Assuring Heart Valve Reliability

Technology derived from early research in the nuclear power industry is proving useful for assuring the quality of mechanical heart valves - crucial for the more than 80,000 adults who undergo procedures to replace damaged heart valves each year.

By James Lankford, Ph.D.

Dr. James Lankford is director of the Materials Engineering Department in the Mechanical and Materials Engineering Division. A specialist in the fracture and fatigue of materials, he has concentrated his research efforts in the growing field of biomaterials for the last several years. Shown in the background is the acoustic emission-based system used to detect cracks in mechanical artificial heart valves.

At present, nearly one million artificial heart valves are in service in the United States. These devices are essentially of two major types: mechanical and bioprosthetic. The bioprosthetic class of devices are derived from living tissue, both human and nonhuman. Both types of artificial heart valves have clinical advantages and disadvantages, yet they are vital in the treatment of children with life-threatening congenital heart defects. In addition, every year more than 80,000 adults in the U.S. undergo surgical procedures to repair or replace damaged heart valves.

Atomic force microscopy reveals the surface structure of pyrolytic carbon, a unique material used for constructing mechanical heart valves.

Mechanical heart valves, while durable, require that the patient remain on continuous anticoagulation therapy, which can produce negative side effects in certain patients. These valves rarely fail, but when they do, the event can be catastrophic to the patient. Bioprosthetic valves, on the other hand, have finite lives. Currently they are generally created from chemically treated pig heart valves and are subject to failure, usually in less than 10 years, because of the gradual onset of calcification and valve leakage. However, they do not require the patient to undergo anticoagulation treatment, and they are especially suitable for children with congenital heart disease because the valves are relatively flexible and can better accommodate a growing heart. Also, because their degradation is slow, bioprosthetic valves are nearly always replaced well before failure occurs.

Researchers eventually hope to create a heart valve using polymeric materials that can closely simulate the flexible opening and closing of the human heart valve. An added incentive is that polymeric valves may not require aggressive anticoagulation therapy. However, progress in this area has been slow, and such valves are probably decades from in vivo service.

Current prototype polymeric devices fail at far too few cycles because of a combination of calcification and fatigue crack growth, which is a crack extension that occurs in microscopic increments as the external load rises and falls cyclically. For a heart valve, the cyclic load is produced by the beating heart approximately every second.

This historical background explains why much effort, including research conducted by engineers in the Southwest Research Institute (SwRI) Mechanical and Materials Engineering Division and elsewhere, is still focused on improving the performance and reliability of mechanical heart valves.

Pyrolytic carbon

All currently produced mechanical heart valves are constructed using a special form of carbon, called pyrolytic carbon (PyC). This unique material, derived from early research in the nuclear industry for elevated temperature applications, is one of the most blood-compatible of all man-made materials, as opposed to metals, which tend to encourage coagulation and ultimately fatal clotting of blood. The human body recognizes implanted metal as a foreign material, and protects itself from the object by coating it with layers of blood. But for a reason yet to be known, pyrolytic carbon and other so-called blood-compatible coatings are unrecognized by the body and accepted.

In its processed form, pyrolytic carbon is a microscopically smooth, hard, black ceramic-like material. Like ceramic, it is subject to brittleness. If a crack is present, the material, like glass, has little resistance to the growth of the crack and may fail under low loads.

Fortunately, PyC possesses a mechanical property that mitigates this fragility in the presence of flaws, making it inherently difficult to accidentally introduce cracks of significant size into the material. In particular, unlike true ceramics, PyC is highly ductile. Thus, if a sharp, hard object (like the tip of a tool) is pressed into pyrolytic carbon, it can respond by deforming locally to accommodate the object elastically. When the object is withdrawn, there may be (unlike for a ceramic) no residual depression, and little or no microcracking surrounding the site. It is this intrinsic, atomic microstructure-derived resistance to externally imposed crack nucleation that permits such an otherwise brittle material to be used in the human body.

However, despite this tendency of PyC to resist microcrack nucleation, it can occur. The high-temperature fluidized-bed process (see Pyrolytic Properties) by which the material is created generates, as a by-product, small (micrometer scale) internal cracks within the PyC near the parent graphite substrate. In addition, assembled valves are occasionally subjected to accidental damage by cardiovascular physicians themselves, whose sharp, hard metal surgical instruments can serve as inadvertent "indenters." Such incidents may produce either surface or hidden subsurface microcracks that can be quite large (tens of micrometers).

Moreover, it recently was demonstrated by researchers at the University of California at Berkeley that cracks in PyC are subject to stable growth under cyclic loading conditions. Once such a crack reaches a critical size, combined with operative loading, sudden catastrophic failure can occur. As the heart beats, a typical valve will undergo 60 cycles a minute. That translates into 40 million cycles a year, or about 2 billion in an average lifetime. It should be mentioned here that mechanical heart valve failure is rare; only about 50 such failures have occurred out of the approximately one million valves in service. However, valve failures are nearly always fatal.

As noted earlier, the locations of such subcritical cracks in PyC may be subsurface, and therefore invisible during standard optical inspection. Because of this, it appeared that a system of proof testing might offer a means of establishing the structural quality of potentially flawed heart valves. In this method of testing, a device is loaded to a certain stress level. If it survives, it is deemed fit for service. Clearly, there exists the possibility that, if a flaw is initially present, it may grow under the proof load, but perhaps not quite to the critical point. More seriously, the proof test may nucleate a crack that did not previously exist, creating the potential for failure during subsequent service. What is needed, then, is some means of monitoring flaw activity during the course of the proof test, so that devices that do not fail, but nevertheless are flawed, can be rejected.

Acoustic emission detection

One means of determining whether a valve contains a flaw uses acoustic emission (AE) technology. This method, which is used to assess the quality of critical structural components such as nuclear pressure vessels, involves the detection, amplification, and interpretation of the minute elastic stress waves that are emitted during the extension or initiation of a crack. In these events, thousands to millions of atomic bonds may rupture within a few microseconds, causing the propagation of stress waves that can be detected in metals and ceramics by sensitive piezoelectric detectors located relatively far from the actual flaw location. However, prior to the research performed at SwRI, it was not known whether AE technology was potentially associated with fatigue crack nucleation or growth in PyC.

This plot of acoustic emission amplitude versus frequency for crack extension shows an emission peak at 90 kHz (arrow), indicating a normal mode crack extension in a pyrolytic carbon test sample.

Thus, the early efforts of Institute researchers trying to establish the possible validity of an acoustic emission-based quality assurance (QA) system for PyC were directed toward determining whether acoustic emission was associated with the initiation of cracks. Crack initiation is a sudden event, involving the rupture of a large number of atomic bonds, and accordingly provides the maximum chance of detectable acoustic emission. Subcritical crack extension is more subtle because only a "line" of atomic pairs along the crack tip may separate - a relatively smaller acoustic emission event that is harder to detect.

Small disks of pyrolytic carbon on graphite substrates, of the same composition, overall size, and substrate-coating dimensions as that characteristic of commercial heart valves, were used to simulate crack initiation in such valves. For the crack nucleation experiments, the disks were indented by sharp diamond indenters over a wide range of loads, simulating damage during surgical insertion. Simultaneously, possible damage-induced acoustic emission was monitored using small lead zirconate titanate crystals mounted on PyC disks. Special filters and amplifiers controlled the AE frequency window that was monitored, corresponding to a regime (80 kHz to 1 MHz) that was above the AE noise of the test apparatus itself, but that was also known to be associated with microfracture in other brittle materials.

In an extensive series of tests conducted at increasingly higher indentation load levels, it was found that discrete bursts of AE were associated with the indentation process. At low loads, no visible cracks were detected on the surface to correlate with the bursts, while at higher loads, surface cracks were present at the indent corners, but not enough to equal the total number of acoustic emission events. By careful sectioning and microscopic examination of each specimen, it was found that the "missing" cracks were indeed within the PyC, but were lined up along the PyC-graphite interface - the infinitesimally thin region where the material transitions from graphite to PyC. These invisible subsurface cracks nucleated at lower loads, and so accounted for the low-load AE. Moreover, no AE was ever detected without the finding of a group of cracks equal to the number of bursts, and no crack was ever found that did not correlate with an indentation-induced AE event.

Quality assurance system

These results encouraged Institute researchers to test the AE approach in a manner that might simulate a prototype QA system for the material. Accordingly, an apparatus was assembled that would pressurize and realistically load a complete valve assembly - that is, one with leaflets, moving parts that open and close to control blood flow, inserted into the valve body.

In a critical series of tests, valve assemblies were loaded to a level sufficient to generate one or more discrete bursts of acoustic emission in some, but not all, of the devices. (These loads were far in excess of those in actual cardiovascular service.) Inspection of the valves and their housings by optical microscopy revealed no evidence of damage. However, extremely tedious and lengthy serial sectioning of each valve revealed separate subsurface microcracks within the housing directly underneath the support pins around which the valves rotate.

Each crack, on the order of just 10-20 micrometers in length, had nucleated independently within the PyC material, beginning at the PyC-graphite interface. In each case, the number of cracks was equal to the number of AE bursts. For those cases in which no AE was detected, no cracks were found through the sectioning process. These findings were considered to constitute demonstrative proof of the ability of the acoustic emission approach to detect flaw nucleation in PyC cardiovascular devices.

However, there still remained the open and urgent question of whether the AE system could detect crack extension. It is important to know this because quality assurance inspection of a flawed valve assembly requires the confidence that any subcritical extension of the crack during proof testing can be detected. What was required was the AE monitoring of observed, stable macroscopic crack extension in a well characterized sample under controlled load conditions.

In single crystal graphite (top), the carbon atoms are bonded in hexagonal arrays that are stacked in ordered layers. Turbostratic carbon (bottom), however, has disordered stacking through random rotation or displacement of ordered layers. This disordered structure is what gives pyrolytic carbon its unique mechanical properties.

Test program

Candidate PyC-graphite heart valve components were fabricated into tiny specimens. To eliminate the possibility of AE from the graphite, the graphite was machined off, leaving only thin sheets of PyC for testing. Miniaturized grips, a low capacity load cell, and other special equipment were developed to apply and control the very low cyclic loads required to activate the cracks.

It was found that during steady operation under realistic cardiovascular conditions (frequency, stress level) acoustic emission was indeed detected, but only intermittently. Often hundreds of cycles would pass before a discrete burst of AE would occur, always at characteristic frequencies of either 90 kHz or 110 kHz, although occasionally double AE peaks would appear, located at both frequencies.

These puzzling and, at first, worrisome phenomena were resolved by coupling the AE experiments with detailed real-time optical microscopy of the (fortunately) visible cracks in the test samples.

In these experiments, it was found that cracks in PyC respond to cyclic loading by extending only occasionally, in fairly large "jumps." The latter corresponded to the intermittent AE bursts. An increment extension always yielded AE, otherwise AE signals did not occur. Moreover, if a load were applied in excess of the current cyclic load level, crack extension and associated AE occurred immediately.

The bimodal frequency was explained by the relationship of the crack to the direction of loading. If the loading were perpendicular to the crack plane, the crack faces simply opened normal to themselves, and AE then occurred at 90 kHz. If the crack were at an angle to the load, it could open in a pure shear mode, so that the crack faces rubbed on each other, shifting the AE frequency to 110 kHz. Combined mode opening causes emission at both frequencies simultaneously.

Conclusions

Based on the research conducted, the feasibility of acoustic emission as a heart valve quality assurance technique was established. Stress wave signals that are emitted from PyC heart valves clearly correlate with either crack nucleation or with subsequent crack growth, and if a crack nucleates above a critical size of a few micrometers it will be detected.

Using the project findings as a guideline, an assembly-line acoustic emission-based quality assurance system was put in place using commercially available AE components, coupled with a special heart valve pressurization fixture. Subsequent laboratory evaluation proved the system to be fast, rugged, stable, and resistant to extraneous laboratory noise. Detection of known flaws was insensitive to both flaw source (within the valve assembly) and detector location, implying that it makes no difference in a practical QA system exactly where the detector is located.

Scientists at SwRI continue to evaluate AE technology for detecting fatigue and fracture in aircraft structures and pressure vessels and to study fatigue-induced composite delamination. Applications such as these are similar to the artificial heart valve in that failure is intolerable. Acoustic emission clearly offers the opportunity to unambiguously detect, and thereby eliminate from service, flawed components that might suffer eventual cyclic fatigue failure.

Published in the Summer 1999 issue of Technology Today®, published by Southwest Research Institute. For more information, contact Maria Martinez.

Technics Summer 1999 Technology Today
SwRI Publications SwRI Home