Medical Device Link.

Originally Published MX January/February 2005

BUSINESS PLANNING & TECHNOLOGY DEVELOPMENT

By integrating system theory into their products, medtech companies can create a defensible portfolio of innovative medical devices.

Gail D. Baura

Innovation to address clinical needs is at the heart of medical device breakthroughs. When Wilson Greatbatch serendipitously discovered an electronic circuit that pulsed in a manner similar to a heartbeat, he recognized that circuit as the potential foundation of what would become the world's first implantable pacemaker. When Paul Lauterbur, corecipient of the 2003 Nobel Prize for Medicine, adjusted the intensity of the magnetic field during a course of imaging studies, he discovered the capabilities of the two-dimensional imaging technique later named magnetic resonance imaging. By simplifying complex problems, innovation breakthroughs lead to competitive advantages, new markets, and higher market share for medtech companies.

Figure 1. The signal extraction technology employed in pulse oximeters by Masimo Corp. (Irvine, CA) is based on adaptive filtering, a system theory technique. (click to enlarge) Photo courtesy MASIMO CORP.

One innovation-engendering technology that is often overlooked even though it can be the basis of breakthrough products is system theory. System theory is a special type of engineering mathematics that enables, among other things, the recognition of original signals in the midst of noise, and the accurate classification of signal subtypes.¹ It has been applied in the design of patient monitoring and diagnostic devices for more than a decade.

The strong potential of system theory for medical device applications was relatively unknown outside the engineering community until 1995, when Masimo Corp. (Irvine, CA) began publicizing its signal extraction technology (SET), which provides accurate pulse oximetry readings during difficult patient interferences such as motion (see Figure 1). SET is based on adaptive filtering, a system theory technique first investigated in the late 1950s.

Figure 2. The bispectral index system by Aspect Medical Systems (Newton, MA) was developed using system theory techniques. (click to enlarge) Photo courtesy ASPECT MEDICAL SYSTEMS

During the early 1990s, other devices also utilized system theory as a key component of signal processing, but their makers did not promote system theory as part of these devices' feature sets. Aspect Medical Systems (Newton, MA) developed the bispectral index (BIS) monitor, which has been approved by FDA for determining a patient's level of consciousness under anesthesia. BIS employs frequency-domain processing and was developed using system theory techniques (see Figure 2). Interflo Medical, which has since been acquired by Baxter Healthcare (Deerfield, IL), used a pseudorandom binary sequence (PRBS) algorithm in its Vigilance monitor to continuously calculate cardiac output via a special thermodilution catheter. IVAC Corp., now Alaris Medical Systems Inc. (San Diego), also used PRBS in its Signal Edition large-volume infusion pump to calculate catheter resistance as a predictor of drug-infusion complications.

Although system theory is still underutilized in patient monitoring, four factors are beginning to bring this technology to prominence: its promise of greater utility; the wide-open field of patentability; regulatory precedent; and time-to-market advantages. This article discusses the benefits device manufacturers can expect to realize from adopting this innovative approach in their products.

Increased Device Utility

One reason the patient monitors just identified used system theory rather than traditional heuristic curve fitting--that is, an iterative technique based on empirical data--was to increase the accuracy of their processing (see Table I). Heuristic algorithms look at simple relationships between waveform curves, which do not generalize well in critically ill patients or under conditions such as patient movement that cause unusual waveform shapes. By contrast, because they are built on mathematical models that can predict a variety of behaviors, system theory­based algorithms remain viable in the face of these unusual waveform shapes. At the time Masimo developed SET, the algorithms used by other manufacturers of pulse oximeters were derived empirically and thus were not effective at minimizing the effect of patient motion on the accuracy of oxygen readings.

The monitors used system theory also to compensate for less-than-ideal monitoring conditions. Without system theory, the small pulses of heat generated by the Interflo thermodilution catheter would not be large enough for cardiac output calculations. Similarly, naturally present blood-pressure pulses could not be separated from IVAC infusion-pressure pulses to allow for catheter-resistance calculations if not for system theory.

As the world's healthcare systems move toward personalized healthcare, system theory, which encompasses artificial intelligence (AI) techniques, offers an ideal platform for customized diagnosis (see Table II). System theory algorithms provide accurate classification of various patient states in real time. For example, the fuzzy model within the AutoPap 300 by NeoPath (now TriPath Imaging Inc.; Burlington, NC) enabled that system to become what is still the only FDA-approved primary, unassisted screening system for identifying cancerous Pap smears.

Wireless Holter monitors of the current generation, such as those marketed by CardioNet (San Diego), detect the possibility of an arrhythmic event, but the event's occurrence must still be confirmed by an off-site specialist. Future Holter monitors, however, may be able to identify arrhythmias on-site with much greater accuracy. Algorithm research involving system theory technologies such as neural networks is being conducted with the aim of greatly improving arrhythmia detection accuracy by comparison with the capabilities of previous-generation monitors.

These types of algorithms will also be applied to other signals besides electrocardiogram and Pap smear images. Sophisticated glucose sensors are now viable, but, to make an implantable insulin pump a reality, a control algorithm for personalized administration of insulin is needed. Insulin dosage, perhaps under fuzzy control, would vary by patient and by glucose levels derived from food intake. As home-use monitors become commonplace, the accurate real-time analysis of vital-sign waveforms could result in an AI-derived prediagnosis that automatically triggers a telephoned request for a physician visit.

Intellectual Property

Defensible intellectual property (IP) is an essential foundation of medtech companies. Device manufacturers eagerly seek out opportunities to file IP as the basis for innovative products.

The U.S. Patent and Trademark Office (PTO) considers the combination of system theory and patient monitoring quite novel--and therefore also quite patentable. To date, few system theory techniques have been applied to patient-monitoring signals. When a knowledgeable patent attorney crafts generalized claims for an innovative technology, the manufacturer may be able to obtain broad coverage. PTO accepted all of the 54 claims submitted by CardioDynamics (San Diego) in 2003, for example, thereby protecting the company's combination of time-scale distributions and impedance cardiography for the next 20 years. The time-scale distributions based on system theory in this case improve noise immunity during the waveform analysis of impedance cardiography, a noninvasive continuous measurement of cardiac output. ²

Further, claims based on technology-combination IP can be successfully defended against patent infringement. In March 2004, a federal jury awarded Masimo $134.5 million in damages after finding that the Nellcor unit of Tyco Healthcare (Pleasanton, CA) willfully infringed four patents containing claims based on adaptive filtering combined with pulse oximetry.³ The U.S. Court of Appeals for the Federal Circuit later overturned the finding of willful infringement, but in August 2004 it also increased the award to $164 million for five additional months of infringement.⁴ Had willfulness been part of the final judgment, damages would have been three times the final award.

Strong patent portfolios enable medtech companies to protect their innovative system theory technologies, including, as the Masimo-Nellcor contest vividly demonstrates, defending them successfully against incursion by competitors. The IP should comprise the combination aspect of the device or technology.

Regulatory Clearance and Indications for Use

FDA is aware of system theory. The agency has addressed the regulatory status of devices that make use of system theory in an appendix to its guidance document on device software.⁵ Special topics covered in the appendix include AI, neural networks, and closed-loop control. The intent of this appendix is "to give reviewers a 'heads-up' to alert them that the software . . . may need special attention or further research." Given such a clear warning from the regulator, it is advisable to file FDA submissions for system theory­based devices in stages, and to include sufficient validation documentation, so that clearance will not be delayed.

Whenever possible, premarket submissions for system theory­based devices should be filed as premarket notifications (510(k)s) rather than as slower-moving premarket approval (PMA) applications.

The average time for PMA approval is currently about one year, with exceptions for special cases. In 1997, Gensia Automedics Inc. (San Diego) received such approval for its combination drug/closed-loop system after an arduous 44-month review process.⁶ The delay occurred because Gensia's device was the first combination product sent to FDA for approval. The long preset clock for a PMA increases the probability that an FDA reviewer will spend more time analyzing system theory validation results and will request more clinical data.

A recommended strategy is to file two serial 510(k)s. The aim of the first submission would be to establish equivalency to a predicate device. Later, the manufacturer could submit a second 510(k) for an additional indication for use that highlights the device's innovative system theory behavior. A successful example is provided by Aspect Medical Systems' BIS. The company's original 510(k) clearance was for a standard electroencephalograph to measure brain wave activity. Later, the firm filed a 510(k) to add this indication for use: "The bispectral index, a processed electroencephalogram variable, may be used as an aid in monitoring the effects of certain anesthetic agents."⁷

The choice of clinical data, when necessarily submitted to demonstrate the safety and effectiveness of a system theory algorithm, should be carefully considered. The protocols and data submitted should efficiently validate an algorithm's utility. Should questions arise when FDA reviews the data analysis, 510(k) clearance will be delayed.

Typically, a 510(k) clears in less than 90 days. However, a March 2003 submission by Zargis Medical Corp. (Princeton, NJ) for an electronic stethoscope with identification of heart murmurs using time-frequency distributions stalled during review of clinical data.⁸ The company's 510(k) was finally cleared in June 2004, after additional clinical data were supplied. ^9,10

Faster Time to Market

During the first, venture-capital- funded, development phase of a start-up company, an undesirable slip in the product-completion schedule could occur. A primary factor in schedule slippage is underestimation of the signal processing work involved in the project.

When new hardware and sensors are under advanced development to solve a difficult, previously unmanageable problem (for example, continuous, noninvasive blood pressure monitoring), chances are good that the signal-processing algorithms being employed are also new. Many start-ups are loath to admit this, but signal-processing algorithm research often is necessary to unite all parts of the device system into an entity capable of cohesive measurement. Delaying signal-processing work until a relatively late stage of the project often causes schedule slippage.

Basing intelligent medical devices on system theory rather than empirical processing rules may actually keep development on schedule, preventing the risk of slippage and the possible need for a bridge loan. Certainly though, more staffing will be required; key system theory positions will have to be filled. But once the system theory researcher and support staff are in place, algorithm research can be conducted neatly in parallel with the advanced development portion of the project.¹¹ System theory is an innovative solution for efficient realization of an accurate algorithm.

Conclusion

System theory has been applied in innovative patient monitoring and diagnostic devices for more than a decade. It enabled technology breakthroughs by companies such as Masimo, Aspect Medical, and Neopath, who have become technology and market leaders in their specialties.

While it is almost impossible to invent a new specialty, an existing specialty can be transformed by means of system theory. System theory increases the accuracy of signal processing beyond what can be achieved with traditional heuristic curve fitting. It compensates for less-than-ideal monitoring conditions. As healthcare becomes ever more personalized, system theory provides an ideal platform for customized diagnosis. The technology can be protected and defended by patents. With all these benefits, it is surprising that system theory has not been deployed in more medical devices.

References

1. GD Baura, System Theory and Practical Applications for Biomedical Signals, Biomedical Engineering Series (Hoboken, NJ: Wiley-Institute of Electrical and Electronics Engineers Press, 2002).
2. GD Baura and SK Ng, "Method and Apparatus for Hemodynamic Assessment Including Fiducial Point Detection," U.S. Pat. 6,561,986 (May 13, 2003).
3. MW Walsh, "Tyco Unit Loses Patent-Infringement Case," New York Times, March 27, 2004, sec. C, p. 4.
4. "Court Enters Final Judgement against Tyco's Nellcor Division for Infringement of Masimo's Patents," Masimo press release (August 11, 2004).
5. Food and Drug Administration, Guidance for FDA Reviewers and Industry: Guidance for the Content of Premarket Submissions for Software Contained in Medical Devices (Rockville, MD: FDA, Center for Devices and Radiological Health, 1998).
6. "FDA Allows Use of Device in Diagnosing Heart Disease," Wall Street Journal, September 15, 1997, sec. B, p. 4.
7. Aspect Medical Systems, "A-1000 EEG Monitor and 1050 EEG Monitor," FDA 510(k) K963644 (October 8, 1996).
8. T Oskiper and R Watrous, "Results on the First Time-Frequency Characterization of the First Heart Sound in Normal Man," in Proceedings of the Second Joint EMBS/BMES Conference (Houston: IEEE Engineering in Medicine and Biology Society, 2002): 126­127.
9. "Zargis Medical Updates on FDA Submission," Zargis press release (April 6, 2004).
10. "FDA Clears Zargis' Pioneering Medical Device," Zargis press release (June 1, 2004).
11. GD Baura, "System Theory in Industrial Patient Monitoring: An Overview," in Proceedings of the 26th Annual Engineers in Medicine and Biology Conference (Houston: IEEE Engineers in Medicine and Biology Society, 2004), 5356­5359.

Gail D. Baura, PhD, is vice president for research at CardioDynamics (San Diego), a medical device company that develops, manufactures, and markets noninvasive cardiac output monitoring technologies.

Copyright ©2005 MX