|Home | About | Journals | Submit | Contact Us | Français|
Compartment syndromes, caused by elevated intramuscular pressure (IMP) and resulting from trauma or chronic overuse, frequently require invasive IMP monitoring for accurate diagnosis. Our objective is to test a non-invasive ultrasound technique for estimating IMP based on fascial displacement waveforms from arterial blood pressure pulses. In this study, IMP was increased in the legs of 23 healthy adult subjects up to 80 mmHg using two blood pressure cuffs covering the region from the knee to the ankle. Receiver operator characteristic (ROC) curves and recursive partitioning were used to determine the sensitivity and specificity of diagnosing elevated IMP using fascial displacement. For one ROC curve, in which several ultrasonic measurement parameters were used along with subject body mass index and blood pressure, the sensitivity and specificity for diagnosing normal IMP (below 30 mmHg) from elevated IMP (30 mmHg and up) was 0.61 and 0.94 respectively. Recursive partitioning, in which IMP was divided into three ranges (normal < 30 mmHg; mid-range of 30-40 mmHg and elevated >= 50 mmHg), resulted in improved diagnostic sensitivity (0.77) with almost no change in specificity (0.93).
Compartment syndrome is a condition in which high pressure within a closed fascial space (muscle compartment) reduces capillary blood perfusion below the level necessary for tissue viability. Compartment syndromes develop in skeletal muscles that are enclosed by relatively noncompliant, osseofascial boundaries, where a build-up of pressure is not easily dissipated.  In acute cases, this pressure build-up may result from tibial fracture, blunt trauma resulting in interstitial edema, hemorrhage, post-ischemic muscle fiber swelling, or venous obstruction caused bv burns or compartment volume constriction (e.g., a tight cast). Chronic cases, in turn, may be caused by repeated strenuous exercise. Treatment of acute compartment syndrome relies on early recognition and prompt fasciotomy to prevent Volkmann's contracture [2-4]. In chronic cases, symptoms may disappear with rest, although in many cases fasciotomy is also required to relieve symptoms.
Compartment syndrome can often be diagnosed on the basis of clinical examination, but recognition can be difficult in cases of severe trauma where the patient is unable to convey the early symptoms of pain. An important adjunct to clinical diagnosis of compartment syndrome is direct measurement of intramuscular pressure (IMP) by insertion of a catheter into the muscle at risk, for example, the use of a slit catheter to provide continuous measurement of IMP . Most clinicians accept direct measurement of IMP as the gold standard for compartment syndrome diagnosis. However, due to the invasive nature of the measurement and because there is considerable disagreement over what is the threshold pressure for diagnosis of compartment syndrome (as time after injury is also an important factor), some clinicians are reluctant to perform direct measurement of IMP in all cases.
Inaccuracy in the current methods of compartment syndrome diagnosis often leads to delays in treatment. Several recent studies have shown that near-infrared spectroscopy (NIRS) can be an effective tool for diagnosis of chronic, exertional compartment syndrome. These studies demonstrated a significant inverse correlation between IMP and oxyhemoglobin level, when IMP reaches critical values some form of shock often accompanies trauma, which leads to low oxygen tissue saturation globally. Low global oxygen saturation can lead to misdiagnosis of acute compartment syndrome. Also, NIRS has trouble measuring oxygen saturation in deep muscle compartments, as the infrared signal can only penetrate tissue 2-3 cm deep [7-10].
In studies of chronic exertional compartment syndrome, magnetic resonance imaging has shown increased T2 signal intensity with anterior compartment pressures [11-13]. However, due to its large size and cost, MRI use is limited to radiology departments and usually impractical. A lower cost, portable alternative is noninvasive measurements of tissue hardness using a tonometer-type device. Initial results indicated low diagnostic sensitivity and specificity (68% sensitivity and 96% specificity) . A more recent study showed improved performance, but even in this case, diagnosis is limited to superficial compartments where the tonometer may be placed over the compartment .
Noninvasive ultrasonic measurements of muscle compartment motion provide a promising alternative to current methods. Lynch and co-workers  first described the technique, in which ultrasonic pulsed phased locked loop (PPLL) was used to measure micron-level tissue displacements caused by saline infusion into a cadaver model. These measurements showed that the expansion of the muscle compartment with saline infusion was related to the nonlinear pressure-volume curve. Further work on human subjects showed that the ultrasonic PPLL could detect arterial pulsations in the muscle compartment, and that automated waveform analysis could be used to estimate IMP without calibration . In this study, the harmonic content of the arterial pulse waveform was used to determine pressure, and the ratio of the fundamental frequency of this waveform to the first harmonic (harmonic ratio) correlated to IMP with an R2 value of 0.89.
Previous work on the harmonic analysis of intracranial pressure waveforms has suggested that the harmonic ratio is related to tissue compliance, and is therefore an indirect measure of pressure [17-21]. In a follow-up study of model compartment syndrome study in pigs, in which plasma infusion to the anterior compartment was used to increase IMP, harmonic ratio correlated poorly with pressure . This suggests that the relationship between compliance and pressure varies with compartment type and species. In this same study, it was found that the amplitude of the arterial pulse waveform was more directly related to IMP than harmonic ratio. The relationship was nonlinear, with pulse amplitude initially increasing with IMP until it approached the mean arterial pressure. This trend is consistent with the findings of Kim and associates  who measured the displacement of the arterial wall using a 2-D ultrasonic imaging array when the wall was compressed using a blood pressure cuff. During the arterial wall motion study, displacement increased with external pressure, which was attributed to a reduction in transluminal pressure on the arterial wall. By reducing transluminal pressure, the external pressure reduced arterial compliance, resulting in increased displacement. This trend was evident so long as the external pressure was less than arterial pressure. As external pressure approaches mean arterial pressure, blood flow is occluded and displacement begins to decrease
In the previous pig study fascial displacement remained elevated over baseline readings even when pressure exceeded 100 mmHg, so that fascial displacement could be used to distinguish elevated pressures (30 mmHg and up) from normal pressures with fair sensitivity 74% and specificity 75%.  In this study, fascial displacement and other ultrasound echo parameters were monitored in human leg model of elevated IMP with respect to subject body mass index and mean arterial pressure. We hypothesized that ultrasonic measurements of fascial displacement can be used to detect elevated IMP in this model compartment syndrome with high sensitivity and specificity.
This study was conducted at the College of William and Mary (Williamsburg, VA) under an IRB-approved protocol, in which IMP was increased in the anterior compartment of the leg in 23 healthy adult subjects (14 males, 9 female with an average weight of 75 +/− 16 kg) by applying one 21-cm wide blood pressure cuff and one 15-cm cuff over the leg, covering the entire length from the tibial tuberosity of the knee to a position proximal to the malleolus flair of the ankle joint. A wide cuff has been previously found to increase IMP uniformly from superficial fascia down to bone [24,25], with IMP approximately equal to cuff pressure . Approximately half the subjects were 18-22 year-old student-athletes while the other half ranged in age from 25 to 70 with varying degrees of physical fitness.
In each subject, a baseline cuff-pressure reading at 0 mmHg was taken, and then cuff pressure was increased from 20 to 80 mmHg in 10 mmHg steps. For each pressure level, the cuff pressure was maintained for 1 minute while ultrasonically monitoring fascial displacement. Fascial displacement was measured in both the cuffed leg and the contralateral control leg using 1 MHz, unfocused ultrasonic transducers, 0.5 inches in diameter. These transducers were placed over the anterior compartment of the lower leg, approximately 4 cm distal to the bottom of the tibial tuberosity and 1-2 cm outside the tibia. In the cuff leg, the cuff was placed over the transducer prior to pressurizing the cuff.
At each pressure setting, small adjustments were made to the transducer position and the receiver depth to obtain a strong echo at a depth around 3-4 cm, which corresponds to the depth of the interosseous membrane and the inner fascia wall of the anterior compartment. Once a strong echo was obtained, the transducer was taped in place and the position of the four largest echoes within the 4-cm sample depth of the ultrasonic receiver was tracked using a digital pulsed phase locked loop (dPPLL) algorithm. The dPPLL is a new implementation of the PPLL that measures changes in the phase of an ultrasonic echo and converts that change into a time-of-flight measurement [27-29]. In this implementation of the dPPLL, provided on the Emergency Noninvasive Tissue and Compartment Tester, or EN-TACT™ (Luna Innovations, Inc.: Roanoke, Virginia), phase measurements are converted into displacement by assuming that ultrasonic velocity is constant. While this assumption has not been conclusively validated, at least one published report backs up this claim .
After setting up the ultrasonic system, displacement and echo backscatter waveforms were obtained at a sample rate equal to the ultrasonic pulse repetition rate of 1 millisecond. Both of these waveforms follow the cardiac cycle, as seen in figure 1. From each of these waveforms, the amplitude and harmonic ratio were extracted once per second, after shifting the starting point of the waveform 1000 samples and performing an auto-correlation routine to filter out non-pulsatile motion artifacts such as those caused by inflation of the blood pressure cuff or muscle fasiculations. After filtering out non-pulsatile waveforms, the fast Fourier transform of the signal was taken, and the amplitude of the fundamental frequency was used to estimate fascial displacement and backscatter amplitude.
While amplitude measurements are most often performed in the time-domain, frequency domain measurements of amplitude correlate well to time-domain measurements of amplitude, and are less noisy . The frequency domain measurement also provides data used in analyzing harmonic content. While numerous waveform analysis techniques have been proposed for measuring harmonic content, for this application the ratio of the fundamental frequency amplitude to root mean squared sum of the next four harmonics was used.
Data were recorded for two minutes at each cuff pressure, yielding a maximum of 120 data points for each pressure. At each cuff pressure, the data point that represented the 90th quantile reading over the two-minute period was used for analysis. The 90th quantile was chosen over the maximum value to eliminate possible outliers due to patient movement or other transient effects that passed through the auto-correlation filter. Use of a median value would also filter outliers from the measurement. However, in many cases the recorded values had a bimodal distribution with multiple echoes were often present and the system sometimes track motion from a tissue boundary that did not correspond to the fascial wall. Use of the 90th quantile reading ensured that the reading was chosen from the fascial wall while eliminating the occasional outlier.
Four ultrasonic parameters were measured: fascial displacement amplitude, displacement harmonic ratio, echo backscatter amplitude and backscatter harmonic ratio. In order to assess the relationship between each of the four ultrasonic parameters measured and cuff pressure, an analysis of variance (ANOVA) followed by contrast analysis was performed using JMP v5.1.1. The multi-factor ANOVA modeled variance for the following values: leg (control vs. cuffed), cuff pressure, body mass index, mean arterial pressure and the interaction between the variables. Next, two ROC curves were constructed to assess clinical diagnostic utility. The first was constructed using a full logistic regression model in which diagnosis was based on the four ultrasonic parameters described above plus subject body mass index and mean arterial pressure. The second ROC curve was constructed using fascial displacement amplitude alone, as this parameter appeared to be most sensitive to changes in pressure. Both ROC curves were constructed using a true positive reading when the cuff pressure was greater than or equal to 30 mmHg. Thus all readings on the control leg were used as readings of normal IMP, as were readings on the cuffed leg taken at cuff pressures of 0 and 20 mmHg.
A diagnostic threshold of 30 mmHg was chosen to be conservative in diagnosis. As mentioned earlier, many clinicians set the threshold at 40 mmHg, while others base diagnosis on perfusion pressure (mean arterial pressure minus IMP) or use a combination of clinical signs and pressure readings. This ambiguity can be incorporated into the diagnostic decision by using recursive partitioning, which divides measured parameters into groups in order to find an optimal diagnostic threshold. This technique is ideal for finding borderline diagnostic groups, in which some readings are neither positive nor negative, but rather in between. In performing recursive partitioning, the diagnostic IMP values were divided into three groups. Ultrasonic readings obtained at cuff pressures of zero mmHg and 20 mmHg, plus all readings on the control leg, were categorized as normal IMP. Readings obtained at cuff pressures of 30 and 40 mmHg were categorized as mid-range IMP, and readings obtained at cuff pressures 50 mmHg and up were categorized as elevated IMP.
For the purpose of comparison to the ROC curves, sensitivity measurements were obtained from the recursive partitioning data using a ratio of true positive (TP) readings to the sum of true positives and false negatives (FN) as follows:
Similarly, specificity was determine using a ratio of true negative (TN) readings to the sum of true negatives and false positives (FP):
In the recursive partitioning analysis, mid-range diagnostic readings introduced some ambiguity as to whether a reading was a true/false positive or negative. We have attempted to resolve this ambiguity by providing two sets of sensitivity and specificity values for the recursive partitioning data. In the first set of calculations, all mid-range displacement values as a true positive when the cuff pressure was 30 or 40 mmHg, while mid-range displacements were considered false positives or false negatives if the cuff pressure was either high or low. Case 2 provides an alternative scenario, where it was assumed that the use of continuous monitoring and the evaluation of clinical signs would allow a clinician to arrive at a correct diagnosis over time. The only exception was for a low displacement reading at a pressure of 30 or 40 mmHg, in which case a clinician might incorrectly assume that further monitoring was not necessary. This was considered a false negative.
The ANOVA results show that each of the four ultrasonic parameters exhibited significant increases in the model compartment syndrome leg versus the control leg (figure 2) once pressure exceeded 50 mmHg. All four ultrasonic parameters increased with pressure with fascial displacement and echo backscatter amplitude exhibiting a strong, nonlinear increase at high pressures. Displacement harmonic ratio and backscatter harmonic ratio exhibited weaker linear increases.
Other sources of variation, including the subjects' body mass index and mean arterial pressure are shown in the multi-factor ANOVA results provided in table 1. The ANOVA model, which included higher order interactions between variables, shows that cuff pressure, leg and the interaction between these two variables were the most significant sources of variation in each of the four measured parameters. Body mass index was a significant source of variation in the displacement harmonic ratio only, while the higher order interactions between body mass index, leg and cuff pressure were significant sources of variation in the displacement amplitude.
Figure 3 provides two ROC curves with the diagnostic threshold set at a cuff pressure of 30 mmHg along with results from the recursive partitioning test. The first curve, formed using a full logistic regression of all parameters, provides only a marginal improvement in the diagnosis of elevated IMP over the use of displacement amplitude alone in the second curve. The sensitivity and specificity readings for the ROC curves and the recursive partitioning tests are summarized in table 2.
Our results are consistent with the hypothesized relationship between fascial displacement and elevated IMP. While the mean values for each of the ultrasonic parameters were not significantly elevated over control values until cuff pressure reached 60 mmHg, this may have been due to a few outliers, as the sensitivity and specificity readings of table 2 are obtained with a diagnostic threshold of 30 mmHg and up a true positive reading.
The sensitivity and specificity values given in table 2 are comparable to those reported for clinical signs reported by Ulmer  and invasive IMP measurements reported by Dickson and colleagues . However, a more direct comparison to gold-standard invasive IMP measurements in a clinical setting is still needed.
For the subjects enrolled in this study, the sensitivity and specificity does not appear to depend on the physiological characteristics such as blood pressure, age or body mass index. However, it should be noted that in this study all subjects had a blood pressure in the normal range. In a few of the subjects with a lower blood pressure reading, the tissue displacement was observed to decrease when IMP approached mean arterial pressure, as illustrated the results for two subjects in figure 4. This is consistent with the results from a previous study performed on a porcine model , in which tissue displacement was observed to first increase with IMP until IMP approached mean arterial pressure. At that point, tissue displacement appears to decrease. Despite this tissue-displacement decrease in the porcine model, pressure remained elevated over baseline values even when IMP = 100 mmHg. The nonlinear response of tissue displacement was more pronounced in the porcine model  than in this study, probably because the human subjects had a higher mean arterial pressure (96 mmHg +/− 11 mmHg) than that in the pigs (81 +/− 12 mmHg) and because the maximum IMP studied in humans was limited to 80 mmHg due to safety concerns.
Based on this observation, ultrasonic measurements of tissue displacement would be a less sensitive diagnostic tool in hypotensive subjects, particularly when IMP approaches mean arterial pressure. However, in subjects with a normal blood pressure, the clinical signs of a compartment syndrome are quite pronounced at the high IMP values needed to reduce fascial displacement below baseline values. However, further study is needed to determine whether clinical signs would be similarly obvious in the case of a severely hypotensive subject with high IMP.
The close relationship between fascial displacement and echo backscatter seen in figure 2 suggests that the two parameters are closely linked, most likely because changes in fascial geometry during tissue motion affect the amount of reflected ultrasonic signal. Thus, it appears that the echo backscatter waveform provides no additional diagnostic information over the fascial displacement waveform alone. In addition, the harmonic ratio readings of figure 2 follow a similar trend to the amplitude readings, and do not appear to provide additional diagnostic information beyond that already provided by displacement amplitude alone. The two ROC curves of figure 3 are consistent with this conclusion, as the full logistic regression model provided only marginal improvement in diagnostic accuracy over the ROC curve formed using fascial displacement alone.
Based on these results, noninvasive measurement of pulsatile tissue motion using ultrasound appears to have comparable diagnostic sensitivity and specificity to invasive measurements and clinical signs. However, this model study has important limitations as it is not clear to what extent the low sensitivity is due to errors in the cuff model or due to errors in the measurement itself. A more controlled study, conducted using gold-standard invasive measurements of intramuscular pressure would provide more definitive results in this regard. Ideally, this testing would be performed in a clinical setting, in which noninvasive measurements are performed in conjunction with invasive measurements of IMP and with diagnosis by clinical signs.
This study was funded by Luna Innovations Incorporated.