Since there are several types of arrhythmia, the analysis must yield the basic shape of a patient's heartbeat waveform for submission to the neural network, where its precise type is determined
The Department of Cardiology at the Long Island Jewish Hospital specialises in the care of diseased heart patients.
The hospital accommodates people with heart cardiac diseases and irregularities.
Life-threatening arrhythmia, or irregular heartbeat, affects 300,000 people per year in the United States alone.
Arrhythmias vary from person to person, so a cardiologist needs as accurate a picture as possible of an individual's arrhythmia in order to treat it correctly.
Steven Evans heads a group at the centre that does just this kind of evaluation.
His team records a patient's heart's electrical activity, extracts and analyses the important part of the cardiac signal, then submits it to a neural network to ascertain what type of arrhythmia the patient has.
Since processing data quickly is essential when patients' lives are in danger, it is vital for this team to have a tool that can meet sudden demands in emergencies, yet is accessible enough for several researchers to be working with it all at once.
Evans's team uses the graphic display data processing software, Dadisp, to meet their data processing needs.
Evans says: "The use of Dadisp has dramatically decreased the time spent extracting and analysing data, markedly increasing research output.
Since more research means more patients helped, Dadisp is a vital part of Evans's team.
To record the signal from an arrhythmia patient's heart, Dr Evans inserts a small catheter with a wire tip into the heart; an operation so exacting that it has to be done under x-ray guidance.
The signal from the beating heart is recorded by an EP Lab System, a cardiac mapping system.
This device digitises and displays in real time 20 channels of data from the waveforms being picked up inside the heart, stores them to optical disk, and sends them to a chart recorder which records them on paper.
Only three or four of the 20 recorded channels are necessary, so these are put into a single binary file by an application supplied by Bard for this purpose and read directly into Dadisp, which is running on a 486-based personal computer. Dadisp is used to extract and analyse relevant information from the cardiac signal.
Since there are several types of arrhythmia, the analysis must yield the basic shape of a patient's heartbeat waveform for submission to the neural network, where its precise type is determined.
To begin the analysis, Dadisp pulls apart the binary file into separate channels, then gets the peaks of each individual heartbeat using a peak detection algorithm.
Each peak is used as a guide for a first-approximation extraction of the significant part of the waveform surrounding it.
The first derivative of the extracted data is taken, which permits a more accurate extraction of the original signal.
The amplitude of the waveform has to be normalised, a process easily accomplished with Dadisp, so that amplitude differences will not interfere with the neural network's assessment of the overall shape.
After this and several other steps, the waveform is concatenated with the rest of the extracted data to create the final data series.
A long segment of cardiac electrical activity is now ready for submission to the neural network for evaluation of the type of arrhythmia present.
Dr Evans says: "Dadisp is simplicity in itself to use".
He appreciates the power and flexibility of Dadisp's many commands and simple command structure, and says he likes the fact that a change in one command or input parameter can enable Dadisp to accommodate a wide range of input data.
Dadisp is supplied and supported in the UK and Ireland by Adept Scientific
