|Home | About | Journals | Submit | Contact Us | Français|
The objective of this study was to validate, both in vitro and in an ex vivo model, a technique for the measurement of forces exerted on surgical sutures. For this purpose, a stainless steel E-type buckle force transducer was designed and constructed. A strain gauge was mounted on the central beam of the transducer to measure transducer deformation. The transducer was tested and calibrated on a single strand of surgical suture during cyclic loading. Further validation was performed using a previously published cadaveric model of laryngoplasty in the horse. Linear regression of transducer output with actual force during calibration tests resulted in mean R2 values of 1.00, 0.99, and 0.99 for rising slope, falling slope, and overall slope, respectively. The R2 was not less than 0.96 across an average of 75 cycles per test. The difference between rising slope and falling slope was 4%. Over 45 846 samples, the predicted force from transducer output showed a mean error of 4%. In vitro validation produced an adjusted R2 of 0.99 when the force on the suture was regressed against translaryngeal pressure in a mixed-effects model. E-type buckle force transducers showed a highly linear output over a physiological force range when applied to surgical suture in vitro and in an ex vivo model of laryngoplasty. With appropriate calibration and short-term in vivo implantation, these transducers may advance our knowledge of the mechanisms of success and failure of techniques, such as laryngoplasty, that use structural suture implants.
L’objectif de la présente étude était de valider, autant dans un modèle in vitro que ex vivo, une technique pour la mesure des forces exercées sur des sutures chirurgicales. À cette fin, un transducteur de force en acier inoxydable de type boucle en E a été élaboré et machiné. Un cadran de force a été installé sur le montant central du transducteur afin de mesurer la déformation du transducteur. Le transducteur a été testé et calibré sur un brin unique de suture chirurgicale lors de chargement cyclique. Une validation additionnelle a été faite à l’aide d’un modèle cadavérique, antérieurement décrit, de laryngoplastie chez le cheval. Une régression linéaire du signal du transducteur avec la force actuelle durant les tests de calibration a résulté en des valeurs moyennes de R2 de 1,00, 0,99 et 0,99, respectivement, pour la pente ascendante, la pente descendante et la pente globale. La valeur de R2 n’a jamais été moins de 0,96 pour une moyenne de 75 cycles par test. La différence entre la pente ascendante et la pente descendante était de 4 %. Sur un total de 45,846 échantillons, la force prédite à partir du signal du transducteur présentait une erreur moyenne de 4 %. La validation in vitro a permis d’obtenir une valeur ajustée de R2 de 0,99 lorsque la force sur la suture était régressée contre la pression translaryngale dans un modèle à effets mêlés. Les transducteurs de type boucle en E ont permis d’obtenir un signal très linéaire sur une étendue de force physiologique lorsque appliquée à des sutures chirurgicales in vitro et dans un modèle ex vivo de laryngoplastie. Avec les calibrations appropriées et des implantations in vivo de courte durée, ces transducteurs permettraient de faire progresser nos connaissances des mécanismes de succès et d’échec de techniques, telle la laryngoplastie, qui utilisent des implants structuraux de suture.
(Traduit par Docteur Serge Messier)
The structural use of one or more strands of suture is inherent to many soft tissue and orthopedic surgical procedures, including joint reconstructions and upper respiratory tract interventions. Although the sutures must often bear considerable cyclic forces after implantation, minimal data exist to document these forces in vivo. This can partly be attributed to the lack of appropriate and accurate techniques for measuring force in vivo. We set out to design an implant that would ultimately be suitable for measuring dynamic force on sutures in vivo, and particularly to determine the forces on sutures used in equine laryngoplasty. This study describes the construction and testing of our implant, with determination of transducer output linearity, both in vitro and in an ex vivo model of laryngoplasty that simulates in vivo conditions.
Laryngoplasty is the most common approach for the treatment of recurrent laryngeal neuropathy (RLN) in horses. During the immediate post-operative period, however, many horses suffer loss of abduction and reduction in the cross-sectional area of the rima glottidis, with subsequent return of exercise intolerance and abnormal respiratory noise (1–5). Cyclic loading of the implant-cartilage construct has been implicated in the failure of laryngoplasty (6–8). As a result of cyclic loading, the suture may cut through the muscular process of the arytenoid or the cricoid cartilage, or both (7,9–11). In vivo, cyclic forces are likely to occur during respiration, due to negative intra-luminal pressure and during swallowing, due to contraction of the intrinsic and extrinsic laryngeal musculature (2). Previous studies have used load cells to measure force on sutures in a variety of models, but these techniques are limited to cadaveric experiments or experiments under general anesthesia (7,12). E-type buckle transducers have been used for many years to measure tendon and ligament force in vivo (13,14). Increasing force on the structure being measured results in deformation of an E-shaped stainless steel template. This force is measured by a strain gauge mounted on the central arm of the ‘E’ (15). Such transducers have been shown to be more accurate and sensitive and provide more reproducible data than alternative techniques for measuring force (16). When used in vivo, these transducers measure change in force during a given activity, rather than measuring actual force at a single point in time. This most important distinction means that the key determinant of measurement accuracy is the linearity of the transducer output over the range of forces anticipated in vivo. Therefore, in previous studies such transducers have been calibrated before and after implantation by application on a single strand of suture or nylon cord and in situ during terminal studies (17,18). Our long-term goal is to develop a minimally invasive technique that can be used to measure the force exerted on surgically implanted laryngoplasty sutures while subjects perform normal activities. For this application, in-situ calibration would not be possible. This paper therefore describes preliminary testing, calibration, and error quantification, including linearity of output, of an E-type buckle force transducer as a prerequisite for in vivo implantation and measurement of the force on surgical sutures.
E-type buckle transducers were constructed from grade-430 stainless steel. The transducers were polished with emery paper and prepared for mounting of the strain gauge using acidic and alkaline surface cleaners (M-Prep Conditioner A and M-Prep Neutralizer 5A; Vishay Micro-Measurements, Shelton, Connecticut, USA). A single element 120 Ω foil strain gauge with pre-attached 3-lead wires (type FLG-1-17-3LT; Tokyo Sokki Kenkyujo, Japan) was bonded to the central arm using cyano-acrylate adhesive (SG496; Omega, Stamford, Connecticut, USA). The entire transducer was coated with xylene-cured polyurethane (M-coat A; Vishay Micro-Measurements). The principle of operation of an E-type force transducer is as follows. The suture is applied to the transducer by passing the suture over the outer arms and under the central arm of the E (Figure 1). As tension is applied to the suture, the central arm deflects; in response to this deformation, the electrical resistance of the strain gauge changes proportionally. A 3-wire 1/4 Wheatstone bridge was used to complete the strain gauge circuit in all experiments (19). This approach compensates for the effect of a change in temperature of the lead wire on bridge balance and increases measurement sensitivity compared to a two-wire configuration. A bridge completion module (BCM-1; Omega) was used in conjunction with an AC-powered signal conditioner (DMD-465; Omega). For all experiments, the bridge was balanced at zero force.
Initial testing and gauge calibration were performed using a bench-top materials testing machine. The following calibration procedure was performed for each gauge before and after each experiment to ensure that no change in transducer sensitivity had occurred. A single 10 cm strand of ultra-high molecular weight polyethylene suture (#5 Fiberwire; Arthrex, Naples, Florida, USA) with loops at each end was mounted between a 225 Newton (N) load cell (Sensotec Model 31; Columbus, Ohio, USA) and the actuator of an electromagnetic axial compressive apparatus (EnduraTEC ELF Series 3220; Bose, Eden Prairie, Minnesota, USA). After balancing the Wheatstone bridge at zero force, the buckle transducer was mounted on the suture as described in Figure 1. Using load control, a cyclic force of 5 to 50 N, selected as the range of forces likely to be encountered in vivo, was applied to the suture at a physiologically relevant frequency of 2 Hz. Actual force and buckle voltage output were recorded simultaneously at 500 Hz for at least 40 cycles in WinTest software (v2.54; EnduraTEC, Columbus, Ohio, USA). Data were analyzed using custom software written in Matlab (The Mathworks, Natick, Massachusetts, USA). Each cycle was divided into rising force and falling force. Linear regression of force versus the voltage output of the buckle transducer was performed for rising and falling force, both separately and for the entire cycle, yielding a slope and an R2 value for each. In most cases, slight hysteresis was seen between rising and falling force. The difference in slope between rising and falling force was quantified as a percentage of the value for rising force. Mean rising, falling, and overall R2 values and percentage difference values were calculated across all transducers. Subsequently, calibration values (correlation coefficients of overall regression across all cycles) were applied to the transducer output at each data point, in order to calculate the predicted force. Actual force was subtracted from predicted force to yield an error value in Newtons, with a negative error indicating underestimation of force and a positive error indicating overestimation. The hypotheses for transducer calibration were that hysteresis between the rise and fall of force (difference in slopes) would be less than 5% for all cycles across all transducers, that actual and transducer-determined force would be highly correlated for each cycle, and that mean predicted force error across all transducers would be less than 5%.
A laryngoplasty suture was placed bilaterally in 10 adult equine larynges. Larynges were collected at necropsy, retaining a section of trachea, frozen within 1 h, and subsequently handled as described by Cheetham et al (20). Laryngoplasty was performed in standard fashion, with a single strand of ultra-high molecular weight polyethylene suture (#5 Fiberwire; Arthrex) placed from the caudal aspect of the cricoid cartilage to the muscular process of the arytenoid cartilage. Suture placement was standardized and both arytenoid cartilages were abducted to the published optimal degree [grade 2; (5)] before each suture was tied with 5 throws. After balancing the Wheatstone bridge at zero force, one of 5 transducers was randomly applied to the lateral arm of either the left or the right laryngoplasty suture on each specimen. The transducer was therefore subjected to force in its baseline state. The baseline force was dictated by the force required to achieve optimal abduction of the arytenoid cartilage and therefore varied from larynx to larynx. Since our interest was in the change in force, the offset was subtracted in subsequent analysis.
Larynges were then placed in a previously described in vitro cyclic air flow model. The model employs a constant vacuum and oscillating valve to draw air through the larynx and has been shown to generate physiological airflows and pressures within the laryngeal lumen (20). Teflon catheters (1.3-mm internal diameter, Neoflon; Cole-Parmer, Chicago, Illinois, USA), connected to differential pressure transducers (Celesco LCVR; Celesco Transducers Products, Cango Park, California, USA), were placed within the tracheal remnant and at the rima glottidis to measure tracheal and pharyngeal pressures, respectively. Pressure transducers were referenced to atmospheric pressure and calibrated from −70 to 70 mmHg using a manometer. To standardize air pressure and flow across larynges of different sizes, the oscillating valve was maximally opened and the vacuum adjusted to give a static tracheal pressure of −33 mmHg before cyclically testing each larynx. The valve was then cycled (open/closed) at a physiologically relevant frequency of 2 Hz. Tracheal and pharyngeal pressures and transducer voltage were recorded simultaneously for 30 s (approximately 60 cycles). Data were collected via a 12-bit AD card (DAQCard 6024E; National Instruments, Austin, Texas, USA) and recorded at 500 Hz using custom software written in Matlab (The Mathworks).
The voltage output of the transducer was converted to suture force [Newtons (N)] by applying the previously determined calibration value (linear regression slope) for the transducer. Minimum absolute tracheal pressure was used to cut data into individual cycles. Translaryngeal pressure was calculated as the difference between the pressure at the rima glottidis and the pressure within the tracheal remnant. Cycles were interpolated to average cycle length and the mean cycle was calculated for each condition. A mixed-effect model was fitted to the mean cycle data with specimen as a random effect, absolute translaryngeal pressure as a fixed effect predictor, and force on the suture as the response variable (JMP; SAS Institute, Cary, North Carolina, USA).
Calibration data were collected over a total of 374 cycles for 5 transducers. An average of 75 cycles was collected per transducer calibration test. In all tests, minimal hysteresis was seen between rising force and falling force in regression of transducer output against actual force. The mean difference between the slopes for rising and falling force was 4% over all transducers and all cycles. The mean R2 values for rising force, falling force, and complete cycles were 1.00, 0.99, and 0.99, respectively. The minimum R2 for any individual cycle was 0.96.
Actual force and force predicted from the output of a single transducer are plotted against time for 2 cycles in Figure 2a. While transducer output follows actual force well in linear regions, the transducer underestimates force at the top and bottom of the curve. Figure 2b shows predicted force against actual force for representative cycles. A total of 45 846 samples was used for error determination over all transducers and all trials. Mean absolute error [± standard deviation (s)] in predicted force was 0.55 ± 0.42 N, which equates to a mean percentage error of 4.4 ± 6.6%. Force predicted from transducer output was within 2% of the actual force for 26 903 samples (59%) and within 4% for 34 071 samples (74%). The distribution of percentage error in predicted force is shown in Figure 3. The effect of underestimation of force can be seen in the tail to the left.
An average of 58 cycles was collected per specimen. The relationship between translaryngeal pressure and force on the suture for each larynx is shown in Figure 4. The mixed-effect model fitted the data extremely closely (adjusted R2 = 0.99). After controlling for differences between larynges, for every 1 mmHg increase in the absolute value of translaryngeal pressure, there was a mean increase in suture force of 0.034 N. This relationship was highly significant (P < 0.0001). There were significant differences between the least squares (adjusted) means for all larynges (Tukey’s post hoc test; P < 0.05).
This study set out to evaluate an E-type force transducer for the measurement of dynamic force on surgical sutures. During calibration using a materials testing machine the transducer showed linear behavior across the force range anticipated in our ultimate application and at a physiological frequency of 2 Hz. Minimal hysteresis between transducer output during the rise and fall of force and good correlation between actual and transducer-determined forces were seen and a mean percentage error of less than 5% was measured. While this error value is clearly not as low as can be achieved with a load cell, it is within our expected error and allows this technique to be further investigated for potential in vivo application. The technique is particularly suited to studies that aim to measure a change in force under dynamic conditions, rather than to determine absolute one-off forces.
The frequency response of the system was not definitively determined by performing calibration at a variety of loading frequencies. Although the authors anticipated that transducer output would be uniform across a range of low frequencies, the data presented here (Figure 2a) and limited preliminary testing at higher frequencies (> 5 Hz) suggest that transducer deformation and recoil may lag behind the force on the suture, yielding less linear results. The actual error during linear changes in force is likely to be much lower than the 4.4% presented here. The frequency response of the system is considered appropriate for measuring the physiological frequency of respiration at exercise (~2 Hz).
To test our transducer in a realistic environment before in vivo implantation, we employed an in vitro model capable of generating physiological laryngeal airflows. This model was able to reproduce one component of the force that is likely to be exerted on laryngoplasty sutures in vivo, namely the force due to changes in intra-luminal pressure associated with respiration. The other likely contributors to cyclic force on sutures in vivo are coughing and swallowing. These activities would be difficult, if not impossible, to mimic convincingly in vitro. Although our validation was therefore limited to cyclic airflows, the results of these experiments demonstrate adequate stability of transducer response and can be extrapolated to other circumstances. Translaryngeal pressure and suture force were significantly correlated, providing evidence that changes in intra-luminal pressure are likely to cause changes in loading of the suture in vivo.
Previous in vitro experiments have measured the cyclic forces on laryngoplasty sutures associated with changing intra-luminal pressure (21). Those experiments used load cells mounted in series with the suture on cadaveric larynges and are a useful comparison in terms of the accuracy of our transducers. In the study by Ducharme, the oscillating force generated by cyclic airflow was determined to be 0.17 ± 0.24 N, which is similar to the force determined here (21).
Since the technique relies on changes in electrical resistance of the strain gauge, fluctuations in environmental temperature often play a role in absolute strain gauge outputs during in vitro or materials testing applications. Attempts were made to minimize this effect by using a 3-wire 1/4 Wheatstone bridge. Changes in temperature are of less concern in the in vivo application of strain gauges in endothermic species. Ultimately, changes in force in response to a given action or activity of relatively short duration are often of greater interest than absolute values, and acute temperature fluctuations are unlikely to be encountered. At the start of all strain gauge experiments, the bridge is balanced and the output zeroed. Differences in intercept do not play a large role in transducer accuracy.
In conclusion, this paper has shown that E-type buckle force transducers are sensitive and accurate enough for measuring the force on surgical sutures. In vitro and in an ex vivo equine laryngeal model they demonstrated a highly linear response over the anticipated range of forces and at a physiological frequency.
The authors acknowledge Drs. Lawrence Bonassar and Jason Gleghorn of the Laboratory for Tissue Engineering, Sibley School of Mechanical and Aerospace Engineering and Department of Biomedical Engineering, Cornell University for assistance with transducer calibration.