Date of Award

Fall 1996

Document Type


Degree Name

Doctor of Philosophy in Electrical Engineering - (Ph.D.)


Electrical and Computer Engineering

First Advisor

Stanley S. Reisman

Second Advisor

Nirwan Ansari

Third Advisor

John D. Carpinelli

Fourth Advisor

Peter Engler

Fifth Advisor

H. Michael Lacker


Electronic control systems for counterpulsation Cardiac Assist Devices (CADs) are an essential part of cardiac assistance. Synchronization of the counterpulsation CAD controller with the cardiac cycle is critical to the efficacy of the CAD. The robustness of counterpulsation CAD controllers varies with the ability of the CAD controller to properly trigger on aortic pressure (Pa) and electrocardiogram (ECG) signals for sinusoid rhythms, non-sinusoid rhythms and non-ideal signals resulting from surgical intervention. An analog-to-digital converter and digital-to-analog converter based CAD controller development platform was devised on a 33Mhz PC-AT.

Counterpulsation Pa systolic rise and dicrotic notch detectors were demonstrated with a 15cc pediatric Intraaortic Balloon (IAB) and 50cc Extraaortic Counterpulsation Device (EACD) CADs using mongrel canine experimental models in which biological variation due to changing heart rate and arrhythmia as well as surgical interference due to mechanical ventilation, electrocautery, signal attenuation and random noise was present. The robust Pa triggering algorithm was based on a derivative comparator "riding clipper" algorithm for the Pa-based controller.

In order to empirically determine the robustness of the Pa triggering algorithms, a simulation platform, Pa trace model, and Pa trace artifact and physiological variation models were devised. Each set of simulation experiments utilized a different Pa trace artifact or physiological variation model to determine the capability of the Pa trigger algorithm to withstand the effects of the Pa detection impediments while maintaining 100% accuracy of the dicrotic notch detection. Multiple simulation experiments were conducted in which the same nominally adjusted interference was increased to benchmark the immunity threshold of the dicrotic notch detector. Biological variation and deviations in Pa artifacts due to clinical conditions experienced in cardiothoracic surgery were investigated. Pa triggering was unhindered by biological variation of a Pa trace with a 3 mmHg dicrotic notch deflection along with a Pa trace with no dicrotic notch deflection present. Pa triggering was unhindered by heart rate variability ranging from 60 to 80 bpm due to respiration. Pa triggering was unhindered by clinical conditions including 40 mmHg changes in the Pa baseline modeling mechanical ventilation, aortic trace attenuation modeling variations in pressure transducer positioning and blood coagulation on the pressure catheter tip ranging from 100% to 200% of the Pa trace amplitude every four seconds, uniformly distributed noise with a mean of 0.5mmHg and standard deviation of 0.289mmHg and Gaussian distributed noise with a zero mean and standard deviation of 0.6nunHg. The results of the simulation experiments performed quantified the robustness of the Pa detection algorithm.

Development of a fault tolerant counterpulsation CAD control system required the development of a robust ECG triggering algorithm to operate in tandem with the Pa triggering algorithm. An ECG detector was developed to provide robust control for a range of ECG traces due to biological variation and signal interference. The ECG R-wave detection algorithm is based on a modified version of the Washington University QRS-complex DD/1 algorithm (Detection and Delineation 1) which uses the associated AZTEC (Amplitude Zero Threshold Epic Coding) preprocessing algorithm and provides accurate ECG-based CAD control R-wave detection for 96.56% of the R-waves stored within the MIT/BIH ECG Arrhythmia database with a maximum detection delay of 8 milliseconds.

Further IAB experiments performed with mongrel canine experimental models demonstrated that the systolic time interval to heart rate relationship existing in humans (essential to human patient CAD control inflation prediction) is not prevalent in canine mongrels particularly when treated with beta-blockers.

In order to execute both Pa and ECG C software detection algorithms for a fault tolerant counterpulsation CAD controller, investigation into the communications throughput of a quad‑transputer board was performed. Development of streamlined communication primitives led to a communication processor utilization of 8.3%, deemed efficient enough for fault tolerant multiprocessor CAD control implementation.