Comparison of Lung Volume Measurements byAntero-Posterior Chest X-Ray and the SF 6 Washout

Technique in Mechanically Ventilated Infants

Ulrich Thome, MD,1* Andreas To pfer, PhD,2 Peter Schaller, PhD,2 and Frank Pohlandt, MD, PhD1†

While anterior/posterior chest x-rays (CXR) are routinely performed to estimate lung volume (LV)and adjust ventilator settings, the precise measurement of LV requires additional sophistication.In 31 infants ventilated because of surfactant deficiency (n = 23), bronchopulmonary dysplasia(n = 7), or pulmonary hypoplasia (n = 1) with either intermittent positive pressure (n = 18) or highfrequency oscillation (n = 13) (gestational age 23–39 weeks (median 26 weeks); birthweight550–2780 g (median 840 g); age at measurement 1–91 days (median 6 days); weight at studytime (WST) 675–3000 g (median 938 g)), we investigated whether LV, as measured by the sulfurhexafluoride (SF6) washout technique, could by estimated from CXR by: 1) the sum (A+B) of theright (A) and left (B) lung fields areas; 2) the product (LxW) of the distances from the right apexto the right costophrenic angle (L) and between both costophrenic angles (W); 3) the diaphragmposition relative to the posterior parts of the ribs (DP); and 4) the lung radiolucency (RL, grades0–4). Correlations between A+B (r = 0.44) or LxW (r = 0.37) and LV were poor, but improvedwhen A+B, LxW, and LV were normalized to WST: (A+B)/WST vs. LV/WST (r = 0.74), andLxW/WST vs. LV/WST (r = 0.67). DP (r = 0.13) and RL (Spearman’s rho = 0.17) did not correlatewith LV/WST. A multiple linear regression analysis led to the following best-fit equation: LV/WST= 2.58 (A+B)/WST − 5.47 DP + 42.2 (r = 0.83). We concluded that an estimate of LV from CXRlacked sufficient accuracy. DP and RL did not correlate with LV measured by SF6 washout.Pediatr Pulmonol. 1998; 26:265–272. © 1998 Wiley-Liss, Inc.

Key words: chest roentgenogram; lung volume; functional residual capacity; sulfurhexafluoride washout technique; mechanical ventilation; newborn infants.

INTRODUCTION

Respiratory management of infants with respiratorydistress syndrome is aimed at maintaining an adequatelung volume (LV) despite the lung’s tendency to col-lapse. This is done by using positive end-expiratory pres-sures (PEEP),1 high frequency oscillatory ventilation(HFOV) with a mean airway pressure (MAP) sufficientlyhigh to optimize lung volume2,3 and surfactant replace-ment.4 Since the appropriate LV is an important deter-minant for successful mechanical ventilation, its mea-surement would be helpful in arriving at the appropriateventilator settings for each patient. While direct measure-ment requires sophisticated devices, anterior/posteriorchest x-rays (CXR) are readily available and are rou-tinely performed in all infants requiring mechanical ven-tilation. CXR findings such as radiolucency and dia-phragm position are commonly used by clinicians5,6 toestimate lung inflation and adjust ventilator settings. Fur-ther, various measurements based on anterior–posteriorand lateral chest radiographs have been shown to corre-

late with LV in a rabbit model,7 in excised lungs,8 inadults with normal lungs,9–14 in patients with variouslung diseases,11,15–17 and in children 4–16 years ofage.18–21 In addition, two methods, which required only

This work was presented in part at the SPR-Meeting 1997 in Wash-ington DC.

1Division of Neonatology and Pediatric Critical Care, Department ofPediatrics, University of Ulm, Ulm, Germany.

2Department of Pediatrics, Technical University, Dresden, Germany.

*Correspondence to: Ulrich Thome, MD, Division of Neonatology,Department of Pediatrics, University of Alabama at Birmingham, 525New Hillman Bldg., 619 South 19th St., Birmingham, AL 35233-7335.

†Reprint requests to: Frank Pohlandt, MD, PhD, Division of Neona-tology and Pediatric Critical Care, University Children’s Hospital,D-82075 Ulm, Germany.

Received 23 February 1998; accepted 1 June 1998.

Pediatric Pulmonology 26:265–272 (1998)

Diagnostic and Therapeutic Methods

© 1998 Wiley-Liss, Inc.

one CXR, have been described in infants with clinicaland radiological evidence of bronchopulmonary dyspla-sia, but not mechanically ventilated and less than 1 yearold,22 and in children after lung transplantation.23 Wehypothesized that the latter two methods for estimatingLV based on CXR measurements may be valid in me-chanically ventilated infants: planimetry of the right (A)and left (B) lung fields (A+B),22 and calculation of theproduct LxW of the distances from the right apex to theright costophrenic angle (L) and from the right to the leftcostophrenic angle (W).23 In addition, we evaluated howtwo items commonly used in clinical practice correlate tothe LV: diaphragm position (DP) relative to the posteriorparts of the ribs and lung radiolucency (RL).

Therefore, we compared LVs calculated from variousradiological parameters with LV measured by an opencircuit washout technique using sulfur hexafluoride(SF6), an insoluble, nontoxic tracer gas.24,25 The tech-nique has been adapted for use with both intermittentpositive pressure ventilation (IPPV)26,27 and HFOV.28

We feel that the term functional residual capacity(FRC), which denotes the end-expiratory lung volume, isnot appropriate when using HFOV. Therefore, we usedlung volume (LV) in this article. In IPPV, LV is theend-expiratory lung volume and is thus identical to FRC;in HFOV, LV denotes the mean lung volume.

METHODS

Patients

All infants who had their lung volume measured in theneonatal intensive care unit at the University Children’sHospital of Ulm from 1993 to 1997 were included in thisretrospective study, provided a CXR of adequate qualityand taken within ±24 hr at the same ventilator settings asthe LV measurement by SF6 tracer washout was avail-able. The infants were ventilated either with IPPV orHFOV. IPPV was carried out with time-cycled, pressure-limited, constant flow infant ventilators, such as theSechrist IV 100B (Sechrist, Anaheim, CA, USA),

Stephan HF 300 (Stephan, Gackenbach, Germany) or theInfant-Star (Nellcor Puritan Bennett, Carlsbad, CA,USA). HFOV was performed according to the high lungvolume strategy6 using an HFV-Infant-Star with softwareversion 83 (Nellcor Puritan Bennett). The HFOV fre-quency was set at 10 Hz throughout.

Chest X-Ray Evaluation

CXRs were taken in the supine position. The infant laydirectly on the plate and the distance to the x-ray tubewas 1 meter. It is the general policy in our unit to takeend-expiratory images in infants ventilated with IPPV bybriefly switching to continuous positive airway pressure(CPAP). In infants ventilated with HFOV, ventilator set-tings were left unchanged while the CXR was taken. ACardio C 200 medical image processor (Kontron, Mu-nich, Germany) was used for CXR evaluation. For axialand planimetry measurements, the processor wasequipped with a sense pad and an electromagnetic mouseand it was capable of sensing the position of the mouseanywhere on the pad with an accuracy of 0.01 cm. Tocorrect for the magnification factor of our x-rays, weimaged a steel ball with a diameter of 3 cm at the samedistance from the x-ray tube and the film as an infantwould be. The resulting image of the steel ball had adiameter of approximately 3.1 cm and was used to cali-brate the image processor.

The translucent sense pad was placed on a light boxand all measurements were made directly on the CXRfilm. Planimetry of the right (A) and left (B) lung fieldswas performed by following the outlines of the lungfields with the crosshairs of the electromagnetic mouse(Fig. 1), as described previously.8,14 Likewise, the dis-tances from the right apex to the right costophrenic angle(L) and from the right to the left costophrenic angle (W)were measured by placing the crosshairs at the respectivepositions and pressing a button. All measurements weremade five times on each x-ray, and the mean values wereused for calculation of the sum of both lung fields (A+B)and the product LxW. In addition, we determined thediaphragm position relative to the posterior parts of theribs and the lung radiolucency. The latter was graded asfollows:

Grade 0: Perfectly radiolucent lung fields with sharp car-diac and diaphragmatic margins;

Grade 1: Slightly reduced radiolucency with still sharpcardiac and diaphragmatic margins;

Grade 2: Markedly reduced radiolucency with retainedcardiac and diaphragmatic margins;

Grade 3: Severely reduced radiolucency with airbroncho-grams and blurred cardiac and diaphragmatic margins;and

Grade 4: Almost completely white lung fields with orwithout airbronchograms and barely visible cardiacand diaphragmatic margins

Abbreviations

A+B Sum of the right (A) and left (B) lung fields by planimetryCPAP Continuous positive airway pressureDP Diaphragm position relative to the posterior parts of the

ribsHFOV High frequency oscillatory ventilationIPPV Intermittent positive pressure ventilationLV Lung volumeLxW Product of the distances from the right apex to the right

costophrenic angle (L) and from the right to the leftcostophrenic angles (W).

MAP Mean airway pressurePEEP Positive end-expiratory pressureRL Lung radiolucency (grades 0–4)WST Weight at study time

266 Thome et al.

Lung Volume Measurement

LV was measured by washout of SF6.24,25The setup ofthe measurement apparatus, the measurement process,and the algorithms for calculation of LV have been de-scribed previously.26–29 In brief, a pneumotachographand a fast mainstream infrared SF6 analyzer (Siemens-Elema, Solna, Sweden) were inserted between the Y-piece of the ventilator circuit and the endotracheal tube.The pneumotachograph30 had a dead space of 0.9 mL anda resistance of 1.1 kPa s/L at 5 L/min, and the cuvette ofthe SF6 analyzer had a dead-space of 1 mL and a resis-tance of 0.26 kPa s/L at 5 L/min. SF6 was added to theventilator bias flow at a concentration of 0.8–1.1% untilthe airway SF6 concentration was constant. LV was de-termined during washout of the tracer by continuous reg-istration of respiratory flow and SF6 concentration andcalculation of the total volume of expired SF6. Divisionof the latter by the initial SF6 concentration provided ameasurement of LV at the transition from wash-in towashout. Any changes of ventilation pattern or LV dur-ing the washout would not change the total amount oftracer gas to be washed out, and thus would not distortthe results. Therefore, IPPV was also used for the SF6

washout in the HFOV ventilated patients,28,31,32a periodwhen exact tidal volume recordings were mandatory. Af-ter equilibration with the tracer gas during HFOV, thetransition from HFOV to IPPV for washout required twoswitches to be set on the HFV-Infantstar.First, HFOVwas switched off (CPAP). This was detected by the com-puter controlling the LV measurement apparatus. SF6

admixture to the bias flow was stopped, its concentrationin the airway was stored in the computer software forlater LV calculation, and registration of the expired SF6

was started.Second,the HFV-Infantstar was switchedfrom CPAP to IMV and tracer gas washout began. Theprevious MAP became the PEEP of the new IPPV. If thePEEP was above 6 cm H2O, it was lowered during the

following breaths to allow for a sufficiently large tidalvolume for a quick washout. The other IPPV settingswere: peak inspiratory pressure, 20–22 cm H2O; inspira-tory time, 0.25–0.33 sec; rate, 60 breaths/min. Correctsynchrony between the change of the ventilation modeand the start of the expired tracer gas registration andvolume calculation was ensured by the software of themeasurement device. After completion of the washout(duration∼1 min), the ventilation was switched back toHFOV. The short IPPV phase did not appear to have aprolonged effect on LV.28

In infants ventilated with IPPV, the end-expiratory LV(functional residual capacity) was measured because thetransition from wash-in to washout occurred at the end ofan expiration. The accuracy and reproducibility of LVmeasurements carried out with our apparatus have beentested in IPPV and HFOV: measurements of a conven-tionally ventilated dummy lung with a known, adjustablevolume resulted in a difference of 0.7 ± 3.2% (mean ±SD). The coefficient of variation across 20 LV determi-nations in five adult rabbits was 1.7% during CPAP and2.0% during IPPV.26 Measurements of a dummy lungwith a known volume (50 ml), using HFOV for wash-inand IPPV for washout with the same method as describedin this work, resulted in a difference of 0.4 ± 2.3%.Dummy lung measurements were repeated on a regularbasis to ensure accuracy. In human infants, the mediancoefficient of variation was 4.0% during IPPV27 and4.9% during HFOV.28 The minimum time between twomeasurements was 2 min. All measurements were madein triplicate and the mean values were used for furtheranalysis. SF6 was supplied by Linde (Munich, Germany)as a 1:1 mixture with nitrogen. This mixture has beenapproved by the Bavarian Government Health Authorityfor diagnostic use in human subjects. The infants werenot paralyzed but in quiet sleep while the measurementswere carried out. Each infant’s heart rate, transcutaneousPaCO2 and arterial oxygen saturation were monitored.Informed consent for LV measurement was obtainedfrom the parents of each infant.

Data Analysis

Division of LV by weight at the time of study (WST)yielded the specific lung volume (LV/WST). Similarly,A+B and LxW were divided by WST, yielding (A+B)/WST and LxW/WST. Linear correlation coefficients andP-values for the relations between A+B and LV, LxWand LV, (A+B)/WST and LV/WST, LxW/WST and LV/WST, and DP and LV/WST were determined. To evalu-ate the relationship between RL and LV/WST, Spear-man’s rho was calculated. Further, a multiple regressionanalysis of the relationship of LV/WST with (A+B)/WST, LxW/WST, DP and RL was performed to deter-mine whether a combination of measurements would im-prove the estimate of the LV from CXR. All regression

Fig. 1. An anterior/posterior chest x-ray indicating the mea-sured parameters A, B, L, and W.

Lung Volume Measurements in Premature Infants 267

analyses were performed with SAS software (SAS Insti-tute, Cary, NC, USA).

RESULTS

Selection criteria were met by 31 infants who wereventilated because of respiratory distress syndrome ofprematurity (n4 23), bronchopulmonary dysplasia (n47), or pulmonary hypoplasia (n4 1, patient 17 in theAppendix), by IPPV (n4 18), or HFOV (n4 13). Thedemographic data and the results of CXR and LV mea-surements are listed in Table 1. The standard deviationsof repeated LV measurements were between 0.24 and3.66 (median 1.14), and the coefficients of variation werebetween 0.01 and 0.16 (median 0.04). As expected fromthe wide range of pulmonary pathology and the patient’s

age, the measured LV covered a wide range as well.LV/WST was significantly greater (P < 0.01) withHFOV (median 29.1 mL/kg) than with IPPV (median22.1 mL/kg), as was (A+B)/WST (HFOV: median 15.4cm2/kg, IPPV: median 13.2 cm2/kg, P < 0.04) and DP(HFOV: median 10, IPPV: median 9,P < 0.03), but nosignificant differences between HFOV- and IPPV-treatedinfants were noted for LxW/WST (HFOV: median 38.9cm2/kg, IPPV: median 34.8 cm2/kg, P 4 0.057), and forRL (HFOV: median 2, IPPV: median 2,P 4 0.48) by theMann Whitney U test.

The clinically used parameters DP and RL did notcorrelate with the LV measured by SF6 (Table 2). Whilethe two radiological measurements A+B and LxW cor-related well with each other (r2 4 0.53), their correlation

TABLE 2—Results of the Regression Analyses

Y x1 x2 s1 s2 o r2 r p

LV A+B — 1.04 — 11.2 0.20 0.44 0.012LV L*W — 0.32 — 14.0 0.14 0.37 0.04LV/WST (A+B)/WST — 1.89 — −0.18 0.54 0.74 <0.0001LV/WST L×W/WST — 0.82 — −3.0 0.45 0.67 <0.0001LV/WST DP — 1.49 — 12.2 0.016 0.13 0.50LV/WST RL — — — — — 0.17* 0.35LV/WST (A+B)/WST DP 2.58 −5.47 42.2 0.68 0.83 <0.0001

Including more than two independent variables into the regression analysis did not substantially improve r2 (less than 1%) and is therefore notshown.Calculated according to the equation y4 s1 x1 + s2 x2 + o, where y4 dependent variable, x1 4 first independent variable, x2 4 secondindependent variable, s1 4 parameter for x1, s2 4 parameter for x2, o 4 vertical offset, r4 correlation coefficient,P 4 probability.LV 4 lung volume measured, A4 right lung field area, B4 left lung field area, L4 vertical distance from the right apex to the rightcostophrenic angle, W4 width from the right to the left costophrenic angle, WST4 weight at study time, DP4 diaphragm position, RL4radiolucency.*Spearman’s rho (ordinal data).

TABLE 1—Median, Minimum, and Maximum Values of Demographic Data, VentilatorSettings, Radiological Measurements, and Lung Volume (LV) Measurements of the31 Infants

Median Minimum Maximum CV

Gestational age (weeks) 26 4/7 23 0/7 39 5/7Birthweight (kg) 0.840 0.550 2.780Weight at study time (WST, kg) 0.938 0.675 3.000Age (days) 6 1 91Time between CXR and LV measurement (h) 3.9 1 17.8PEEP (IPPV, cmH2O, n 4 18) 5 3 8Mean Airway pressure (HFOV, cmH2O, n 4 13) 10 5 15Radiolucency (0–4) 2 1 4Diaphragm position (DP, ribs) 9 8 11A (right lung field area, cm2) 7.4 5.5 12.9 0.01B (left lung field area, cm2) 5.8 4.1 10.6 0.01L (vertical distance, cm) 4.9 3.6 7.2 0.003W (width, cm) 7.1 6.3 9.6 0.002A+B (cm2) 13.2 10.9 22.5L×W (cm2) 34.7 23.9 55.4Lung volume (LV) by SF6 wash-out (mL) 24.9 12.0 39.0 0.04(A+B)/WST (A+B normalized by weight, cm2/kg) 14.3 4.4 20.0L×W/WST (L×W normalized by weight, cm2/kg) 36.1 17.6 49.7LV/WST (LV normalized by weight, mL/kg) 26.2 7.9 45.4Calculated LV (2.58 (A+B)/WST−5.47 DP+42.2) 26.5 6.6 42.3

CV 4 median coefficient of variation of repeated measurements in the same subject or CXR.

268 Thome et al.

with the measured LV was poor (Table 2). Somewhatbetter correlations were obtained with A+B, LxW, andLV normalized to body weight: (A+B)/WST correlatedwell with LV/WST (r2 4 0.54), resulting in the follow-ing linear regression equation: LV/WST4 1.89 (A+B)/WST − 0.18. (Fig. 2, Table 2). The correlation betweenLxW/WST and LV/WST was less favorable (r2 4 0.45);the regression equation was LV/WST4 0.82 LxW/WST− 3.0 (Fig. 3, Table 2). The correlation coefficients werenot improved by separating the analysis of IPPV- andHFOV-treated patients, or by the exclusion of the infantsin whom the time delay between CXR and LV measure-ment exceeded 6 hr.

A multiple linear regression analysis revealed that theinclusion of both (A+B)/WST and DP improved the es-timate of LV/WST. By combining these variables, r2

increased to 0.68; the regression equation

LV/WST 4 2.58 (A+B)/WST − 5.47 DP + 42.2

provided the best estimate of LV as determined by SF6

washout. Interestingly, a negative factor was assigned toDP by this analysis (Fig. 4, Table 2). Further improve-ments of r2 by including more independent variables intothe multiple regression analysis, such as LxW/WST orRL, were negligible (less than 1%).

DISCUSSION

Since the first attempt in 1933,9 a number of methodsto estimate lung volume from the CXR have been sug-gested. Most of them were based on anterior–posteriorand lateral images of adults. One approach was calcula-tion of volume of the chest by simplifying its shape to fitit into geometric models, such as a truncated cone10 or astack of multiple elliptical cylindroids.11,13,15,17,19An-other approach consisted of planimetry of the right (A),left (B), and lateral (C) lung fields. Various formulashave been developed for calculation of LV by planime-try, such as

[(AxC)0.75 + (BxC)0.75] 0.67 + 320,8,18,20

42.9 [(A+B) C/100]0.679,18,21

0.67 (A+B+C)1.37,18,21

8.5 (A+B+C) − 1,200,11,14,16and10.3 (A+B+C) − 2,930.12

The third approach was making linear measurementsbetween selected points.14,23 Recently, two methods forestimating LV in children and infants which require onlyan anterior–posterior image have been published. Onewas based on linear measurements,23 the other on pla-nimetry of the two lung fields, resulting in the formula:22

Fig. 2. Scatterplot of lung volume measured by SF 6 washoutdivided by body weight (LV/WST) vs. the sum of right and leftlung field areas divided by body weight ((A+B)/WST). Solidcircles represent patients ventilated with IPPV, triangles withHFOV. The dashed line represents the regression line (LV/WST= 1.89 (A+B)/WST − 0.18; r 2 = 0.54).

Fig. 3. Scatterplot of lung volume measured by SF 6 washoutdivided by body weight (LV/WST) vs. product of length andwidth divided by body weight (LxW/WST). Solid circles repre-sent patients ventilated with IPPV, triangles with HFOV. Thedashed line represents the regression line (LV/WST = 0.82 LxW/WST − 3; r2 = 0.45).

Fig. 4. Scatterplot of lung volume measured by SF 6 washoutdivided by body weight (LV/WST) vs. radiographic LV/WST cal-culated from right + left lung field areas divided by body weight((A+B)/WST) and DP, according to the results of a multiple linearregression analysis (LV/WST = 2.58 (A+B)/WST − 5.47 DP +42.2). Solid circles represent patients ventilated with IPPV, tri-angles with HFOV. The dashed line represents the regressionline (r 2 = 0.68).

Lung Volume Measurements in Premature Infants 269

Inspiratory lung volume4 3.5 (A+B) − 13.5

with r 4 0.7. Although most of the equations cited abovehave been developed by using plethysmographic or gasdilution measurements as reference, none has achievedroutine use in clinical practice, probably because of thedifficulty of planimetry when an image processor is notavailable and the confusing multitude and complexity ofthe published equations. Clinicians tend to use simplerassessments based on the DP5,6 and RL to estimate LV.Despite their widespread use, however, the latter param-eters have not been validated.

The accuracy of LV measurement by SF6 washout hasbeen shown previously by different groups: LV measure-ments obtained by SF6 washout were highly correlatedwith results by the nitrogen washout method in a lungmodel33 as well as in human subjects.34,35 Likewise, re-sults of the SF6 washout method did not differ from thehelium dilution method in adult rabbits36 and human sub-jects.37,38 Our own measurement apparatus has beentested with a dummy lung,26 and such tests were repeatedon a regular basis during this study. Therefore, the SF6

washout method is sufficiently validated to serve as a‘‘gold standard’’ when estimating LV from CXR.

This study is the first in which LV was estimated fromCXRs in mechanically ventilated infants. We analyzedthe relationship between LV obtained by SF6 washoutand values obtained by planimetry of the right and leftlung fields (A+B), linear measurements (LxW), DP andRL. These parameters are especially interesting to neo-natologists, as they can be determined without lateralimages. A+B and LxW were poorly correlated to LV, butthe estimates improved markedly by relating them toWST. This improvement may be explained by the factthat WST introduces information about the third dimen-sion, which is lacking in the pure CXR measurements.By simple linear regression analysis, the best correlationbetween LV/WST and radiological parameters was ob-tained from the measurements of (A+B)/WST (r40.74). Despite their widespread recognition in clinicalpractice, DP and RL were not correlated to LV/WST.

Multiple linear regression analysis indicated that thecorrelation to LV/WST can be improved by combiningthe derived values of (A+B)/WST and DP. Interestingly,a negative factor was assigned to DP (LV/WST4 2.58(A+B)/WST − 5.47 DP + 42.2, r4 0.83; see Table 2).This illustrates that a caudal diaphragm position does notnecessarily mean a large LV, but leads to an increasedlung field area on CXR and thereby results in overesti-mation of LV when planimetry is performed. This iscorrected by assigning a negative factor to DP. For clini-cal decision making, however, the accuracy of LV esti-mation by this equation was still insufficient. Since thecorrelations were poor, a further assessment of agree-ment according to the technique of Bland and Altman39

was considered futile and was, therefore, not done.

The patients in this study featured a wide range of lungpathology and age. Some were in the acute phase ofrespiratory distress syndrome and had relatively low lungvolumes, others were older and had developed broncho-pulmonary dysplasia, with high lung volumes, especiallyat high PEEP settings. Further, the patients were treatedby two different ventilation modes (IPPV and HFOV),and both modes were used with different levels of PEEPand MAP. These are complex conditions for any type ofindirect LV estimation, but they realistically reflect clini-cal practice. To be useful clinically, any method for LVestimation must be reliable in a wide range of patients.

This study is retrospective, and the time elapsed be-tween LV measurement and CXR was quite long in somepatients. However, excluding the patients with the long-est delay between LV measurement and CXR, or exclud-ing IPPV- or HFOV-treated patients and repeating theanalyses, led to similar regressions and correlation coef-ficients, thus strengthening the validity of our results.However, a prospective study should be undertaken.

The equation derived from our multiple linear regres-sion analysis is different from ones published earlier. Itincludes the measurement of planimetry and diaphragmposition and allows compensation for the large lung fieldarea when the diaphragm is displaced caudally, resultingin an improved correlation between measured and calcu-lated LV/WST. However, we do not think that this is thecomplete solution, since accuracy of LV/WST predictionwas still insufficient (r2 4 0.68, see Fig. 4) when com-pared to LV measurements obtained by a validatedmethod, such as the SF6 washout technique.26,33–38Thiswas only the best fit equation for our study population,and it may be different for other groups of infants.

The numerous methods and equations published pre-viously and the results presented here indicate that areliable calculation of LV from CXR is probably notpossible. If knowledge of a patient’s LV is clinicallyimportant, there is probably no substitute for direct mea-surement. Adding lateral images and, thus, informationon the third dimension may improve the accuracy of LVestimation by CXR, but the shape of the thoracic cavityis much too complicated to be sufficiently described bytwo images. The additional disturbance of the infant,radiation exposure, and cost of a lateral image may not bejustified to gain a small improvement in accuracy. DPand RL may not be sufficiently reliable to make clinicaldecisions regarding ventilator settings.

ACKNOWLEDGMENTS

The authors thank D. Lang (Division of Pediatric Car-diology, University Children’s Hospital of Ulm, Ger-many) for making the image processor available, J. Ho¨-gel (Department of Biometry and Medical Documenta-tion, University of Ulm, Germany) for statistical advice,and R.J. Kopotic for reviewing the manuscript.

270 Thome et al.

APPENDIX—The patients enrolled in this study, and the individual results obtained from x-ray evaluation and lungvolume measurements.

Pt.No. GA BW WST Vent. Age TE AP RL DP A B L W A+B L×W LV

A+B L×W LV

LVCWST WST WST

1 24 610 710 IPPV 22 1.8 6 2 8.3 6.5 4.4 3.6 6.8 10.9 24.6 30.7 15.4 34.6 43.2 36.82 31 870 960 IPPV 7 3 4 3 9.0 6.9 4.7 5.0 7.0 11.6 34.7 21.5 12.1 36.1 22.4 24.23 28 550 1020 IPPV 35 3 5 2 9.0 9.3 5.7 4.9 7.9 15.0 38.5 16.9 14.7 37.8 16.6 30.84 25 790 1090 IPPV 10 3.5 7 3 10.0 9.9 6.7 5.5 7.4 16.5 40.7 32.3 15.2 37.3 29.6 26.65 36 2780 3000 IPPV 1 7.3 4.5 1 8.0 6.8 6.4 5.5 9.6 13.2 52.8 35.7 4.4 17.6 11.9 9.86 26 970 970 IPPV 1 5.5 3 1 9.0 6.2 5.1 4.9 6.5 11.3 32.2 18.6 11.6 33.2 19.2 23.07 28 1020 2020 IPPV 45 4 5 1 8.5 10.7 10.6 5.3 9.6 21.3 51.0 39.0 10.5 25.2 19.3 22.98 26 890 868 IPPV 3 2.5 4 2 10.0 6.4 5.3 4.3 6.9 11.7 29.4 19.0 13.5 33.9 21.9 22.29 27 710 930 IPPV 17 17.8 6 2 9.0 7.7 6.1 4.8 7.5 13.7 36.2 33.5 14.8 38.9 36.0 31.1

10 25 590 689 IPPV 14 2 5 2 9.0 7.0 6.2 4.7 7.2 13.2 34.3 26.3 19.1 49.7 38.2 42.311 26 1030 1030 IPPV 3 6.3 6 2 10.0 7.7 4.1 5.1 7.2 11.8 36.7 12.0 11.4 35.6 11.7 17.012 26 705 894 IPPV 6 2.3 4.5 3 9.0 7.1 4.5 4.4 6.8 11.6 30.0 22.8 13.0 33.6 25.5 26.513 26 705 930 IPPV 26 4 5 4 9.0 6.4 4.6 4.4 7.3 10.9 32.1 25.3 11.7 34.5 27.2 23.214 26 780 880 IPPV 14 8.4 5 2 10.5 8.6 6.5 5.0 7.5 15.0 37.6 20.0 17.1 42.7 22.7 28.815 29 1640 1580 IPPV 2 1 8 1 10.0 12.9 9.6 7.2 7.7 22.5 55.4 33.2 14.2 35.1 21.0 24.216 30 1010 1280 IPPV 13 3.3 4 3 9.0 8.1 5.2 4.9 7.1 13.3 34.6 14.9 10.4 27.0 11.6 19.717 39 1690 2600 IPPV 91 10.5 4 2 8.8 5.6 6.8 5.0 9.2 12.4 45.7 20.6 4.8 17.6 7.9 6.618 26 1120 1040 IPPV 6 3.9 3 1 10.0 10.1 7.3 5.3 7.3 17.3 39.1 23.8 16.6 37.6 22.9 30.419 24 660 710 HFOV 4 2.5 11 2 10.8 6.8 6.1 4.7 6.5 13.0 30.5 20.6 18.3 43.0 29.1 30.220 28 710 712 HFOV 2 2.3 12 2 9.0 6.0 5.0 4.4 6.7 11.0 29.3 25.9 15.4 41.2 36.4 32.721 27 960 980 HFOV 4 3 6 3 10.0 8.7 4.5 6.2 6.6 13.2 40.7 25.9 13.5 41.5 26.4 22.322 25 750 824 HFOV 7 17 7 2 10.5 6.7 5.8 4.0 6.7 12.5 26.5 21.6 15.1 32.2 26.2 23.723 24 630 675 HFOV 4 7.3 9 2 10.0 7.1 6.4 4.7 7.1 13.5 32.8 30.7 20.0 48.7 45.4 39.124 27 990 1090 HFOV 3 1.5 11 3 10.0 8.0 5.8 5.1 7.5 13.8 38.0 24.9 12.6 34.9 22.8 20.125 26 960 938 HFOV 3 1 12 3 10.5 8.7 8.3 5.5 6.7 17.0 36.3 24.0 18.1 38.7 25.6 31.626 24 700 700 HFOV 3 17.2 9 2 11.0 6.9 5.7 4.9 6.3 12.6 30.7 23.4 18.0 43.9 33.5 28.427 25 900 860 HFOV 7 7.5 5 2 10.0 7.8 6.9 5.5 7.0 14.7 38.3 29.2 17.1 44.6 34.0 31.728 26 840 1015 HFOV 10 2.6 10 3 10.3 9.4 5.2 5.4 7.2 14.6 38.5 31.8 14.3 37.9 31.3 23.129 23 640 764 HFOV 10 6.7 14 3 9.0 7.6 5.5 4.5 6.7 13.0 29.7 30.3 17.0 38.9 39.7 36.930 24 790 850 HFOV 3 5.3 9 1 8.8 5.5 5.4 3.7 6.5 10.9 23.9 23.6 12.8 28.1 27.8 27.331 31 1140 1250 HFOV 3 11 15 1 9.0 7.4 6.3 4.7 7.0 13.7 32.7 35.7 11.0 26.2 28.6 21.3median 26 840 938 6 3.9 6 2 9.0 7.4 5.8 4.9 7.113.2 34.7 24.9 14.3 36.1 26.2 26.5

GA 4 gestational age (wk); BW4 birthweight (g); WST4 weight at study time (g); Vent.4 ventilation mode; IPPV4 intermittent positivepressure ventilation; HFOV4 high frequency oscillatory ventilation; age4 chronological age (days); TE4 time elapsed between CXR andLV measurement (hours); AP4 airway pressure (IPPV: AP4 PEEP; HFOV: AP4 mean AP); RL4 radiolucency; DP4 diaphragm position;A 4 right lung field area cm2; B 4 left lung field area cm2; L 4 vertical distance from the right apex to the right costophrenic angle (cm);W 4 width from the right to the left costophrenic angle (cm); LV4 lung volume measured (mL); LVC4 LV/WST calculated from(A+B)/WST and DP by the following equation: LVC4 2.58 (A+B)/WST − 5.47 DP + 42.2; A, B, L, and W are mean values of fivemeasurements, LV of three measurements.

REFERENCES

1. Duncan AW, Oh TE, Hillman DR. PEEP and CPAP. AnaesthIntens Care. 1986; 14:236–250.

2. Froese AB, Bryan AC. High frequency ventilation. Am RevRespir Dis. 1987; 135:1363–1374.

3. Clark RH, Gerstmann DR. Controversies in high-frequency ven-tilation. Clin Perinatol. 1998; 25:113–122.

4. Svenningsen NW. Pulmonary functional residual capacity andlung mechanics in surfactant-treated infants. Sem Perinatol. 1992;16:181–185.

5. Clark RH, Gerstmann DR, Null DM Jr, deLemos RA. Prospectiverandomized comparison of high-frequency oscillatory and con-ventional ventilation in respiratory distress syndrome. Pediatrics.1992; 89:5–12.

6. Gerstmann DR, Minton SD, Stoddard RA, Meredith KS, MonacoF, Bertrand JM, Battisti O, Langhendries JP, Francois A, ClarkRH. The Provo multicenter early high-frequency oscillatory ven-tilation trial: Improved pulmonary and clinical outcome in respi-

ratory distress syndrome (see comments). Pediatrics. 1996; 98:1044–1057.

7. White KS, Muelenaer AA Jr, Beam CA, Effmann EL. Determi-nation of functional residual capacity from digital radiographs ofthe normal neonatal chest: Studies in a rabbit model. Am J Roent-genol. 1991; 156:1209–1214.

8. Pratt PC, Klugh GA. A method for the determination of total lungcapacity from posteroanterior and lateral chest roentgenograms.Am Rev Respir Dis. 1967; 96:548–552.

9. Hurtado A, Fray WW. Studies of total pulmonary capacity and itssubdivisions. II. Correlation with physical and radiological mea-surements. J Clin Invest. 1933; 12:807–823.

10. Gildenhorn HL, Hallett WY. An Evaluation of radiological meth-ods for the determination of lung volumes. Radiology. 1965; 84:754–756.

11. Herman PG, Sandor T, Mann BE, McFadden ER, Korngold E,Murphy MA, Mellins HZ. Rapid computerized lung volume de-termination from chest roentgenograms. Am J Roentgenol. 1975;124:477–483.

Lung Volume Measurements in Premature Infants 271

12. Ries AL, Clausen JL, Friedman PJ. Measurement of lung volumesfrom supine portable chest radiographs. J Appl Physiol. 1979;47:1332–1335.

13. Pierce RJ, Brown DJ, Holmes M, Cumming G, Denison DM.Estimation of lung volumes from chest radiographs using shapeinformation. Thorax. 1979; 34:726–734.

14. Gamsu G, Shames DM, McMahon J, Greenspan RH. Radio-graphically determined lung volumes at full inspiration and duringdynamic forced expiration in normal subjects. Invest Radiol.1975; 10:100–108.

15. Barnhard HJ, Pierce JA, Joyce JW, Bates JH. Roentgenographicdetermination of total lung capacity. A new method evaluated inhealth, emphysema and congestive heart failure. Am J Med. 1960;28:51–60.

16. Harris TR, Pratt PC, Kilburn KH. Total lung capacity measured byroentgenograms. Am J Med. 1971; 50:756–763.

17. Loyd HM, String ST, DuBois AB. Radiographic and plethysmo-graphic determination of total lung capacity. Radiology. 1966;86:7–14.

18. Campbell SC. Estimation of total lung capacity by planimetry ofchest radiographs in children 5 to 10 years of age. Am Rev RespirDis. 1983; 127:106–107.

19. Hiller EJ, Kirkpatrick JA, Huang NN. Radiographic determinationof total lung capacity in patients with cystic fibrosis. J Pediatr.1971; 78:435–440.

20. Shephard RJ, Seliger V. On the estimation of total lung capacityfrom chest X-rays. Radiographic and helium dilution estimates onchildren aged 10-12 years. Respiration. 1969; 26:327–336.

21. Salam H, Warwick WJ. Measurement of total lung capacity by aroentgenography-planimetry method in children 4–16 years ofage. Respiration. 1978; 36:177–182.

22. Fumey MH, Nickerson BG, Birch M, McCrea R, Kao LC. Aradiographic method for estimating lung volumes in sick infants.Pediatr Pulmonol. 1992; 13:42–47.

23. Schlesinger AE, White DK, Mallory GB, Hildeboldt CF, Huddles-ton CB. Estimation of total lung capacity from chest radiographyand chest CT in children: Comparison with body plethysmogra-phy. Am J Roentgenol. 1995; 165:151–154.

24. Hlastala MP, Meyer M, Riepl G, Scheid P. Solubility of helium,argon, and sulfur hexafluoride in human bood measured by massspectrometry. Undersea Biomed Res. 1980; 7:297–304.

25. Cander L. Solubility of inert gases in human lung tissue. J ApplPhysiol. 1959; 14:538–540.

26. Schulze A, Schaller P, To¨pfer A, Kirpalani H. Measurement offunctional residual capacity by sulfur hexafluoride in small-volume lungs during spontaneous breathing and mechanical ven-tilation. Pediatr Res. 1994; 35:494–499.

27. Thome U, To¨pfer A, Schaller P, Pohlandt F. The effect of positiveendexpiratory pressure, peak inspiratory pressure, and inspiratorytime on functional residual capacity in mechanically ventilatedpreterm infants. Eur J Pediatr. 1998 (in press).

28. Thome U, To¨pfer A, Schaller P, Pohlandt F. Effect of mean air-way pressure on lung volume during high frequency oscillatoryventilation of preterm infants. Am J Respir Crit Care Med. 1998(in press).

29. Schulze A, Schaller P, Toepfer A, Kirpalani H. Resistive andelastic unloading to assist spontaneous breathing does not changefunctional residual capacity. Pediatr Pulmonol. 1993; 16:170–176.

30. Schaller P, Ma¨dler HJ, Schulze A, Bo¨hme B, Leupold W. EinLamellen-Spirorezeptor fu¨r die Pneumotachographie bei Fru¨hge-borenen und Sa¨uglingen. Z Klin Med. 1985; 40:947–949.

31. Karna P, Wood B, Adams A, Stenzler A. Evaluation of the rela-tionship at varying mean airway pressures between compliance,functional residual capacity and gas exchange during high fre-quency oscillatory ventilation. Pediatr Res. 1994; 35:339A (Ab-stract).

32. Wood B, Karna P, Adams A, Stenzler A. Respiratory mechanicsand functional residual capacity during high frequency ventilation.Pediatr Res. 1994; 35:358A (Abstract).

33. East TD, Andriano KP, Pace NL. Automated measurement offunctional residual capacity by sulfur hexafluoride washout. J ClinMonit. 1987; 3:14–21.

34. Jonmarker C, Jansson L, Jonson B, Larsson A, Werner O. Mea-surement of functional residual capacity by sulfur hexafluoridewashout. Anesthesiology. 1985; 63:89–95.

35. Larsson A, Linnarsson D, Jonmarker C, Jonson B, Larsson H,Werner O. Measurement of lung volume by sulfur hexafluoridewashout during spontaneous and controlled ventilation: Furtherdevelopment of a method. Anesthesiology. 1987; 67:543–550.

36. Vilstrup CT, Bjoerklund LJ, Larsson A, Lachmann B, Werner O.Functional residual capacity and ventilation homogeneity in me-chanically ventilated small neonates. J Appl Physiol. 1992; 73:276–283.

37. Yamamura T, Okamura A, Kikuchi N, Fukuda M, Kemmotsu O.Measurement of functional residual capacity by sulfur hexafluo-ride washout. Masui Jpn J Anesth. 1992; 41:925–931.

38. East TD, Wortelboer PJ, van Ark E, Bloem FH, Peng L, Pace NL,Crapo RO, Drews D, Clemmer TP. Automated sulfur hexafluoridewashout functional residual capacity measurement system for anymode of mechanical ventilation as well as spontaneous respira-tion. Crit Care Med. 1990; 18:84–91.

39. Bland JM, Altman DG. Statistical methods for assessing agree-ment between two methods of clinical measurement. Lancet.1986; 1:307–310.

272 Thome et al.