Project Design

We need to determine the nature and scope of our project. It is crucial to go into as much detail as possible to support a decision that makes optimum use of resources to achieve desired results. Everyone is more than welcome to contribute.

Mission
Who will benefit
Where and how would our device be used
Hardware
Software
Proof of Concept
Software Architecture
Education
Other projects
References

Mission

To develop an open source, low cost, and clinically functional electrocardiography solution.

The ECG should be of good quality, and safe for patients. This is not a hobby project!

Should the device be able to pass international certification?
Professor Robert A. Malkin in his EWH study has identified that standards probably play only a negative role (Malkin 2007). He says "...we can conclude that international standards are not creating an indigenous medical equipment manufacturing capability. In fact, imposing standards probably amounts to “kicking out the ladder”—preventing the emergence of local companies that might have produced low-cost devices, devices that work in the developing world, but do not meet the international standard." This said, it is still important that regardless of official compliance any device meets the following outcomes:
a) Full optoisolation - no chance of back current to patient or damage to equipment from a defibrillator or immersion.
b) Reliability - If not redundancy, simplicity will help ensure continuous operation under stressful conditions.
c) Operating/Maintenance endurance - The device must be able to endure cleaning procedures and hospital solvents, and require little maintenance. Maintenance failure is one of the key causes of equipment shortage in the developing world (Roberts 2000).
d) If an auto diagnosis routine is included, this will need to have statistically predicted sensitivity, specificity, positive and negative predictive values published. This presents a barrier for such software.

Who will benefit

Physicians from high income countries

In Europe, the number of hospitals in decreasing which means that general practitioners have to accommodate the demanding population. They require basic diagnostic devices like the ECG to carry out primary examination. Commercial solutions do not come cheap even to these physicians.

Physicians from outreach centers in high income countries

Geographically diverse countries like Australia have community outreach centers and medical stations separated by large distances. Indigenous communities in these areas have the highest incidence of heart disease in the general Australian population (Penm 2008). Many of these facilities suffer from doctor shortages and equipment failures (Milgate 1999). A cheap ECG device that can transmit trace information to city hospitals has potential. Usage would be dependent on compliance.

Physicians from middle income countries

As the countries are getting richer the burden of disease caused by infectious diseases and undernutrition decreases, and noncommunicable diseases such as heart disease, diabetes, and stroke cause the largest share of the total burden of disease (Stevens et al. 2008).
Similar commercial devices are very expensive and numerous physician cannot afford them.

Physicians from low income countries

Lifestyle disease presents a burden on developing as well as developed countries. It should be noted that medical equipment is often present but inactive due to poor maintenance and consumables shortages. (Roberts 2000)Cheap, fail proof medical technology developed with telemedicine in mind is promising for developing countries.

Industry

The ECG is of mandatory concern in the development of drugs to screen for the liability a drug may have on the electrophysiology of the heart. No drug goes on the market with out a thorough analysis of the drug-induced effects on the ECG. By a developing a open-source system (hardware and software) for the detection and analysis of the ECG smaller biotechnology companies will have a better platform to develop their medications with. It is unlikely that it will radically benefit Big Pharma because it is much better at absorbing these costs. Giving industry a viable alternative to the proprietary ECG systems available on the market will reduce the cost of doing such experiments which in turn could reduce the cost of developing new drugs.

Education of health scientists

In the training of new health scientists (Pharmacologists, Biochemists, Physiologists) companies subsidize equipment for the students to work on (AD Instruments). This incentive is to encourage students to be familiar with a given system they will demand it from their future employers and is a standard marketing tactic. By developing a open system you help to break the cycle to which classical business maintains long-term customers. Instead you may prompt scientists to re-ignite their connection with their tools - Einthoven whom coined the term ECG and invented the "PQRST" nomenclature in 1904 and Holter who built the first ambulatory device to measure ECGs are examples of cardiologists who built their own devices. It may also aid in helping scientists to better understand the systems they use because they are given access to the very source code that the software is written in. Ultimately, hopefully, by instilling the spirit of open-source development you encouraged learned scientists to contribute to ever improving the system.

Other projects?
Small local companies

Could such a device be produced locally, thus also helping the local economy. This could most definitely be archived in some low and all of the middle income countries like India for example.

Where and how would our device be used

Mostly in peri-urban and rural health care. To asses patients before arrival of the ambulance service and their referral to the hospital. In some rural and remote communities the open ECG will form a bridge between diagnosis and treatment, as medical evacuation is not always possible and can take time. Time pressures when treating a ST elevation MI, for example, necessitate fast diagnosis and treatment well inside the typical hours/days a flying doctor service takes, even in developed countries.

The usage of the openECG is inversely proportional to the wealth of the country. The wealthier the country, the less mainstream the use of the device. In poorer countries, the open ECG may well be used at tertiary medical referral centers. In wealthier countries, it may only be used at outreach missions or in remote communities.

The openECG should also be capable of monitoring, especially if alternative equipment is not available. Monitoring may only include a continuous lead II trace.

How long will the recording last?
The ECG should record 12 leads for a few minutes, for diagnostic purposes. If recording is PC based (not cached locally) then provisions should be made available for longer monitoring if required. The bitrate of a lead II continuous recording with digital waveform recording is roughly 100Kb/min using the FreeECG software (external link)

. Provision should be made in any software to allow continuous lead II monitoring if desired. This is unlikely to add design complexities, and is merely a software addition.

What parameters would be measured?
ST-segment is very important, as is QT interval, asystole, brady and tachycardia and VF/VT. As outlined earlier, it may not be possible to have software recognition of these complexes. However, it is relatively simple to automate rate warnings (brady or tachy) which would be convenient for monitoring.

Hardware

Main unit

Our ECG should be PC-based USB device. This would allow it to be connected to desktop computers, laptops, but mobile phones too.

What about the number of leads?
From a medical stand point 12-lead is the best option.
It is not beneficial to do a 12-lead ECG recording in an acute situation, it takes too much time?
Quite the opposite, it has been conclusively demonstrated that information obtained from the prehospital ECG reduces the time to hospital-based reperfusion treatment (Ferguson JD et al. 2003).
Analysis of the cost of providing this service to the community must include consideration of the demonstrated benefits of more rapid treatment of patients with myocardial infarction and the resulting time savings advantage shown to accompany the use of 12ECG programs (Garvey JL 2006).
There is perhaps a problem with education and training, but it has been shown that hospital-based paramedics and nurses can successfully be taught to evaluate (ie, sight read) a prehospital ECG for the presence of AMI with accuracy in rural areas (Foster DB et al. 1994).

What about the 5-lead approach?
There is good data to support 12 lead synthesis using orthogonal lead placement of five electrodes and the use of the Dower Transform. Philips's trade name for this is EASI (external link)

. However, the VCG (and transformation to 12 leads) can be faked in many ways. SVD of five or more spatially diverse leads can provide a good 12 lead reconstruction. For AMI three precordial leads are going to be needed and at least 2 other orthogonal leads. More may be required.

What about sampling frequency and quantization?
At least 10 bit assuming a full dynamic range.
If QT analysis is being considered, then 500 Hz to 1kHz is needed.

What about noise analysis and rejection?
Impedance checking, lead selection (particularly if Dower-like transform is used).

What about the power?
It would have to be either battery powered or use an isolation transformer plus defib protection in front of the op-amps on the front-end.

Electrodes

Disposable or reusable?
See the reusable electrode page here: Reusable electrode development
There is a mandate for reusable electrode usage, including demonstrated need in developing nation hospitals.

PC Hardware requirements

See PC Hardware

Software

What languages should the software be written in?
Since a GUI is probably essential (as you would want to have human training and some oversight), Qt (external link)

is a good choice, another might be Java. The underlying code might be written in ANSI C. To ensure a simple and OS-independent deployment, Qt should be considered, or perhaps Java Webstart? Even better, an interface that allowed many coding languages would be useful, so contributors could add modules in languages of their own choice. Java is a good choice because it would allow easy porting of software to cellphones. Java is also future-proofed as it the native operating language of many hard coded devices (PDA/mobile). Qt is a good choice because it is written in C++, which is very effective and portable.

Again about the leads?
From a software point of view the only difference is multiplexing and switching the 3 leads each 2.5 seconds. Alternatively, the ability to simultaneously acquire all leads is a feature of the best current ECGs. If this is possible, it should be pursued. Numerous cardiac events are best analyzed from multiple directions simultaneously.

In what format should the ECG data be stored?
Using the DICOM file format may allow the data to be viewed in PACS software. EDF (external link)

, or more flexibly, WFDB, would probably be better choices. Open source viewers exist for both these libraries.
The format chosen should ideally meet the following criteria:
a) Built from open source libraries.
b) East to read (using software).
c) Support de-identification methods.
d) Allow annotations.
d) Be compact - to facilitate rapid trasmission rates in telemedicine.
See this page for more details.

What about interpretation?
Automated ECG interpretation still cannot beat a qualified clinician on relatively short recordings (Hongo RH & Goldschlager N. 2006), so should we bother with it at this point? For now it may be more beneficial to have the output readily available for telemedicine (transmission and remote review). However, when skills are in short supply, or connections are out, local software could be useful. The software could also be adjusted to provide training.

Proof of Concept

This flow chart is a proof of concept for one potential interface method. This includes the use of an off-the-shelf USB to Serial converter, available for roughly five dollars ($5) (external link)

.
Simplicity is king, so it may be appropriate to remove the relay controlling power - a switch is good enough.
Similarly, for multiple lead acquisition I am unsure of whether it is possible to constantly feed 12 leads over RS232 or whether switching would need to occur in the device. If you are more familiar with serial protocols please update this.

You are welcome to change this image or add things. Image was created in microsoft visio - ECGdesign.vsd

Software Architecture

I am proposing the following software architecture. Please review it and give your comments.

DSP Module: this module is a central repository for implementations of various signal processing algorithms (e.g: filtering, downsapling, interpolating). All the other modules rely on it for their signal processing needs.
Data Acquisition Modules: these modules will read data from the hardware or files, filter the signal (if necessary) using the DSP module and then pushes the data to the cardiognosis and display modules.
- Interfacing modules: read ECG signals from the USB device. However, during development, programmers must be able to work even if they do not have access to the actual signal acquisition hardware. For this reason we will need to develop a simulator that will provide pre-recorded signals (e.g. MIT/BIH recordings) as if it were received from a USB device. The interfacing module must provide device control functionalities (e.g: on/off, baud rate control).
- File Reader: this module shall read signals that have been previously recorded to file.
Recorder: records ECG signals to file. The recorded files can be read back using the file reader module.
Cardiognosis Module: this is where the ECG signal will be analyzed and monitored. Heart beat and other parameters that need to be displayed to the user will also be calculated in this module.
Display Modules: these modules will display the actual graph and values calculated by the cardiognosis module. Before the graph is displayed, the digital trace might need to either downsample or interpolate the signal it received from the data acquisition modules.
Printing module: will enable the user to print out the output generated by display modules. In addition it should provide an option for printing grid lines if ECG papers are not available and the user wishes to print on a blank paper

Education

Education is extremely important. We should aim to include cardiologists, other physicians and paramedics in the process of writing open educational materials. Also, extensive instruction books should be available for the device. All of the material should be made available in many different languages.