Abstract
Purpose::
to develop a computer algorithm for identifying intraocular pressure (IOP) pulses so that IOP and pulsatile ocular blood flow (POBF) analyses can be performed under computer control, simultaneously with the IOP measurement process.
Methods::
The computational tools are signal processing, pulse recognition, pulse comparison, and statistical averaging. Waveforms from a database of 1600 de-identified IOP measurements taken with a pneumatic tonometer were used to develop and test the algorithm. The IOP measurements were performed in accord with the Declaration of Helsinki.
Results::
The IOP pulsates in synchrony with the heart beat. In systole, blood is driven into the eye and the IOP rises and then falls as blood drains out of the eye during diastole. Sampling the IOP at 100-200 Hz over a 5-10 second period produces a pulsatile waveform containing a train of IOP pulses. Not all pulses are useable for analysis due to unintended eye movements or other distractions. As the IOP is measured, the algorithm identifies individual pulses, conditions them to remove inherent high frequency noise, and compiles their pulse properties (average values, pulse amplitudes, systolic and diastolic times, and waveform curvatures) for pulse-to-pulse comparisons. When pulse properties between pulses are found to match within tolerance limits, individual matching pulse properties are replaced by their mean values. As each additional pulse is acquired, its pulse properties are compared with the previous pulses or mean values of matching pulses and new mean values are calculated if there is a match. When five not necessarily consecutive matches are found, the algorithm stops the process, omits the unusable pulses, and delivers the mean and standard deviation results based on the five acceptable pulses. From the known pressure-volume relation for the eye, the IOP pulses are transformed into pulse volumes and POBF using published procedures. The algorithm was validated by comparing the automatically selected pulses and calculated results with manual selection of pulses and corresponding IOP and POBF values, using the 1600 waveform database.
Conclusions::
Pulse recognition and analysis are accomplished in real time automatically by the computer algorithm, thus enabling the calculation of IOP and POBF from the properties of five IOP pulses. Since the matching process does not involve a normal set of pulse properties, the five pulses only match one another and therefore are specifically characteristic of the patient being measured.
Keywords: computational modeling • intraocular pressure • blood supply