TY - JOUR
AU - MS, Kevin Ho,
AB - ABSTRACT Objectives: We present a method to account for the effects of the hearing protection devices (HPDs) for use with the 8 hours equivalent A-weighted energy (LAeq8) criterion. The method involves the calculation of the LAeq8 equivalent unprotected free-field dose (LAeq8EUFF), which is obtained by using the insertion loss (IL) data of the HPD together with free-field pressure measurements. Methods: The method was validated against the historical the U.S. Army Medical Research and Materiel Command walk-up study data with volunteers exposed to simulated large weapon noise wearing a range of HPDs. The IL data were obtained using standard acoustical test fixtures fitted with the matching HPDs in replicated field tests and using shock tubes at conditions comparable to the actual exposure intensities. Logistic regression calculations were performed to correlate the LAeq8EUFF values against the walk-up study outcomes to determine the L(95,95) threshold for the protection of 95% of the population with 95% of the time. Results: Data comparison shows that L(95,95) is 89 dBA, which is slightly higher than the 85 dBA criterion but falls in the 80 to 90 dBA range as used by various NATO nations. Conclusions: Therefore, considering the limitation of the walk-up dataset, it is conservative to adopt the 85 dBA threshold for general application. INTRODUCTION Hearing loss (HL) is one of the most disabling conditions for both the military and civilians. According to the U.S. Department of Veterans Affairs, 21% of nearly 7 million cases contain the pathological outcomes of auditory damage.1 The World Health Organization has also pointed out the significance of HL with a staggering 5.3% of the world population suffering from a variety of causes including noise exposures. Noise-induced HL remains one of the top military medical problems. Soldiers are routinely exposed to high-level impulse noise from large weapons during training exercises or on the battlefield that causes severe and permanent hearing damage. Impulse noise injury can be more effectively prevented if an adequate medical standard can be provided to set the appropriate limit on exposure levels. This task can be achieved if a representive injury dataset from impulse noise exposures is available for modeling. As part of the Blast Overpressure Project, the U.S. Army Medical Research and Materiel Command conducted a test series at the blast test site at Kirkland Air Force Base in Albuquerque, New Mexico, commonly known as the “Walk-up” or “Albuquerque” study. The historical human walk-up study collected valuable data but did not provide a new standard.2 The data include the waveforms of the simulated large weapon noise for several distances (1, 3, and 5 m) from the noise source and the resulting human temporary threshold shift (TTS) data for five types of hearing protection devices (HPDs) worn during the study. The HPDs included the regular (unmodified) and modified earmuffs, the French, Rucker, and Perforated earplugs. The use of the modified muff was to simulate improper fitting of the earmuff that was considered a common (unintentional) occurrence. Using the modified earmuff data alone, Chan et al3 performed statistical analysis that confirmed the deficiency of various noise standards. The data predicted that the current standards overestimate the injury threshold by about 10 dB. The data also showed that the 8 hours equivalent A-weighted energy (LAeq8) model4 provides a better fit compared to the peak-duration based standards. The analysis, however, was based on the crude approximation of 15 dB attenuation for the modified earmuffs due to the lack of insertion loss (IL) data. The American Institute of Biological Sciences recommended the use of the LAeq8 model as an interim standard until a robust biomechanically based auditory standard is found.5 This choice was motivated by the extensive research conducted on the LAeq8 model and its popular use across nations. A system acquisition standard was recently adopted by the military, MIL-STD-1474E, which, however, is not a medical standard. MIL-STD-1474E includes a variant of the LAeq8 that has not been validated and the current auditory hazard assessment algorithm for the human that is still under rigorous investigation for validation and improvement. In order to use the LAeq8 standard as a general procedure, a method to account for HPDs is needed. The LAeq8 standard limits the daily unprotected noise exposure to 85 dBA, but this threshold was established based on free-field pressure measured at the ear position without HPDs and in continuous noise conditions. For intense impulse noise for operational application, HPDs are required. In this case, the use of the free-field pressure data is not appropriate for correlation with injury because the free-field data does not account for the effects of the HPDs, which can become nonlinear at high-exposure levels.6 For this reason, the use of a constant attenuation is not adequate. A procedure is, therefore, needed to account for the effects of HPDs and to set the threshold. In this article, we present an injury correlation based on the LAeq8 method that includes a test procedure to incorporate the HPDs using IL data derived from acoustical test fixture (ATF) measurements. The LAeq8-based correlation method is validated against the human walk-up study data. METHODS The method consists of (1) establishing an ATF method to characterize the HPDs, (2) incorporating the HPD characteristics to calculate an exposure dose based on the LAeq8 model, and (3) performing statistical analysis to correlate the dose to the human injury data to predict the L(95,95) threshold for the protection of 95% of the population, 95% of the time against auditory TTS exceeding 25 dB. ATF Method for HPDs Characterization Shock Tube Evaluation of the ATF An ATF suitable for high-impulse noise was selected and its response was verified against benchmark data. The G.R.A.S. 45CB (G.R.A.S. Sound & Vibration, Twinsburg, Ohio) was selected for this study based on microphone specs, the maximum level permissible of 174 dB at the eardrum, and the ability to handle both earplugs and earmuffs. The G.R.A.S. 45CB ATF is fully ANSI S12.42 compliant. During the tests, the pinna/ear canal material was heated to 37°C in accordance to the ANSI S12.42 requirement. The transfer function of the open-ear (TFOE) was measured using shock tube experiments to evaluate the ATFs for impulse noise exposure (Fig. 1). The results were compared to the human TFOE data from Shaw and Vaillancourt,7 obtained with continuous noise exposures. FIGURE 1. View largeDownload slide Shock tube test setup for acoustical test fixture (ATF) evaluation. FIGURE 1. View largeDownload slide Shock tube test setup for acoustical test fixture (ATF) evaluation. Replication of Field Tests to Collect ATF Eardrum Data The human walk-up tests were replicated to collect the ATF eardrum data at matching conditions. To quantitatively compare the original and the replicated free-field waveforms, matching criteria based on the waveform characteristics were defined. The criteria were that when the peak level, A-duration, impulse, and A-weighted sound exposure level (SELA) of the replicated waveform lie simultaneously within one standard deviation of the original blast overpressure (BOP) waveform, the two waveforms are considered matched. Thus, a series of calibration tests were performed to search for the distance from the blast center that would satisfy the matching criteria for each test condition designated by the (nominal) distance and level. With the respective optimal distances determined, the ATF fitted with HPDs and the pencil gauges were placed at these distances from the blast, and the eardrum data were collected. Six to seven shots per condition were measured for each distance and blast level. Figures 2 and 3 show the test setup for the 1, 3, and 5 m tests, respectively. The recovered HPDs are shown in Figure 4, and Figure 5 shows the ATF with the recovered earmuff and other protective gear worn in the walk-up study. FIGURE 2. View largeDownload slide One and three meter setup. Two acoustical test fixtures (ATFs) fitted with hearing protective devices, helmet, and goggle are shown with the right ear facing the explosion, which originates from underground guided by a center tube. Pencil gauges shown are placed at the level of the ATF's ear at the same distance from the blast. FIGURE 2. View largeDownload slide One and three meter setup. Two acoustical test fixtures (ATFs) fitted with hearing protective devices, helmet, and goggle are shown with the right ear facing the explosion, which originates from underground guided by a center tube. Pencil gauges shown are placed at the level of the ATF's ear at the same distance from the blast. FIGURE 3. View largeDownload slide Five meter test setup. Two acoustical test fixtures fitted with the hearing protective device, helmet, and goggle are shown paired with pencil gauges. The explosive charge is hung above the ground between three tall cylindrical protection tubes. FIGURE 3. View largeDownload slide Five meter test setup. Two acoustical test fixtures fitted with the hearing protective device, helmet, and goggle are shown paired with pencil gauges. The explosive charge is hung above the ground between three tall cylindrical protection tubes. FIGURE 4. View largeDownload slide Recovered hearing protective devices. (A) Unmodified earmuff, (B) modified earmuff with eight tubes inserted in the seal, (C) French No. 1 earplug with insertion filter (black), (D) Rucker earplug with insertion filter (transparent), and (E) perforated earplug without filter inserted in the hole. FIGURE 4. View largeDownload slide Recovered hearing protective devices. (A) Unmodified earmuff, (B) modified earmuff with eight tubes inserted in the seal, (C) French No. 1 earplug with insertion filter (black), (D) Rucker earplug with insertion filter (transparent), and (E) perforated earplug without filter inserted in the hole. FIGURE 5. View largeDownload slide Acoustical test fixture equipped with earmuff, helmet, and goggle. FIGURE 5. View largeDownload slide Acoustical test fixture equipped with earmuff, helmet, and goggle. Characterization of HPDs The HPD was characterized by its IL, defined as the difference between the pressures recorded at the eardrum without and with the HPD for the same free-field pressure:   (1) where denotes the pressure measured under the HPD at the location of the eardrum. The IL was calculated using the TFOE and the ATF eardrum data, as in Equation 1, for all five HPDs used during the human walk-up study. For the free-field peak pressure level greater than 180 dB as observed above Level 4 in the walk-up study, eardrum measurements could not be made using the ATF due to microphone saturation. The ILs for the high-intensity levels were obtained by extrapolating from the ILs at lower levels. LAeq8 Dose Calculation With HPD Using the IL, the LAeq8 equivalent unprotected free field (LAeq8EUFF) dose with the HPD worn is defined as:   (2) where the summation is based on 1/3-octave band frequency (k) intervals, F(k) is the A-weighting correction factor for the 1/3-octave band frequency interval, T8h is the 8-hour conversion factor, and the last term represents the traditional dosage accumulation rule with N being the number of shots. LAeq8 Correlation With Human Walk-Up Test Data The logistic regression fit to the BOP injury data with the LAeq8EUFF dose as the predictor variable was used to validate the LAeq8 method. The calculations were carried out for the BOP HPD series to determine the threshold dosage for 25 dB TTS. In the present analysis, all human BOP data collected in the free field were used, which included the data from the modified and unmodified earmuff and earplug test series. Data from the 1, 3, and 5 m series were pooled together for LAeq8 correlation with the injury data. The statistical analysis procedure is identical to that in Chan et al.3 The population average logistic regression model8 is used to correlate the human injury data with LAeq8EUFF to establish the threshold for TTS2 ≥ 25 dB at 2 minutes after exposure. The predictor variable x is the LAeq8EUFF dose and the response variable y is 1 when injury (TTS ≥ 25 dB) occurs and 0 when no injury occurs. The logit function g(x) is thus given in terms of the conditional probability of injury given a LAeq8EUFF dose, π (y = 1 | x) as:   (2) The logistic regression model coefficients a and b were estimated by fitting the model with the pooled human injury response and the associated LAeq8EUFF dose for all test cases using the STATA software (StataCorp, Stata: Release 11. Statistical Software, 2009, StataCorp LP: College Station, Texas). The (1 − α)100% confidence interval (CI) for the estimated probability of injury is calculated from the mean estimates and covariance matrix Covab (i, j) of the regression coefficients and by assuming normal distribution of the error of the regression. The formulas for the CIs for the estimated probability of injury, logit, and standard error of the logit are summarized in Equation 4 to 6.   (4)  (5)  (6) In Equation 6, zx satisfies the cumulative standard normal distribution function, ϕ(zx) = x and z0.975 = 1.96 corresponds to 95% confidence in the injury prediction. The LAeq8EUFF dose threshold for the protection of 95% of the population with 95% confidence, L(95,95) is given by the upper bound of the CI with probability of injury of 5%, i.e., π+ = 0.05. RESULTS G.R.A.S ATF Evaluation The G.R.A.S ATF TFOE measurements show comparable results with the TFOE measured from human subjects. Figures 6 and 7 show the TFOE comparison result of ATF vs. human for normal and grazing noise exposures, respectively. Figure 6 shows that the TFOE generated from shock tube testing at normal incidence at 163 dB peak pressure level (PPL) is fairly close to the historical curve from Shaw and Vaillancourt7 obtained from human subjects at low sound levels. Similarly, Figure 7 shows that at grazing incidence, the TFOE obtained from shock tube test at 164 dB PPL compares closely to that from humans. The similarity between the TFOE obtained from the ATF and humans appear to be independent of direction for the highest intensity of ∼164 dB PPL tested, but slightly dependent on the impulse intensity level. FIGURE 6. View largeDownload slide Comparison of transfer function of the open-ear (TFOE) for the acoustical test fixture left and right ears from shock tube measurements and the historical TFOE data from human exposures with continuous noise for normal incidence noise exposures. FIGURE 6. View largeDownload slide Comparison of transfer function of the open-ear (TFOE) for the acoustical test fixture left and right ears from shock tube measurements and the historical TFOE data from human exposures with continuous noise for normal incidence noise exposures. FIGURE 7. View largeDownload slide Comparison of transfer function of the open-ear (TFOE) for the acoustical test fixture left and right ears from shock tube measurements and the historical TFOE data from human exposures with continuous noise for grazing noise exposures. FIGURE 7. View largeDownload slide Comparison of transfer function of the open-ear (TFOE) for the acoustical test fixture left and right ears from shock tube measurements and the historical TFOE data from human exposures with continuous noise for grazing noise exposures. ATF Eardrum Data Comparison Figure 8 shows the results of the statistical comparison between the ATF eardrum and the undermuff waveforms from the walk-up tests for the peak pressure and the SELA as a function of the blast level. The peak and SELA increase linearly for the ATF eardrum data, and somewhat less linearly for the human undermuff data toward the high levels. The difference between the mean peak and SELA values is ∼5 to 7 dB, which represents the amplification from the ear canal entrance to the eardrum. It is noted that the ATF eardrum data show very little variability compared to the undermuff data. FIGURE 8. View largeDownload slide Comparison of acoustical test fixture (ATF) eardrum waveforms and the human blast overpressure (BOP) undermuff waveforms for increasing exposure intensity level-3 m test. (A) Peak pressure comparison and (B) A-weighted sound exposure level (SELA) comparison. FIGURE 8. View largeDownload slide Comparison of acoustical test fixture (ATF) eardrum waveforms and the human blast overpressure (BOP) undermuff waveforms for increasing exposure intensity level-3 m test. (A) Peak pressure comparison and (B) A-weighted sound exposure level (SELA) comparison. IL Comparison With Real-Ear Attenuation at Threshold (REAT) The IL results are shown in Figures 9 and 10 for the earmuffs and earplugs, respectively, with comparison to the REAT data. In Figure 9, the IL data of the earmuffs somewhat overlap with the REAT data. However, depending on the level, differences between the IL and the REAT as high as 10 dB can be observed. The overall trend indicates that the IL of the earmuffs is closer to the REAT than that for the earplugs. As shown in Figure 10, there are indeed large differences between the IL of the earplugs and the REAT. For the French no. 1 plug, the difference is as high as 25 dB; and for the Rucker and the perforated plug, the differences are less pronounced, but at least 10 dB. The results indicate that REAT should not be used as IL. FIGURE 9. View largeDownload slide Insertion Loss vs. Real-Ear-Attenuation-at-Threshold (REAT) for earmuffs. (A) l m test with modified earmuff, (B) 3 m test with modified earmuff, (C) 5 m test with modified earmuff, and (D) 5 m test with unmodified earmuff. FIGURE 9. View largeDownload slide Insertion Loss vs. Real-Ear-Attenuation-at-Threshold (REAT) for earmuffs. (A) l m test with modified earmuff, (B) 3 m test with modified earmuff, (C) 5 m test with modified earmuff, and (D) 5 m test with unmodified earmuff. FIGURE 10. View largeDownload slide Insertion Loss vs. Real-Ear-Attenuation-at-Threshold (REAT) for earplugs. (A) 3 m test with French No. 1 earplug, (B) 3 m test with Rucker earplug, and (C) 3 m test with perforated earplug. FIGURE 10. View largeDownload slide Insertion Loss vs. Real-Ear-Attenuation-at-Threshold (REAT) for earplugs. (A) 3 m test with French No. 1 earplug, (B) 3 m test with Rucker earplug, and (C) 3 m test with perforated earplug. Interim LAeq8 Validation Figure 11 shows the result of the logistic regression calculation with injury data from all HPDs when the IL derived from the historical TFOE are used in the LAeq8EUFF dose calculation. The L(95,95) threshold predicted is 89 dBA as shown by the large, open square on the upper bound of the CI. This value is 4 dB higher than the standard threshold of 85 dBA shown by the vertical bar. The data points shown represent the mean failure rates based on 10-bin data grouping, and each short horizontal bar is the standard deviation of the dose for each data bin. The data shows that all injuries occur to the right of the vertical bar, i.e., above the 85 dBA threshold. FIGURE 11. View largeDownload slide Logistic regression fit for LAeq8EUFF dose calculated: from the acoustical test fixture data with all blast overpressure hearing protection device data included. FIGURE 11. View largeDownload slide Logistic regression fit for LAeq8EUFF dose calculated: from the acoustical test fixture data with all blast overpressure hearing protection device data included. DISCUSSION The TFOE obtained from the historical human tests7 was used in the IL calculation. As shown in the evaluation results, the TFOE compared fairly well with that obtained from the shock tube measurements with an ATF. In addition, the effects on the LAeq8EUFF dose calculation of the differences in the TFOEs were verified and produced a dose variation of less than 1 dB. Therefore, the TFOE measured either with an ATF or human can be used for the LAeq8 dose calculation. Three factors that may affect the predictions are the effects of the head orientation, bone conduction (BC) and the ILs for high-intensity levels. These factors are discussed in turn. The orientation effects on IL measurements and prediction should be considered. However, although these effects on IL measurements can be readily assessed using an ATF with the HPD, it is difficult to evaluate the significance on the threshold prediction. The main reason is that the available injury data do not cover all orientations. Animal studies may help guide the inclusion of the orientation effects. The effects of BC on the IL and LAeq8 prediction were considered to be minimal and BC correction was not made. Since the ATF does not model the BC effect, an adjustment of the IL measured with the ATF is commonly made based on the BC limits. For impulse noise, however, this sort of correction appears not to change the ILs significantly.9 A verification of the extrapolated IL results is desirable. A high-pressure microphone sensor installed in the ATF may be used for this purpose. Our results show that the extrapolated IL values for the high-intensity follow the data trend well, and we do not expect the measured IL data to deviate significantly from their current values. Therefore we do not expect the verification results to significantly affect the current LAeq8 predictions. The L(95, 95) threshold prediction by Chan et al3 with their LAeq8 analysis is ∼25 dB higher than that found in the present analysis. As previously noted, the former analysis did not include the earplugs, which caused the most injuries, nor did it include the unmodified earmuff data, which represent the largest dataset and contain the least injuries. In addition, the analysis by Chan et al3 did not use adequate attenuation data for the earmuff. In the present analysis, using the traditional dose accumulation according to the 10log(N) rule as used in Chan et al3, all BOP HPD data were pooled together to predict the injury threshold. The IL for all HPDs used was measured under real life exposure conditions using an ATF. All these factors explain the difference between the threshold predictions. Although the use of an ATF constitutes the most practical method for characterization of HPDs, it is recognized that the method does not account for human factors such as ergonomics, HPD fit and fitting methodology and thus prevents systematic generalization of the ATF results to real world performance. Fitting the HPD to the ATF is equivalent to the investigator fitting the HPD with human subjects during testing, where all the necessary precautions are taken by the inverstigor to ensure proper fit of the HPD. In a real world, however, the subjects wear the HPD themselves, and fitting inconsistency is thus unavoidable. HPD misfit is the predominant factor that can cause the ILs as measured with the ATF to deviate from their performance during operations. The earmuffs can easily be moved during operations, whereas the insertion depth of the earplug can vary from one human subject to another. The variability in these human factors can have a pronounced influence on HPD. On the other hand, no one method for measuring IL is perfectly accurate. The used of REAT is not adequate for high-level impulse, even when corrected for physiological noise masking and BC; this is probably true at least for the earplugs used in the BOP walk-up study as shown in the results. It is recognized that the human BOP walk-up study data is still limited. The data simulate a limited class of blasts although the inclusion of additional weapon noise types would certainly improve the predictions. Although representing both earmuffs and earplugs, double protection is not simulated, and the HPDs used in the study may not represent modern HPD types. Modern earplugs or earmuffs appear to show different attenuation characteristics even among the same category when measured at relatively low levels.10 The E.A.R. classic foam, E.A.Rsoft Yellow Neons foam, E.A.R. Push-Ins pod-style, and the E.A.R Ultrafit premolded earplugs (3M, St. Paul, Minnesota) would give attenuation characteristics that are different by 5 to 10 dB across frequencies. Likewise, an E.A.R Muff Model 1000, Peltor Optime 105 earmuff, and Peltor Optime 95 earmuffs (3M, St. Paul, Minnesota) would differ by 5 to 10 dB in attenuation. However, the response of these HPDs as measured with the same ATF but at high-level impulse conditions are not known. Another limitation is that the injury data result from normally incident blast conditions only. These conditions are clearly different from the operational conditions. For these various reasons, the use of the LAeq8-based correlation method that is not ambitiously relaxed with the risk of causing injury such as the LAeq8 method presented in this article is appropriate. CONCLUSION An LAeq8 standard is presented that includes the use of the ATF to account for the effects of the HPDs by measuring attenuation under realistic exposure conditions. The head orientation effects are captured by the ATF and the intensity of the exposure can be matched. The ATF method shows that the 85 dBA threshold is valid for large weapon noise as demonstrated using the most complete and scientifically relevant injury dataset available to date. The method provides a major step forward to harmonize with environmental noise and unprotected rifle noise criteria and should be validated against small arms. ACKNOWLEDGMENT This work was funded through the U.S. Army Medical Research and Materiel Command (MRMC) under contract W81XWH-11-D-0011. REFERENCES 1. U.S. Department of Veterans Affairs Annual Benefits Report, Fiscal Year 2010, 2011 , Department of Veterans Affairs. Available at http://www.benefits.va.gov/REPORTS/abr/2010_abr.pdf; accessed July 7, 2015. 2. Johnson DL Blast Overpressure Studies with Animals and Man: A Walk-Up Study . EG & G Special Projects, 1993. Available at http://www.dtic.mil/dtic/tr/fulltext/u2/a280240.pdf; accessed January 19, 2015. 3. Chan PC, Ho KH, Kan KK, Stuhmiller JH, Mayorga MA Evaluation of impulse noise criteria using human volunteer data. J Acoust Soc Am  2001; 110( 4): 1967– 75. Google Scholar CrossRef Search ADS PubMed  4. Dancer AL, Franke RP Hearing hazard from impulse noise: a comparative study of two classical criteria for weapon noises (Pfander criterion and Smoorenburg criterion) and the LAeq8 Method. Acta Acust  1995; 3: 539– 47. 5. Wightman FL, Flamme G, Campanella AJ, et al.   AIBS peer review of injury prevention and reduction research task area impulse noise models . American Institute of Biological Sciences, 2010. Available at https://www.arl.army.mil/www/pages/343/AHAAH_AIBS_revew_Public_Release_11Aug14.pdf; accessed March 2, 2016. 6. Price GR Validation of the auditory hazard assessment algorithm for the human with impulse noise data. J Acoust Soc Am  2007; 122( 5): 2786– 802. Google Scholar CrossRef Search ADS PubMed  7. Shaw EAG, Vaillancourt MM Transformation of sound-pressure level from the free field to the eardrum presented in numerical form. J Acoust Soc Am  1985; 78( 3): 1120– 3. Google Scholar CrossRef Search ADS PubMed  8. Zeger SL, Liang KY Longitudinal data analysis for discrete and continuous outcomes. Biometrics  1986; 42( 1): 121– 30. Google Scholar CrossRef Search ADS PubMed  9. Khan A, Fackler CJ, Murphy WJ Comparison of Two Acoustic Test Fixtures for Measurement of Impulse Peak Insertion Loss. In-Depth Survey Report . Department of Health and Human Services, 2013. Available at http://www.cdc.gov/niosh/surveyreports/pdfs/350-13a.pdf; accessed January 19, 2015. 10. Berger E, Kieper R, Stergar M Insersion-loss and transfer function performance of two new acoustical test fixtures complying with ANSI S12.42-2010, relative to performance of prior test fixtures and to real-ear data. Indianapolis Acoustical Society of America Presentation (Session 3aNS4); November 2, 2011, San Diego, CA. Available at https://acousticalsociety.org/sites/default/files/162_full_week.pdf; accessed March 2, 2016. Footnotes 1 This work was presented at the Military Health System Research Symposium, Fort Lauderdale, FL, August 18–21, 2014. Reprint & Copyright © Association of Military Surgeons of the U.S.
TI - An Interim LAeq8 Criterion for Impulse Noise Injury
JF - Military Medicine
DO - 10.7205/MILMED-D-15-00185
DA - 2016-05-01
UR - https://www.deepdyve.com/lp/oxford-university-press/an-interim-laeq8-criterion-for-impulse-noise-injury-BdhZddBqtw
SP - 51
EP - 58
VL - 181
IS - suppl_5
DP - DeepDyve
ER -