A. Specific Aims
The second leading cause of death in the United States is heart disease. It still remains that woman are affected more than men. In 2003, despite the fact over 700,000 Americans died of this disease, only 148,000 were diagnosed that year. Thus, 552,000 Americans were either not diagnosed or misdiagnosed (1).
Many individuals remain unaware of the symptoms of heart attack or dismiss possible symptoms as being unrelated or not important enough to visit a doctor. Such was the case in 2002 with11% of adults that were told by their physicians that they had previously had heart disease without knowing it(2).
This Phase I proposal presents a solution to this public health problem via a simple, over-the-counter tool that would provide the user with an analysis of risk of heart disease.
The Phase I goals of this plan are:
1. Design a device that will be used to continuously record an individual’s electrocardiogram (ECG) over a period long enough to accurately diagnose heart-related issues. This Phase I device will include a differential amplifier, a microcontroller, and a USB compatible storage device.
2. Develop a user-friendly computer interface, which will analyze the data retrieved from the ECG collection device. The program will use the results of accepted studies on ECG waveforms and heart disease in order to determine level of risk.
3. Fabricate a prototype of the ECG collection device.
4. Perform testing with patients not diagnosed with heart disease to assess basic functionality of the system. The result of this goal will aid in future adjustmentsto the device to achieve a more accurate analysis.
B. Significance
Heart disease is a major concern in America. There is a need for people to understand their risk for this disease and other heart related problems so that the necessary steps can be taken to prevent serious complications before they start. In times of increasing costs of healthcare and increasing risk of heart complications, cheap, at-home and portable heart diagnostic tools are becoming more and more critical in the early stages of personal healthcare.
Our device is for the general public mainly towards adults over the age of forty. There are four main factors that will help determine whether you are likely to develop heart disease. These are family history, hypertension, obesity, and hypertension. Based on the 2002 Census, 119,386,252 people (42% of the population) belonged in the appropriate age group. Hypertension, or high blood pressure, affects approximately 28 million adults. Obesity impacts the lives of 41 million people. 20 million adults over the age of 40 are smokers.
This device would potentially detect and diagnose several disorders. Arrhythmiasare disorders of the regular rhythmic beating of the heart and are fairly common. While they can occur in a healthy individual and not cause a problem, they can also indicate a serious health problem and lead to disease, stroke, or death. There are two types of arrhythmia. One is ventricular fibrillation which is an often fatal form of arrhythmia characterized by rapid, irregular fibrillar twitching of the ventricles of the heart in place of normal contractions, resulting in a loss of pulse. The other is atrial fibrillation, which is fibrillation in which the normal rhythmical contractions of the cardiac atria are replaced by rapid irregular twitchings of the muscular wall that cause the ventricles to respond irregularly.
Heart murmurs are most oftencaused by defective heart valves. A stenotic heart valve has a smaller-than-normal opening and can't open completely. A valve may also be unable to close completely. This leads to regurgitation, which is blood leaking backward through the valve when it should be closed.
People with sinus tachycardia have an abnormally rapid sinus rhythm; specifically a rhythm at a rate greater than 100 beats per minute. Whereas those patients with sinus bradycardia have an abnormally slow sinus rhythm; a rhythm at a ratelower than 60 beats per minute.
Finally, the monitor should detect heart block. This is a condition in which faulty transmission of the impulses that control the heartbeat results in a lack of coordination in the contraction of the atria and ventricles of the heart.
While there are heart monitors on the market, these devices are used almost exclusively when directed by a physician. They also use wires to collect data. This device would be unique in two ways. One, it would be designed to be purchased over the counter. This way, anyone could assess their personal risk of heart problems. Second, this system would be wireless. This will minimize the chances of the device interfering with the person’s normal activities.
There is currently a system known as Holter monitor, in which the doctor gives the patient a monitor to take home. In this device, the leads connect to a small recorder. One of the major problems of this device is that it costs upwards of $6000 dollars. However, this device indicates that it is feasible for the investigators to design and market the Wireless Heart Monitor.
C. Relevant Experience
James Cook has extensive experience in computer programming—particularly applicable to this study are LabVIEW and Matlab. He also has experience in analog and digital circuit design. He has worked with human subjects in biomechanics studies of keyboarding and gait under Drs. Rakié Cham, Mark Redfern, and Nancy Baker at the University of Pittsburgh, and is currently developing a device for the early detection of diabetic retinopathy under Dr. Michael Gorin.
Carmen Hayes is a bioengineering student at the University of Pittsburgh. She has carried out research in angiography and musculoskeletal imaging. At the University of Wisconsin-Madison, she conducted research comparing two methods of imaging the carotid artery using the MRI.
Joe Konwinski has worked for the department of neurosurgery for the past year developing a miniature implantable EEG for wireless signal transmission. In the past, he has carried out research in The University of Pittsburgh’s Department of Orthopaedic Surgery and done a clinical rotation with The Bradford Regional Medical Center’s Department of Microbiology.
D. Experimental Design & Methods
1. Design a wireless device that will be used to continually record an individual’s electrocardiogram (ECG) over a 24-hour period.
This Phase I device will include electrodes, a differential amplifier, a microcontroller, power supply, and USB compatible flash memory will be integrated into the device. Size constraints will be determined by further testing, but from previous work with miniature amplifiers, power supplies, and A/D converters it is expected that the complete device will be no bigger than 4.0 in. x 2.0 in. x 0.5 in.
Electrode Placement and Viable ECG Recording
Our main task in manipulating a basic ECG circuit (fig. 1) was to develop a filtering system that would allow us to bring positive, negative, and grounded electrodes of a three-lead ECG closer together without losing viable ECG recording(fig.2). Because the main constraints of our phase 1 study are product size and power consumption, we have limited ourselves to interpretation of cardiac rhythm from just one recording. This differs severely from a clinical standard 12-lead system which produces multiple ECG recordings and allows clinicians to better see malfunction in all parts of the heart.
Fig. 1. Basic ECG circuit (Differential Amplifier) that is used as a base-line for initial testing of electrode placement. (Taken and adapted from reference 1.)
The risk of bringing electrodes closer together is that it becomes more difficult to remove low-frequency noise (<.03 Hz) caused by respiration as well as wide-bandwidth frequency noise caused by muscles contraction (1-5000Hz). Sixty-Hertz interference from power sources also can cause problems in filtration of the ECG signal (4).
Fig. 2. Evolution of standard one-lead ECG electrode placement.
Einthoven’s Triangular method allows us to place electrodes in a triangle around the heart and still be able to detect and amplify heart-induced differential signals through the skin and fluid resistances. Initial analysis of electrode placement involves bringing electrodes along their angular paths in very short increments and detecting what frequency noise is causing interferences in the signal. After detection of noise, manipulation of both the high and low pass filters in the ECG circuit was carried out in an effort to remove as much interference as possible, and bring electrodes as close as possible. Figure 3 below shows the final circuit manipulation used to filter out noise. It consists of an instrumentational amplifier which filters out DC offset due to differences in both skin and electrode resistivity. The high-pass filter implemented has a low cut-off of 1 Hz to filter to verify that DC has been eliminated and a stable baseline voltage can be recorded.
Fig. 3. Final Phase I circuit design
Power consumption also becomes an issue in portable solutions. The Instrumentational amplifier we are using is TI®’s INA332 which consumes 450uA of current. The Operational amplifier is TI®’s OPA2336 which consumes only 20uA of current. These two chips can be powered for over 40 hours with normal watch batteries. Using the Small-Outline-Package chips size of the ECG amplifier becomes less of a problem (fig. 4).
Fig. 4. Package dimensions for Instrumentational and Operational amplifiers
Portable Memory Control:
The most complicated task in development of this novel portable ECG system is the storage of the ECG signal onto a removable memory device. Past systems have shown storage of an ECG signal onto bulky tape recording devices(5). With the advent of modern computer technologies non-bulky memory of time dependent physiological signals has yet to be “ideally” developed. Unfortunately we are still in the early stages of developing this tool.The process starts by converting the ECG signal from analog to digital form using Nyquist criterion constraints. Microchip’s® PIC18F2455 which was developed in the past 5 months is a useful microcontroller for our instrument. It has a 10 bit A/D converter which we hope will give the retrieved data enough clarity to distinguish peak-to-peak voltages. It also can transfer data through USB port at a rate of 2 MHz which is way faster than we need. The digital signal is then written to a flash memory in ASCII format and can be interpreted by a home PC as a series of numerical voltages.
Fig. 5 Schematic of Portable ECG system from the ECG amplifier to the PC interface.
2. Develop a user-friendly computer program, which will analyze the data retrieved from the ECG collection device.
A computer interface for the diagnosis of heart disease was developed using National Instruments (NI) LabView 7.1. The analysis system was broken down into four subdivisions: ECG Import; Signal Conditioning; ECG Analysis; and User Feedback. These subdivision of the program will be described in detail below.
ECG Import
The first step of the program takes the data collected from the patient and stores it to be analyzed. Built-in data acquisition (DAQ) software as well as NI DAQ PCI card and connector block was used to acquire the signals from the ECG amplifier (Specific Aim 1) at a frequency of 240 Hz and store them as ASCII data. An example signal acquired using this procedure is shown in (Fig. 6).
Fig.6, Imported ECG signal
Signal Conditioning
This step of the program imports the data from the saved ASCII file and attempts to condition the data for further analysis. A moving-median filter with a bandwidth of 240 frames (1 second) is used to calculate the baseline of the ECG signal (Fig.7). The baseline is then subtracted from the original signal, which moves it to a new baseline of zero. Areas of muscle contraction drown out the signal from the heart (Fig. 8), so they must be removed from the data. A threshold detector is run, which detects points in the signal that have a voltage greater than 0.4 V or less than 0.4 V (recall: ECG amplifer has an amplification level of 500). Then the largest section of continuous data between areas of muscle contraction is located to be used in the analysis step, since the larger section would improve the accuracy of the system. If the largest gap is less than twelve seconds (standard time-period for cardiologists) then the user is prompted to recollect data and try to relax his or her body. Finally, a second-order Butterworth lowpass (fl= 15Hz) filter is used to remove high-frequency noise from the signal (Fig. 9). The continuous section of data and its filtered version are used in the next step of the program.
Fig. 7, Unfiltered ECG signal (white line) with moving-median
baseline superimposed (red line).
Fig.8, Area of ECG signal showing muscle contraction.
Fig. 9, Final ECG signal after baseline shift, removal
of areas of muscle contraction, and filtering.
ECG Analysis
This step of the program takes the conditioned signal and attempts to locate peaks and valleys and label them according to the standard ECG waveform (Fig. 10).
Fig. 10. Anatomy of typical ECG waveform and normal physiological parameter times.
A peak-detection algorithm fits a quadratic polynomial to every three consecutive frames of data to determine the existence of a peak or valley within the set. The algorithm locates the peaks along with the amplitudes, first derivatives, second derivatives, and third derivatives of the signal at their location (Fig. 11). K-means cluster analysis is used to group the peaks based on their relative locations to each other in reference to amplitude, second derivative, and third derivative (Fig.12). Heuristics based on derivatives, amplitudes, and relative locations of the clusters would be used to give each cluster a label based on the standard ECG waveform (note: the ECG signal of at least 40 normal patients must be analyzed before these heuristics are developed and refined).
Fig. 11, Example of peak-detection showing the second derivative
of the peaks verses amplitude of the peaks.
Fig.Fig 12, Example of k-means cluster analysis on the result of the peak-detection algorithm. Red points indicate cluster centroids.
User Feedback
This, the final step of the program, aims to use the analysis data to determine risk for heart disease and provide feedback to the user. The peak-to-peak timing and slopes of important intervals are calculated. Programmatic statistical hypothesis testing would then be used to evaluate significant differences between the calculated periods and slopes and those of the normal population (Fig.13). This would result in a p-value, which would be translated into a percent-risk for each disease. A total risk percentage would be calculated by weighing each disease by a severity index and taking the sum of the risks. Finally the user would be provided with feedback on each disease; his or her risk; the causes, results, and usual treatments of each disease; and a prompt to see his or her physician for a more accurate evaluation.
Fig.13, Hypothetical distribution of the PR interval of the population of normal individuals. The red section indicates risk for first-degree heart block
3. Fabricate the device in sufficient quantity to support Phase I testing.
Phase 1 construction required the development of the breadboard model of the ECG amplifier (fig. 14) as well as further development of the miniature ECG shown in figure’s 15 and 16 below.
Fig. 14 Breadboard model of ECG amplifier
Fig. 15 Development of miniature PCB
Fig. 16. Development of Miniature ECG amplifier
Later construction will incoropate this amplifier into a complete package with a microcontroller, power supply, and electrodes (Fig. 17).
Fig. 5. A schematic example of the final construction of the portable ECG.
4. Perform actual ECG testing to assess basic functionality.
To test the capability of this system, three healthy patient ECG waveforms weresent directly from the ECG amplifier into the DAQ and interpreted by Labview®. Phase 1 goals were only to assess the capabilities of the computer model to detect peak-to-peak signal times. We were able to successfully detect the signal peaks of these healthy individuals with reasonable accuracy. Unfortunately time constraints did not allow for a statistical measure of success. Later work will lead to the statistical measure of success of the computer to detect peak-to-peak times as well as its ability to work on non-healthy patients. Researchers will manually assess the peak-peak time durations and voltages and compare them with Labview’s © results.
E. Future Plans
The problems in the flash memory system still need to be worked out. Once this is accomplished, we can move onto the next step of making a working prototype. Also, while we have gotten the computer program to detect the peaks, our next job is to work the code for the assessment of the disorders. Finally, and perhaps most important, the user interface will have to be created with complete instructions for use of the device and interpretation of results.
F. References
- NationalCenter for Health Statistics. Fast Stats: Heart Disease. 2004.
- U.S. Department of Heath and Human Services. Vital and Heath Statistics. Series 10,Number 222. 2002.
- MEDICAL and HEALTH RELATED SCHEMATICS
- Filtering of ECG and Other electrophysiological signals…Alango solutions for sound science…
- MEDLINE PLUS MEDICAL ENCYCLOPEDIA…..Holter Monitor…