This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Secomski, W. Articles by Lewin, P. A. Search for Related Content PubMed PubMed Citation Articles by Secomski, W. Articles by Lewin, P. A.

Noninvasive In Vivo Measurements of Hematocrit

Institute of Fundamental Technological Research, Polish Academy of Sciences, Warsaw, Poland (W.S., A.N.); Electronics and Telecommunications Department, University of Florence, Florence, Italy (F.G., P.T.); and School of Biomedical Engineering, Science, and Health Systems and Department of Electrical and Computer Engineering, Drexel University, Philadelphia, Pennsylvania USA (P.A.L.).

Address correspondence and reprint requests to Wojciech Secomski, Institute of Fundamental Technological Research, Polish Academy of Sciences, Swietokrzyska 21, Warsaw 00-049, Poland.

	Abstract

Top Abstract Introduction Materials and Methods Results Discussion References

 
Objective. To develop a clinically applicable method for noninvasive acoustic determination of hematocrit values in vivo. Methods. The value of hematocrit was determined initially in vitro from the pulse-echo measurements of acoustic attenuation. The testing was carried out in a laboratory setup with an ultrasonic transducer operating at 20 MHz and with the use of human blood samples at 37°C. The attenuation coefficient measurements in blood in vivo were implemented by multigated, 20-MHz pulsed Doppler insonation. The Doppler signal was recorded in the brachial and radial arteries. Both in vitro and in vivo hematocrit data were compared with those obtained by the centrifuge method. Results. The attenuation coefficient in vitro was determined from the measurements of 168 samples with hematocrit values varying between 23.9% and 51.6%. The attenuation from 20-MHz data was equal to 3.66 + 0.089 · hematocrit (decibels per centimeter). The uncertainty of in vivo measurements in the brachial artery was determined to be within ±5% hematocrit. However, the measurements in the radial artery resulted in a clinically unacceptable uncertainty of ±20% hematocrit. Conclusions. The method proposed appears to be promising for in vivo determination of hematocrit, because 5% hematocrit error is adequate for monitoring changes in patients in shock or during dialysis. It was found that the multigate system largely simplified placement of an ultrasonic probing beam in the center of the blood vessel. Current work focuses on enhancing the method’s applicability to arbitrarily selected vessels and to reducing the hematocrit measurement error to much less than 5% hematocrit.

Key Words: blood • Doppler ultrasound • hematocrit • multigate Doppler ultrasound • power Doppler ultrasound

Abbreviations: DPP, Doppler power profile • HCT, hematocrit

	Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

 
Noninvasive monitoring of hematocrit (HCT) is of importance in several clinical applications. Constant monitoring of HCT is essential during dialysis procedures in which microfiltration of the blood can lead to inadvertent blood volume reduction and thus to an increase of HCT. Rapid determination of HCT is also essential in emergency departments and during open heart surgery.

Hematocrit is defined as the ratio of the volume of packed red blood cells to the volume of whole blood, and ultrasound-based continuous HCT measurement was already proposed by Johner et al.¹ They determined the value of HCT by monitoring changes of ultrasonic wave velocity propagation in plasma as a function of red blood cells. Although they reported very good agreement (r > 0.96; SD, 1.7%) comparing their results with those determined by conventional, centrifuge-based HCT measurement, they noted that the uncertainty of the method depended markedly on even minute temperature variations. At 37°C a 10% increase in HCT caused merely a 0.7% to 0.8% change in the wave velocity.² To overcome this problem, Johner and coworkers¹ had to use a temperature sensor that was capable of measuring the temperature to within 0.1°C and had to design and include a special algorithm in their data processing. Hughes et al³ carried out ultrasonic velocity measurements in whole blood at 10 MHz using a pulse-echo method and noted a "parabolic" minimum at 8% HCT that led to ambiguity in the HCT for values between 0% and about 16%.

In contrast to the quasi-invasive or in vitro ultrasonic velocity-based HCT measurements, the method proposed here is capable of determining HCT values noninvasively in vivo by using Doppler measurements. More specifically, HCT is determined as a function of ultrasonic wave attenuation in blood, whereas the attenuation coefficient is calculated from the Doppler power spectrum.

In this report, the principle of the novel HCT meter is described, including initial in vitro verification of the method and final in vivo implementation of the system operating at 20 MHz. The results of the attenuation coefficient measurements in animal and human blood both in vivo and in vitro are presented, together with the corresponding values of HCT, and the overall uncertainty obtained in the HCT measurements is discussed.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

 
Below, 2 basic measurement methods used in the development of the noninvasive Doppler HCT meter are described. Both methods were initially used to determine animal and human blood HCT in vitro. On the basis of the results of these measurements, the Doppler HCT meter for in vivo clinical applications was designed, and its operation principles are also outlined below.

Pulse-Echo and Transmission Methods
In Vitro Animal HCT Measurements
In the first method used, ultrasonic attenuation was determined in vitro in animal blood by the transmission measurement approach detailed by Carstensen et al.² Yuan et al⁴ reported that acoustic properties of animal blood, including scattering, are very similar to those of human blood. This is consistent with results published by Vandegriff et al,⁵ who found that animal and human blood cells have almost identical biochemical properties, physical dimensions, and shapes. Their findings were also confirmed by Standl et al,⁶ who proposed the use of bovine blood as an oxygen carrier in humans.

Briefly, a 2-mL cylindrical blood sample container filled with porcine blood was placed between 2 unfocused 20-MHz ultrasonic transducers similar to those used in the final noninvasive Doppler HCT meter described below. The transducers were positioned at a well-known distance; 1 transducer was operated as a transmitter, and the other was operated as a receiver. The transmitter was excited by an 8-cycle tone burst with an amplitude of 36 V peak to peak. (These excitation conditions were also used later in the noninvasive Doppler HCT meter described below.) The amplitude of the received signal was expected to depend on attenuation of the acoustic energy in blood and was first measured in distilled water to determine the reference level. The subsequent measurements were carried out in assays with HCT values ranging from 1% (plasma) to 65% (centrifuged red blood cells). A commercially available hematology analyzer (K-4500; Sysmex Corporation of America, Long Grove, IL) having measurement accuracy of approximately 1% was used to determine the reference level of HCT. The centrifuge-determined HCT values were used to calculate the correlation coefficient. A slightly modified method using 1 transducer operating in the pulse-echo mode was also used. In this method, the attenuation was determined at 37°C from the amplitude of the echoes reflected from either the bottom of the cylindrical container or a 3-mm-diameter stainless steel reflector positioned at 9.6 mm in the near field of the 3-mm-diameter transducer.

In Vitro Human HCT Measurements
These measurements were performed in the 2-mL pulse-echo chamber described above. Human blood was drawn from 168 healthy volunteers: 30 women and 138 men; age range, 19 to 54 years; their HCT values ranged from 23.9% to 51.6%. To prevent coagulation K₃EDTA salt (number 366601; Becton, Dickinson and Company, Franklin Lakes, NJ) was added. Samples of plasma were prepared using a serum separator tube (number 368986, Becton, Dickinson and Company). Reference HCT measurements were carried out on 168 different samples and 5 plasma samples with the Sysmex K-4500 automatic HCT analyzer.

The arrangement for in vitro human HCT measurements is shown in Figure 1. The measurements were carried out 4 to 8 hours after the blood was drawn. The samples were heated up to 37°C ± 0.5°C and poured over to the measurement chamber shown in Figure 1. The blood was constantly stirred to ensure a uniform distribution of red blood cells. Before each blood sample attenuation measurement, the reference measurements in distilled water were carried out. These measurements were performed in the near field of the transducer.

View larger version (13K): [in this window] [in a new window]   Figure 1. Arrangement for in vitro measurement of ultrasonic attenuation in blood by the pulse-echo method.

Doppler Method
In the second method, the attenuation was measured in vitro from the spectrum of the Doppler signal. The porcine blood moving at constant velocity in a flow phantom was investigated. In Figure 2, the initial measurement system is shown. It used a 2-gate pulsed Doppler ultrasonic system operating at the frequency of 20 MHz to maximize ultrasonic attenuation. During these studies, the Doppler gates were positioned symmetrically around the axis of the plastic tube filled with moving blood. As discussed below, this symmetry requirement limited the usefulness of a simple 2-gate Doppler system and prompted the development of a clinically acceptable HCT meter, which is described below.

View larger version (19K): [in this window] [in a new window]   Figure 2. Principle of operation of the initial 2-gate Doppler HCT measurement system.

(1)

(2)

where PT denotes the transmitted acoustic power; T is equal to total loss of the signal between the transducer and the gate Q₁; ₁ and ₂ denote the backscattering coefficients of red blood cells in the gates Q₁ and Q₂, respectively; is equal to the total acoustic attenuation; and z is the axial distance between the 2 gates (see Figure 2).

The backscattering coefficient depends on HCT, cell aggregation, and the concentration of roleaux.⁹ The cell aggregation changes with the spatial shear rate, acceleration, and turbulence.¹⁰ The backscattering coefficient depends on the angle between the flow direction and the transducer axis.¹¹ When the gates (sample volumes) are positioned symmetrically with respect to the center of the vessel, that is, in regions where flow conditions are equivalent, it can be assumed that

(3)

Then, from Equations 1 and 2, the attenuation coefficient can be determined as

(4A)

converting to the power attenuation coefficient:

(4B)

and the HCT value can be then expressed as

(5)

where ₀ is the acoustic attenuation in plasma, and A is a constant determined from the linear regression analysis of the experimental data.

As mentioned earlier, the 20-MHz Doppler signal was chosen to maximize the sensitivity of the HCT meter. The 3-mm-diameter 20-MHz transducer was made of a lithium niobate crystal, and special attention was paid to ensure symmetric distribution of the field generated.

This distribution is shown in Figure 3; it was measured at the axial distances from the transducer surface ranging from 2 to 12 mm in 1-mm steps. The lateral distribution that was recorded from 0- to 14-dB levels in 2-dB intervals is also shown in Figure 3. The measurements were carried out with a 0.5-mm-diameter bilaminar hydrophone (Sonora Medical Systems, Longmont, CO). As can be noticed, the transducer produced an axially symmetric uniform field distribution at axial distances from 3 to 12 mm.

View larger version (24K): [in this window] [in a new window]   Figure 3. Distribution of the axially symmetric field generated by the 20-MHz Doppler transducer.

For the on-axis measurements, the multigate Doppler system described in the next section was used. The power of the Doppler signal scattered on the 0.1% starch dispersion in distilled water was recorded. Similarly to the in vitro experiment described above, the water was flowing in the plastic tube. The scattering coefficient was determined to be constant across the vessel, and the received power signal depended only on the changes of the ultrasonic field and attenuation in water. The distribution of the scattering power versus distance is presented in Figure 4. The total attenuation measured in water was = 1.2 dB/cm (r = 0.99; P < .001). Again, subtracting the attenuation in water = 0.56 dB/cm,⁷ that is, 1.2 – 0.56 dB/cm, yielded attenuation correction of 0.64 dB/cm. This value was then used to correct the in vivo data, because the Doppler power spectrum signal attenuation in blood was on the same order.

View larger version (11K): [in this window] [in a new window]   Figure 4. Distribution of scattering power versus distance from the 20-MHz Doppler transducer.

View larger version (18K): [in this window] [in a new window]   Figure 9. Doppler power profile signals from in vivo measurements as a function of the axial distance from the 20-MHz transducer and the corresponding gate number. For clarity, only 16 distributions recorded during the systolic phase are shown. The regions used for actual HCT determination are marked by dotted vertical lines. A, Radial artery. B, Brachial artery.

View larger version (21K): [in this window] [in a new window]   Figure 5. Principle of the multigate Doppler system designed to determine HCT in vivo.

An array of digital signal processors (TMS320C50; Texas Instruments Incorporated, Dallas, TX) performed the processing task, which consisted of a 64-point fast Fourier transform analysis for each range-gated Doppler signal. The true spectra of the Doppler signals related to the 64 range cells were computed and displayed in real time on the host personal computer monitor. During real-time operation, it was always possible to store all spectral profiles computed over a suitable time interval in the hard disk for possible sequence replay or postprocessing.

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

 
The results of the HCT measurements in vitro and in vivo are presented. In vitro data were obtained by the pulse-echo and 2- and 4-gate Doppler methods, whereas the in vivo data were obtained by the multigate (64-gate) Doppler system.

In Vitro Pulse-Echo Measurements of Human Blood
The results presented below correspond to values of HCT ranging from 0% (plasma) to 80% (separated blood cells).

In Figure 6, attenuation in human blood measured by the pulse-echo method at 20 MHz is plotted against HCT in percent. Discrete data, including plasma (i.e., HCT = 0), are denoted by circles; the solid line represents the outcome of regression analysis. As shown in Figure 6, at 20 MHz attenuation increases linearly with increasing HCT according to the relationship

View larger version (15K): [in this window] [in a new window]   Figure 6. Attenuation in human blood measured in vitro by the pulse-echo method at 20 MHz against HCT.

(6)

The corresponding correlation coefficient was calculated as r = 0.9 (P < .001), and the SD was determined to be 0.21 dB/cm. These values were obtained by taking into account attenuation in water and using the value of 0.56 dB/cm at 20 MHz.⁷

From Equations 4 and 5, the HCT for human blood is calculated as

(7)

In Vitro Doppler Measurements of Animal Blood
Similar to the pulse-echo measurements carried out on human blood, the Doppler measurements in vitro were performed with the use of animal blood assays also ranging from 1% to 65% HCT.

In Figure 7, 4 time-averaged power Doppler spectra, obtained in the frequency range of 0.5 to 3.5 kHz and corresponding to the gates positioned at the axial distances of 3.6, 4.8, 6, and 7.2 mm, are shown. As expected, the shapes of the spectra corresponding to 3.6- and 7.2-mm and 4.8- and 6-mm gates are very similar; the difference in the amplitudes was caused by attenuation of the ultrasonic wave in the 37% HCT blood sample. Hence, the velocities determined at 3.6 and 7.2 mm were equal to 0.11 m/s, and those at 4.8 and 6 mm were equal to 0.14 m/s. As already mentioned, the value of HCT was determined from the ratios of the integrated Doppler power from the 2 sets of measurements, 1 corresponding to signals recorded from 3.6- and 7.2-mm gates and the other from the signals measured within 4.8- and 6.0-mm sample volumes.

View larger version (17K): [in this window] [in a new window]   Figure 7. In vitro HCT measurements by the 4-gate Doppler meter. Doppler spectra corresponding to gates positioned at axial distances of 3.6, 4.8, 6, and 7.2 mm are shown.

View larger version (10K): [in this window] [in a new window]   Figure 8. Comparison of the results obtained by the conventional HCT centrifuge and the 4-gate Doppler method (circles). A, The discrete results (circles) were obtained for 3.6- and 7.2-mm gates. The dotted line was obtained by regression analysis and can be analytically described as y = 0.9x – 4.7. For 14 data points, the correlation coefficient was r = 0.96 (P < .001). B, The discrete results (circles) were obtained for 4.8- and 6.0-mm gates. The dotted line was obtained by regression analysis and can be analytically described as y = 0.63x + 2.7. The corresponding correlation coefficient was r = 0.79 (P < .001).

In both cases, the HCT values measured by the Doppler method were lower than those obtained by a conventional centrifuge. This might have been caused by the different temperatures of the blood during measurement of pulse-echo attenuation and the Doppler measurements. When the distance between gates was short (Fig. 8B), the change in the Doppler amplitude in the gates was small, whereas the variance of the Doppler signal caused by the random phenomenon of scattering was large. As a result, the measuring points were randomly distributed, and the correlation coefficient was lower. In these measurements, the sample volume had a 2-mm diameter (–3 dB) and 0.4-mm thickness. The sample volumes had to be positioned totally within the blood vessel without interfering with the vessel walls. The smallest diameters of the vessel were 2.54 mm for the gates at a 1.2-mm distance and 4.24 mm for the 3.6-mm gates.

In Vivo Doppler Measurements
In vivo measurements of HCT were performed on radial and brachial arteries. For each volunteer, 250 sets of Doppler spectra were recorded, each set being related to all 64 gates across the vessel. The recording was done over 2.5 seconds, which corresponded to 2 heart cycles. After the spectra were acquired, the data were processed with MATLAB software (The MathWorks, Natick, MA). For each spectrum recorded in each gate, the power of the flow signal was calculated, and the Doppler power profile (DPP; the distribution of the Doppler power across the vessel diameter) was obtained. Two different DPPs in radial and brachial arteries are shown in Figure 9. There, the power Doppler signal, in decibels, is plotted against the axial distance from the transducer and the corresponding gate number. For clarity, only 16 power level distributions are shown. Figure 9A corresponds to the data obtained from the radial artery, and the data corresponding to the brachial artery are shown in Figure 9B.

The DPP data were used to calculate the value of HCT in vivo. As can be seen from the data presented in Figure 9, the backscattering of the Doppler signal is random, and the received signal exhibits noise characteristics. Therefore, to calculate the HCT, the following procedure was developed.

First, the value of attenuation was calculated for each DPP according to the procedure for in vitro measurements described above. To ensure that only the spectra from within the blood vessel were recorded, the signals were collected from 16 or 24 gates positioned symmetrically with respect to the center of the vessel. In other words, the data for the radial artery (Fig. 9A) were collected from the signals acquired from gate 14 (corresponding to an 4.4-mm axial distance from the transducer) through gate 30 (6.0-mm distance). Similarly, for the brachial artery (Fig. 9B) the data processed were from gates 16 (4.6-mm distance) through 40 (7.0-mm distance). Both analyzed regions are marked by the dotted vertical lines in Figure 9. The internal diameters of the arteries were 2.5 mm for the radial artery and 3.9 mm for the brachial artery.

Next, the averaged linear regression was calculated from the DPP curve. The slope of the regression line divided by 2 is equal to the attenuation coefficient (Eq. 4B). The HCT value was calculated from Equation 7. Next, all systolic DPPs were averaged to yield the final value of HCT. The average attenuation coefficients were calculated as 8.74 and 7.53 dB/cm for the radial and brachial arteries, respectively. These values, corresponding to HCT values of 57% (radial artery) and 44% (brachial artery), were subsequently compared with those measured by the reference laboratory analyzer method. Blood for the analytical measurements was drawn from the basilic vein, and the measured HCT was 42%. The Doppler results were within ±5% HCT absolute accuracy for data acquired from the brachial artery; however, for the radial artery, the error was greater, approximately ±20% HCT. No variations in the values of the scattering coefficient were observed in the central part of the vessel. To verify the validity of the approach described, 6 patients having similar HCT ranges (42%–46%) were also examined, and a close agreement between those 2 data sets was observed. The measured data are summarized in Table 1.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

In Vitro Doppler Measurements of HCT
Measurements of HCT in vitro by the 2- and 4-gate Doppler system prototypes confirmed the results obtained by other researchers² and indicated that the overall uncertainty in HCT determination depends on the distance between the 2 measurement gates; the uncertainty decreased with increasing gate distance. It is appropriate to note that this behavior is to be expected, provided that the measurements are not affected by the reflections from the blood vessel walls; that is, as mentioned above, the measurement gates have to be positioned within the blood vessel. The HCT values obtained from 168 human blood samples indicated that at 20 MHz the attenuation increased with increasing HCT.

In Vivo Doppler Measurements of Human Blood HCT
The results of the measurements in vivo showed that HCT could be noninvasively determined in that brachial artery to within 5% HCT with the use of the Doppler method described; Such a degree of error is clinically acceptable. It should be noted that the HCT error exceeded 5% when the measurements were performed in the radial artery. This was partly due to the smaller diameter of this artery and, hence, the lower number of gates positioned inside it. Also, it should be pointed out that the multiple-gate Doppler circuitry was essential for in vivo measurements to eliminate the influence of the lateral movement of the vessel measured. As already noted, early in vitro measurements indicated that the HCT error depended on the location of the gates. The 64-gate system developed here allowed this error to be minimized by excluding the measurements influenced by the attenuation within the vessel wall and by averaging the data acquired from multiple gates.

In conclusion, we developed a new, noninvasive, clinically applicable 20-MHz pulsed Doppler ultrasonic HCT meter. The transcutaneous device was initially tested in a laboratory environment, and its performance was verified subsequently in a clinical environment. It was found that the multigate system largely simplified placement of an ultrasonic probing beam in the center of the blood vessel. Although the ±5% HCT uncertainty is clinically acceptable and considered adequate for monitoring changes in patients in shock or during dialysis, the method’s applicability to arbitrarily selected vessels is being examined, and work is under way to reduce the HCT measurement error to much less than 5% HCT.

	Footnotes

 
Received July 11, 2002, from the Institute of Fundamental Technological Research, Polish Academy of Sciences, Warsaw, Poland (W.S., A.N.); Electronics and Telecommunications Department, University of Florence, Florence, Italy (F.G., P.T.); and School of Biomedical Engineering, Science, and Health Systems and Department of Electrical and Computer Engineering, Drexel University, Philadelphia, Pennsylvania USA (P.A.L.). Revision requested August 1, 2002. Revised manuscript accepted for publication December 16, 2002.

This work was supported by Polish State Committee for Scientific Research grant KBN 8T11E02317.

	References

Top Abstract Introduction Materials and Methods Results Discussion References

 

1. Johner C, Chamney P, Schneditz D, Krämer M. Evaluation of an ultrasonic blood volume monitor. Nephrol Dial Transplant 1998; 13:2098–2103.[Abstract/Free Full Text]
2. Carstensen E, Schwan H. Acoustic properties of hemoglobin solutions. J Acoust Soc Am 1959; 31: 305–311.
3. Hughes D, Geddes L, Babbs C, Bourland J, Newhouse V. Attenuation and speed of 10 MHz ultrasound in canine blood of various packed-cell volumes at 37°C. Med Biol Eng Comput 1979; 17:619–622.[Medline]
4. Yuan Y, Shung K. Ultrasonic backscatter from flowing whole blood, I: dependence on shear rate and hematocrit. J Acoust Soc Am 1988; 84:52–58.[Medline]
5. Vandegriff K, McCarthy M, Rohlfs R, Winslow R. Colloid osmotic properties of modified hemoglobins: chemically cross-linked versus polyethylene glycol surface-conjugated. Biophys Chem 1997; 69:23–30.[Medline]
6. Standl T, Burmeister M, Horn E, Wilhelm S, Knoefel W, Esch J. Bovine haemoglobin-based oxygen carrier for patients undergoing haemodilution before liver resection. Br J Anaesth 1998; 80:189–194.[Abstract/Free Full Text]
7. Duck F. Acoustic properties of tissue at ultrasonic frequencies. In: Physical Properties of Tissue: A Comprehensive Reference Book. London, England: Academic Press; 1990:73–135.
8. Rudgers A. Acoustic pulses scattered by a rigid sphere immersed in a fluid. J Acoust Soc Am 1969; 45:900–910.
9. Lim B, Cobbold R. On the relation between aggregation, packing and backscattered ultrasound signal for whole blood. Ultrasound Med Biol 1999; 25:1395–1405.[Medline]
10. Cloutier G, Quin Z, Durand L, Teh B. Power Doppler ultrasound evaluation of the shear rate and shear stress dependencies of red blood cell aggregation. IEEE Trans Biomed Eng 1996; 43:441–450.[Medline]
11. Allard L, Cloutier G, Durand L. Effect of the insonification angle on the Doppler backscattered power under red blood cell aggregation conditions. IEEE Trans Ultrason Ferroelectr Freq Control 1996; 43: 211–219.
12. Tortoli P, Guidi F, Guidi G, Atzeni C. Spectral velocity profiles for detailed ultrasound flow analysis. IEEE Trans Ultrason Ferroelectr Freq Control 1996; 43: 654–659.
13. Guidi F, Guidi G, Ricci S, Atzeni C, Tortoli P. High-speed parallel processing of biomedical ultrasound signals. In: Proceedings of the Second European DSP Education and Research Conference, Parigi, Italy. Paris, France: Texas Instruments; 1998:327–330.

This article has been cited by other articles:

				J. W. McMurdy, G. D. Jay, S. Suner, and G. Crawford Noninvasive Optical, Electrical, and Acoustic Methods of Total Hemoglobin Determination Clin. Chem., February 1, 2008; 54(2): 264 - 272. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Secomski, W. Articles by Lewin, P. A. Search for Related Content PubMed PubMed Citation Articles by Secomski, W. Articles by Lewin, P. A.