Weaning patients from ventilator dependency is a major problem in pulmonary medicine, with large human and economic cost. At least some patients in the so-called "difficult to wean"' category probably suffer from one or another form of poorly understood respiratory center dysfunction. Severe illness such as shock, sepsis, etc., as well as a wide spectrum of functional and metabolic disturbances, can cause profound but usually temporary dysfunction.
Many patients who fail to wean are known to have excessive neurological drive from the respiratory center, as determined from the P 0. 1 second ("P 100") test. Jubran and Tobin recently reported that patients who fail a weaning trial immediately display a rapid and shallow breathing pattern on removal from ventilatory support, while those patients who succeed in weaning display this typical exhaustion breathing pattern only as they become fatigued (See: "Pathophysiologic Basis of Acute Respiratory Distress in Patients Who Fail a Trial of Weaning from Mechanical Ventilation;" Jubran, Amal, and Tobin, Martin J.; American Review of Respiratory and Critical Care Medicine 1997; v 155 pp906-915).
This raises the possibility of strong but corrupted neurologic driving signals from the respiratory center sending out poorly coordinated and/or otherwise improper breathing pattern signals.
Therefore, if proper breathing patterns were visually provided to these patients as a biofeedback model breathing pattern they presumably should be able to follow same and maintain ventilation and therefore accelerate the weaning process. And during this time the respiratory center presumably would also be undergoing a training or reconditioning process to become more functional, thus expediting complete removal from ventilatory support.
Holliday and Hyers using relaxation techniques and a simple visual biofeedback Tidal Volume prompting method have demonstrated significant effectiveness (reduction in average ventilator days from 32.6 to 20.6) in weaning "hard to wean'' patients from ventilator dependency (See: "The Reduction of Weaning Time from Mechanical Ventilation Using Tidal Volume and Relaxation Biofeedback;" Holliday, Jerome E. and Hyers, Thomas M.; American Review of Respiratory Diseases, 1990; v 14 1: pp 1214-1220).
Postulating that "high neural respiratory drive (P 100 >4.5 cmH20)" was the underlying problem, Holliday and colleagues extended their work in a report before the American Association for Respiratory Care convention in 1996 (See: "Reduction in Neural Respiratory Drive to Reduce Ventilator Weaning Failures Using Biofeedback:" Holliday, JE, Haake, R., Range, M; Respiratory Care Open Forum, American Association for Respiratory Care convention, 1996).
The apparatus used was the "Computerized Diaphragmatic Breathing Retraining (CBDR)" from RFB Technologies, Deerfield Beach, Florida. This device consists of virtual reality goggles and earphones, and chest/abdominal motion sensing by infrared radar. Rapid/shallow/irregular breathing produces an unpleasant audio- visual sensation, whereas slow/deep breathing produces pleasant sounds and visual images.
In an uncontrolled study involving nine hard to wean patients, there was significantly suggestive evidence this technique aided the weaning process. These patients developed lower respiratory rates and increased tidal volumes, despite the lack of specific prompting of pulmonary mechanics. And most interesting, while inducing what seems to be these generally accepted desirable breathing patterns, the P 100 values dropped significantly, and, this in conjunction with a reduction of the EEG frequency almost into the alpha range.
The significance of these observations, if any, is not clear. Perhaps the EEG changes noted are casual and unrelated. But the possibility that a technique to induce relaxation and EEG changes should result in breathing pattern modifications felt to be desirable to those interested in the pulmonary mechanics of hard to wean patients is interesting, The perhaps central role of the EEG in this pathophysiologic process seems to be an area worthy of exploration.
In the late 1970's Dr. Warren Smith began development of a single channel EEG monitoring system for clinical use. The object was to use advanced mathematical programs to extract the information content of complex EEG waveforms, and to make a simplified real time visual display icon that could be used by physicians who were unfamiliar with the complexities of standard EEG interpretation.
A common method of complex waveform analysis was, and is, the use of Fast Fourier Transformation (FFT) algorithms, to extract and display the Power Spectrum. This results in a complex graphic output, with multiple irregular peaks that can be difficult to interpret. Moreover, more subtle and potentially important information may be obscured by the complexity of the graphics.
In 1978-79 Dr. Smith was a summer visiting scholar at the Lawrence Livermore Laboratory, where he was introduced to the advanced signal processing techniques which had been developed for military and intelligence purposes. He was permitted to use this technology for civilian purposes.
This resulted in further development of a technique known as Spectral Component Parameter Analysis (SCPA) which computes a Power Spectrum using Parametric Analysis instead of FFT analysis. This results in a smoothing of the typical jagged display obtained from the FFT. More significant however, Parametric Analysis breaks the complex waveform down into up to four individual dominant components.
This information is output in graphic form to a Continuous Power Spectrum Curve. Additionally, these individual components are translated into a simple icon for clinical use.
In this manner, the truly dominant portions of the Power Spectrum are easily detected, and in addition the more subtle components, which may contain relevant information, are appropriately displayed for clinical correlation. The system provides real time information as the output display is refreshed every ten seconds.
The system originally was implemented in an Apple 11 computer, and has been in regular use in several hundred cases in Dr. Smith's dominant interest in EEG analysis during anaesthesia. Dr. Smith plans to convert this program into a modern computer system by use of the LabVIEW development system.
This is a Power Spectral display using typical FFT techniques. Unipolar electrode, with the patient rapidly progressing from very light to deep anaesthesia. Note the jagged, complex FFT display.
The Parametric Analysis method takes the dominant data peaks and provides an analysis of the entire component of each of these areas of data.
Power Spectrum of the same patient using the Parametric Analysis technique. Note the data displays are now more focused, and with a smoother display.
CENTER FREQUENCY: Central frequency of the peaks of the Power Spectrum, Indicated by relative position of the icon within the output graphic, and centered at the base of the side arm.
HALF WIDTH: Widths of the peaks of the Power Spectrum. Indicates the regularity / irregularity of the waveform oscillations by the length of the horizontal bar. (Short = regular --- Long = irregular)
POWER: Spectral Power. The area under the spectral peak of interest. Indicated by the length of the sidearm.
(Short = Low Power --- Long = High Power)
Center Frequency is indicated at the juncture of the side stick graphic. The frequency noted in these examples is 7 and 19 Hz.
The Half Width is a measure of waveform regularity. The top example notes a quite regular waveform, with a short lower horizontal stick indicator. The lower waveform is quite irregular, and this is indicated by a long horizontal indicator.
The Power, or Spectral Power, is the area under the spectral peak. Note, there may be greater Spectral Power in a lower peak of data, if the overall width and base of the spectral peak is wider; i.e. the most dominant peak as seen in an FFT display may not necessarily have the greatest Spectral Power.
Here the upper peak has relatively low power, as indicated by the short sidearm. The lower peak with greater power has a long sidearm.
The present system will evaluate up to four peaks of interest. Here the same patient anaesthesia data is displayed dynamically over a 200 second interval using the Spectral Component Parameters technique.
Here the Spectral Component Parameters are displayed alongside of the Continuous Power Spectrum Curve.
Note the higher (more prominent) middle peak actually has less Power than the left peak. Note the inconspicuous peak on the right, which may in fact contain important data, is fully characterized.
Continuous Power Spectrum Curve on left;
Spectral Component Parameters on the right.
Each of the component parts of the display icon (Center Frequency, Half Width, and Power) are drawn and displayed based on a specific numeric value. These numeric values are therefore available for correlative analysis to associate with relevant clinical events.
Correlative analysis has been attempted by using:
A clinically oriented EEG monitoring system with parallel research capability has been developed for anaesthesia technology. Real time complex data is refreshed every ten seconds and displayed as a simplified icon.
The system analyzes individual waveform components of the EEG by use of Spectral Component Parameter Analysis. The digital output of the individual components of the Power Spectrum can be analyzed by using methods such as Discriminant Analysis, Neural Networks and Logistic Regression.
The system would be suitable for clinical and research needs in respiratory psychophysiology, and other psychological or physiological projects using EEG technology.
1.) Dutton, R. C., Smith, W. D., and Smith, N. T.."The use of the EEG to predict patient movement during anaesthesia." p. 72-82 in Consciousness, Awareness, and Pain in General Anaesthesia, Eds. M. Rosen and J. N. Lunn, London: Butterworths, 1987.
2.) Smith, W. D., Fung, D. L., and Bennett, H. L., "Monitoring the electroencephalogram during anesthesia by spectral parameter analysis," J. Clinical Monitoring, Vol. 1, pp. 95-96, 1985.
3.) Smith, W. D., and Lager, D. L., Evaluation of simple algorithms for spectral parameter analysis of the electroencephalogram," IEEE Trans. Biomed. Eng., Vol. BME-33, pp. 352-358, 1986.