periphery over the further expulsion of blood from the heart. Accordingly the curve sinks after the first third of the ejection period somewhat more sluggishly. The second heart sound indicates the end of systole. With the valves all closed, the congestion of blood awaiting entry into the ventricles produces rheocardiograpically a new ascent of the curve. The continuing diastolic ascent of the curve falls in the rapid filling phase of the ventricles and is explicable by increasing filling of the venae cave in this period. This phenomenon may be attributed to increased pressure on vein walls as a result of increased circulation rate.
AD5934
The AD5934 output voltage and measurement frequency are fully programmable and the communication is provided by an I2C interface. An application of the AD5934 in biological diagnostics research has been reported by many authors. The AD5934 was used in the blood coagulation detection, biosensor applications, general bioimpedance measurements, and was also used in technical object monitoring.
Designers of impedance meters dedicated for biological objects should not forget about the requirements that have to be met. Besides, a measurement system should offer a wide frequency range, high degree of integration, portability and accuracy. AD5934 absolute corresponds to this condition. The designed, fabricated and tested device enables measurements in the frequency range from 1 Hz to 100 kHz and in the range of impedance from 1 Ω to 10 MΩ.
Application of a method of rheography on a chip of AD5934 is the blood resistance research for detection of a virus. When a known strain of a virus is added
40