Extracting Knowledge from an Image: An Example of VAD Stroke Volume Calculation

Abstract:

This paper presents the method of calculation of the blood stroke volume of Ventricular Assist Device (VAD). The method was developed for a pneumatic extracorporeal myocardial support pump, which was equipped with a miniature camera during the research. The aim of the publication is to show the method of “extraction of knowledge” from a 2D image showing the view from above of a multilayer flaccid diaphragm. The knowledge is being “extracted” during image processing and analysis, as well as during 3D modeling of the diaphragm shape. The result of the calculation is a three-dimensional view of the diaphragm and the value of the VAD stroke volume. The developed technique calculates the three-dimensional shape of the diaphragm and the blood stroke volume on the basis of only one image. This enables the stroke volume to be measured at any diaphragm position and at any time, and the frequency of the results is not less than 30 Hz. Basically it depends only on the time of image acquisition of a camera used. Experiments, during which the developed method was tested, have shown that image processing and analysis allows to “extract knowledge”, from which it is possible to obtain a hitherto unknown stroke volume with an average measurement error not exceeding ±2 ml.

nsdlogo2016