Original Research

Navigator-Triggered Oxygen-Enhanced MRI withSimultaneous Cardiac and RespiratorySynchronization for the Assessment of InterstitialLung Disease

Francesco Molinari, MD,1,2 Monika Eichinger, MD,2 Frank Risse, PhD,3

Christian Plathow, MD,2,4 Michael Puderbach, MD,2 Sebastian Ley, MD,2

Felix Herth, MD,5 Lorenzo Bonomo, MD,1 Hans-Ulrich Kauczor, MD,2 andChristian Fink, MD2,6

Purpose: To evaluate an optimized method for oxygen-enhanced MRI of the lung, using simultaneous electrocar-diograph (ECG) and navigator triggering. To correlate oxy-gen-enhanced MRI with lung function tests assessingalveolar-capillary gas exchange.

Materials and Methods: A total of 12 healthy volunteers(aged 20–32 years) and 10 patients (aged 37–87 years) withinterstitial lung diseases (ILD) underwent oxygen-en-hanced MRI and pulmonary functional tests (PFTs) assess-ing alveolar-capillary gas exchange. The paradigm room-air–oxygen–room-air was acquired with a nonselectiveinversion-recovery half-Fourier single-shot turbo spin-echosequence (inversion time � 1200 msec; acquisition time �134.5 msec; slice thickness � 20 mm; matrix size � 128 �128), using simultaneous double triggering (navigator plusECG trigger). Cross-correlation was performed in regions ofinterest (ROIs) encompassing both lungs. The number ofoxygen-activated pixels over the total number of pixels inthe ROIs (OAP%) of volunteers and patients was compared.OAP%s were correlated with PFTs.

Results: The mean OAP% of patients was significantlylower than that of volunteers (36.7 vs. 81.7, P � 0.001).OAP% correlated with the transfer lung factor for carbonmonoxide (Tlco) (r � 0.64; P � 0.002), the transfer coeffi-cient (Kco) (r � 0.75; P � 0.001), the arterial partial pres-sure (r � 0.77; P � 0.001), and the saturation (r � 0.70; P� 0.001) of oxygen.

Conclusion: Navigator-triggered oxygen-enhanced MRI ofthe lung may have a potential role in the quantitative as-sessment of lung function in ILD.

Key Words: lung; magnetic resonance (MR); oxygen; diffu-sion; navigator echoJ. Magn. Reson. Imaging 2007;26:1523–1529.© 2007 Wiley-Liss, Inc.

OXYGEN-ENHANCED MRI is a promising tool to inves-tigate lung function (1–4). Following diffusion from thealveoli into the capillaries, molecular oxygen acts as aT1-shortening contrast agent and can be used to visu-alize ventilated lung with functioning oxygen transfer(1,2,4,5). As the paramagnetic effect of molecular oxy-gen is weak, the measured signal intensity (SI) changesin the lung are small. Therefore, subtraction of imagesacquired while the subject breathes room air from thoseacquired while the subject breathes pure oxygen is re-quired. Additionally, cardiac or respiratory functionmay lead to artificial signal variations and signal loss(6). The effects of this signal variability have been suc-cessfully reduced using electrocardiograph (ECG) or re-spiratory triggering (7).

Recently, some investigators have proposed to com-bine ECG and respiratory triggers in oxygen-enhancedMRI of the lung, proving the effectiveness of sequentialdouble triggering in healthy subjects (8). However, theuse of a mechanical pneumatic belt has been indicatedas a limitation for respiratory synchronization (8,9).Additionally, because respiratory and heart rate do notmatch, the sequential activation of the two triggers may

1Department of Bioimaging and Radiological Sciences, Catholic Univer-sity of Rome, Rome, Italy.2Department of Radiology, Deutsches Krebsforchungszentrum (DKFZ),Heidelberg, Germany.3Medical Physics in Radiology, Deutsches Krebsforchungszentrum(DKFZ), Heidelberg, Germany.4Department of Diagnostic Radiology Eberhard-Karls University Tu-bingen, Tubingen, Germany.5Department of Pulmology, Thoraxklinik, University of Heidelberg, Hei-delberg, Germany.6Department of Clinical Radiology, University Hospital Mannheim,Medical Faculty Mannheim-University of Heidelberg.Contract grant sponsor: Deutsche Forschungsgemeinschaft; Contractgrant number: FOR 474.*Address reprint requests to: F.M., MD, Department of Bioimaging andRadiological Sciences, Catholic University of Rome, Rome, Italy, L.go F.Vito n. 1, 00168 Rome, Italy. E-mail: fmolinari@rm.unicatt.itReceived September 16, 2006; Accepted May 3, 2007.DOI 10.1002/jmri.21043Published online 26 October 2007 in Wiley InterScience (www.interscience.wiley.com).

JOURNAL OF MAGNETIC RESONANCE IMAGING 26:1523–1529 (2007)

© 2007 Wiley-Liss, Inc. 1523

delay the subsequent acquisition of image data. Withlarge delays between two sequential triggering events,the long inversion time used in oxygen-enhanced MRImay further contribute to shifting the anatomical dataacquisition out of the targeted respiratory phase, lead-ing to misregistration.

Technical developments in cardiac MRI have led toimproved algorithms for cardiorespiratory synchroniza-tion, implementing navigator echoes (10). Differentlyfrom the pneumatic belt, whose signal is based on ab-dominal wall motion (11), navigator echoes allow fordirect on-line tracking of diaphragm displacement.Therefore, the use of navigators might provide moreefficient respiratory synchronization in clinical oxygen-enhanced MRI of the lung.

In clinical practice, the alveolar-capillary diffusion ofoxygen is estimated globally by the transfer lung factorfor carbon monoxide (Tlco) and the transfer coefficient(Kco). These two functional parameters are useful inpatients with interstitial lung disease to assess the im-pairment of lung function and monitor the evolution ofthe disease (12). In the future, a more accurate evalu-ation of these patients might be obtained by oxygen-enhanced MRI, assessing the local transfer of oxygen inthe lung. So far, few clinical studies have proved acorrelation between the oxygen-induced SI changes inthe lung and Tlco (2,13). In addition, a correlation hasnot been investigated with Kco, a more accurate param-eter for the assessment of lung diffusion, because itdoes not depend on alveolar volumes (14). Therefore,more clinical data on the correlation of oxygen-en-hanced MRI with parameters of lung diffusion seemhighly desirable to assess the potential clinical rele-vance of this imaging technique.

The aim of this study was two-fold: 1) to evaluate anoptimized protocol to perform oxygen-enhanced MRI ofthe lung, using ECG and navigator triggering for simul-taneous cardiac and respiratory synchronization; and2) to correlate oxygen-enhanced MRI with lung functiontests that assess the alveolar-capillary gas exchange.

MATERIALS AND METHODS

Subjects

Before the beginning of the study, the protocol had beenapproved by the institutional ethic review board. Afterthe nature of the procedure had been fully explained,written informed consent was obtained from all volun-teers and patients.

The study population included 12 healthy nonsmok-ing volunteers (nine men and three women; mean age �24.6 years; age range � 20–32 years) and 10 consecu-tive patients with diffuse restrictive pulmonary disease(seven men and three women; mean age � 65.4 years;age range � 37–87 years). Patient diagnoses were: in-terstitial pulmonary fibrosis (IPF, N � 8); nonspecificinterstitial pneumonia (NSIP, N � 1); and sarcoidosis(N � 1). The diagnosis was based on lung biopsy andhigh-resolution CT in all patients.

Imaging System

All MRI examinations were performed on a 1.5-T whole-body scanner (Magnetom Symphony; Siemens, Erlan-

gen, Germany) offering a maximum gradient strength of30 mT/m and a slew rate of 125 T/m/second. Forsignal reception, a combination of the standard bodyphased array coil with spine array coils was used.

Double-Triggering Scheme and Sequence

Simultaneous respiratory and cardiac triggering wasobtained using a commercial software package (2D-PACE; Siemens, Erlangen, Germany) (15,16) availableon the system (Fig. 1). Navigators monitored the posi-tion of the diaphragm dome, allowing a tolerance of2-mm diaphragm mismatch. Image acquisition wasperformed in end-expiration and in diastole.

A nonselective inversion-recovery half-Fourier acqui-sition single-shot turbo spin-echo sequence (IR-HASTE) with an inversion time of 1200 msec was used(5,17). A total of 68 phase-encoding steps were sampledwith a linear ordering scheme. Other imaging parame-ters were: effective TE � 12 msec; interecho spacing �1.98 msec; pixel bandwidth � 660 kHz; section thick-ness � 20 mm; field of view (FOV) � 480 mm; matrixsize � 128 � 128; pixel size � 3.75 � 3.75 mm2; imageacquisition time � 134.5 msec. TR varied accordingly totwo R-R periods. All measurements were obtained withthe subject in supine position, in one coronal planepassing through the pulmonary hila.

MR Ventilation Imaging

Each MR experiment consisted of three ventilationphases according to the paradigm room-air–100%-oxy-gen–room-air. Before the start of the paradigm, the sub-ject’s breathing pattern was checked in real-time byscout navigator monitoring. If the navigator traceshowed regular diaphragm motion, 20 images were ac-quired for each of the three ventilation phases (totalnumber of images � 60). If the subject’s breathing pat-tern was irregular, the ventilation paradigm was ob-tained with 40 images per phase (total number of im-ages � 120). No specific breathing instructions weregiven to the subject before or during the examination.

For oxygen and room-air delivery, a face mask wasapplied to the subject before the MR experiment. Themask included a two-way nonrebreathing valve withthe outlet open to air and the inlet connected to a 3-mtube (internal diameter of 3.4 cm). To allow for oxygen/room-air ventilation switch, the end of the tube wasconnected to and detached from a 60-liter reservoirbag, previously filled and continuously supplied withoxygen at a flow rate of 8–10 liter/minute. Five minuteswere allowed for wash-in and wash-out of oxygen. Thetotal time required for the imaging study, including thetime allowed for the equilibrium of oxygen concentra-tion, ranged from 20 to 30 minutes.

Image Postprocessing and Data Analysis

All images were analyzed using an in-house-developedsoftware based on Interactive Data Language (IDL; Re-search Systems, Boulder, CO, USA) and implementedon a PC platform.

SD maps were automatically computed to evaluatethe variability of the SI in each image series. In case

1524 Molinari et al.

diaphragms appeared wider than two pixels on the SDmaps, retrospective image selection was performed toreduce diaphragm mismatch. A representative pixelwithin the diaphragm was chosen to calculate the me-dian value of the signal-time course. Assuming no mo-tion, the signal only changes because of the oxygenenhancement. Thus, images with signal values above orbelow the median � P � median were considered asmismatching and, therefore, removed. The P value wasset to 0.1 or to 0.2 if the remaining number of images atP � 0.1 was considered as too low for the subsequentcomputation. Therefore, a relative median signal errorof �10% or �20% was accepted. This error does notcorrespond directly to the error of the diaphragm posi-tion, since it reflects the variance of the signal itself(noise) and, additionally, the variance caused by thediaphragm’s movements. The whole procedure of imageselection was repeated several times by choosing differ-ent representative pixels within the diaphragm. This

allowed for collecting the highest number of imageswith matching diaphragm.

Selected images were processed using cross-corre-lation analysis (18). The time response function ofeach pixel was correlated to the input reference pat-tern of the ventilation paradigm in regions of interest(ROIs) encompassing both lungs. The confidence levelthreshold for the correlation was set to 0.001. Theresulting correlation coefficient (cc) expressed thesimilarity between the temporal SI variations in time-course data and the user-specified reference pattern.In the correlation maps, only pixels with a cc in therange of 0.5 (statistical cutoff level) to one weredisplayed. No spatial filtering was used. The coverageof the maps was assessed computing the percentageof pixels with a cc higher than 0.5, over the totalnumber of pixels in the ROIs (OAP% � % of oxy-gen activated pixels; ROI � both lungs). Image post-processing, including both retrospective selection

Figure 1. Double-triggering technique (navigator and ECG). After an initial learning phase, in which five complete respiratorycycles are monitored by navigator pulses, the algorithm proposes acceptance windows (AW) enclosing the end-expiratorysegment of the diaphragm trace (gray boxes). The horizontal edges (w) of the boxes are automatically calculated. The centralposition (c) was manually adjusted at 5% to 10% of the vertical interval of the tidal breathing curve (end-expiration � 0%,end-inspiration � 100%). The vertical edge (h) was set to 2 mm (spatial tolerance between different end-expiratory diaphragmlevels). In the imaging phase (see the figure), continuous navigator pulses monitor diaphragm position and generate triggeringevents. Navigators with triggering potentials (in black) are only those preidentified by the algorithm as having a fixed time delayfrom triggering R waves. The 180° inversion pulse is triggered only if one of those selected navigator echoes matches anacceptance window, thus securing simultaneous cardiac and respiratory triggering. Once a double-triggering event occurs,navigator monitoring is immediately suspended and the imaging sequence bloc is started. A time delay (TD) of 400 msec isrequired by the system to automatically resume navigator monitoring after the readout. The timing parameters refer to those notdepending on the time of the R-R cycle (TRR). To fit the readout in the diastolic phase, an additional time delay before the inversionpulse (not shown in the figure) was necessary in some subjects according to their heart rate.

Double-Triggered Oxygen-Enhanced MRI in ILD 1525

and cross-correlation, required approximately 30minutes.

Pulmonary Functional Tests

Pulmonary functional tests (PFTs) were measured in 10of 12 volunteers and in all 10 patients, according to theAmerican Thoracic Society standards (19–21). Tlco(mmol/minute/kPa) and Kco (mmol/minute/kPa/liter)were measured using the single-breath method (Mas-terScreen Diffusion; Viasys Healthcare, Wurzburg, Ger-many). The partial pressure of oxygen (pO2 [mmHg])and oxygen saturation (O2Sat [%]) in peripheral arterialblood were measured with a blood gas analyzer (ABL520; Radiometer, Copenhagen, Denmark). The mediantime interval between PFTs and oxygen-enhanced MRIwas eight days for patients (range � one to 70 days) and42 days for volunteers (range � 10–180 days).

Statistical Analysis

All results were expressed as means with 95% confi-dence interval (CI). A Mann-Whitney test (nonparamet-ric statistic for unpaired groups) was performed to testthe null hypothesis that the mean OAP% of the volun-teer group was equal to that of the patient one. TheSpearman correlation of OAP% and PFTs was calcu-lated for all subjects. A P value of less than 0.05 wasconsidered as significant. Statistical analysis was per-formed with Excel 2003 (Microsoft Corp., Redmond,WA, USA), Statistica 6 (StatSoft, Inc., Tulsa, OK, USA),and GraphPad Prism 4.00 (GraphPad Software, SanDiego, California, USA).

RESULTS

All examinations were successfully completed and theimages were eligible for further analysis.

Two examples of the ROIs used to enclose both lungsand corresponding cross-correlation maps from a vol-unteer and a patient with interstitial pulmonary fibro-sis are shown in Fig. 2. Because no postprocess filteringwas used, the original spatial resolution of the relativeimages is maintained in both cross-correlation maps.In the volunteer (Fig. 2b), both lungs are almost com-pletely covered by oxygen-activated pixels. The predom-inance of yellow pixels in the map indicates a highcorrelation between the measured SI and the input ref-erence function, confirming oxygen-induced SI changesin the lung. Slightly lower, but still statistically signifi-cant, correlation is shown along the outer edge of bothlungs. Few pixels in the map are not colored becausethe computed ccs were lower than 0.5 (i.e., statisticalcutoff level). The overall percentage of oxygen-activatedpixels over the total number of pixels in the ROIs(OAP%) was 94.9%. In the patient (Fig. 2c), thin, irreg-ular bands of high SI, caused by fibrosis, are evident inthe lung. In addition, the minor fissure is thickened.Diaphragm position is also higher on the right side thanon the left, for asymmetric reduction of lung volumes.The relative cross-correlation map (Fig. 2d) shows nu-merous defects indicating either absence of correlationor ccs below the statistical cutoff level (OAP% � 36 %).

Mean data from volunteer and patient groups aresummarized in Table 1. The percentage of images ret-rospectively selected for improving respiratory synchro-nization was in the range of 35% to 40%, without sig-nificant differences between volunteers and patients.Mean OAP% from the patient group was significantlylower than that from the volunteer group (36.7 � 16.6vs. 81.7 � 7.1, P � 0.001; Fig. 3). Mean PFTs showedsignificant differences between volunteers and pa-tients.

There was a good, statistically significant correlationbetween all PFTs and OAP% (Table 2, Fig. 4). All ccswere in the range of 0.64–0.77. The correlation washigher with Kco than with Tlco (0.75 vs. 0.64; Table 2,Fig. 4a and b). OAP% also correlated well with pO2 andO2Sat (0.77 and 0.7; Table 2, Fig. 4c and d).

Figure 2. Two examples, from a volunteer (a,b) and a patientwith interstitial pulmonary fibrosis (c,d), showing the ROIsenclosing both lungs (a,c) and the relative cross-correlationmaps (b,d). In the volunteer, both lungs are almost completelycovered by oxygen-activated pixels (OAP% � 94.4%). In thepatient, the cross-correlation map shows numerous defectsindicating either absence of correlation or correlation coeffi-cients below the statistical cutoff level (OAP% � 36%). SeeResults for further details.

Table 1Mean Data � 95% Confidence Intervals Calculated From theVolunteer and the Patient Group

Data Volunteers Patients P

% of selected images 35 � 8.2 39.2 � 6.3 nsOAP% 81.7 � 7.1 36.7 � 16.6 0.001Tlco (mmol/minute/kPa) 11.7 � 1.4 5 � 1.5 �0.0001Kco (mmol/minute/kPa/liter) 1.9 � 0.1 1.1 � 0.2 �0.0001pO2 87.9 � 3.7 69.7 � 3.6 0.0001O2Sat 97.1 � 0.5 94 � 0.9 0.0003

OAP% � percentage of pixels with a correlation coefficient higherthan 0.5 (statistical cutoff level in the cross-correlation analysis) overthe total number of pixels in the ROIs, pO2 � partial pressure ofoxygen in peripheral blood (mmHg), O2Sat � oxygen saturation inperipheral blood (%), ns � not significant.

1526 Molinari et al.

DISCUSSION

In this study, we have demonstrated the feasibility of anew technical method to perform oxygen-enhancedMRI of the lung, using navigator and ECG triggering forsimultaneous respiratory and cardiac synchronization.Results of quantitative analysis of ventilation maps cor-related well with the Tlco, the Kco, the partial pressure,and the saturation of oxygen in peripheral blood.

Double triggering based on ECG and navigator pulsesis an advanced MR technique to obtain cardiac andrespiratory synchronization (9). Respiratory synchroni-zation obtained by pneumatic belt assumes thatchanges in the circumference of the abdomen reflectthose of diaphragm position. This may not be true insubjects with an abdominal breathing pattern or inpatients (11). Navigator pulses monitor directly, and inreal-time, the vertical position of the diaphragm. There-fore, navigator triggering is also virtually insensitive tochanges in respiratory rate. When navigator and ECGtriggering are combined in a sequential fashion (cardiacafter respiratory) (8), respiratory synchronization may

be lost waiting for cardiac trigger activation. In fact, thedifference between heart and respiratory rate yields un-matched R-waves and end-expiratory phases. A vari-able delay occurs after end-expiration is detected andbefore an R-wave triggers the sequence. If the delayextends for several milliseconds, the diaphragm maymove as the subject is free-breathing. By prospectivelyselecting navigators with triggering potentials, namelythose with a fixed time delay from triggering R-waves,the two triggering conditions are interlocked. This tech-nique reduces the number of triggering events. How-ever, the increased interval between the long inversionpulses used in oxygen-enhanced MRI of the lung alsoallows for complete magnetization recovery.

Despite these technical advances, respiratory syn-chronization between IR-HASTE images was less effec-tive than expected. To achieve correspondence of dia-phragms, retrospective selection of images wasrequired. The mean number of selected images in thepatient and volunteer groups was not significantly dif-ferent. This result was surprising, because, consideringthe pathological limitation of breathing these patientsmay have because of restrictive disease, we would haveexpected a poorer performance of navigators in thisgroup. On the contrary, in both groups the number ofimages matching the same respiratory phase wasequally limited. This suggests a different mechanism,other than a pathological breathing pattern, to explainthe unsatisfactory respiratory synchronization ob-tained before postprocessing.

Figure 3. Comparison of the mean percentage of oxygen-ac-tivated pixels (OAP%) between volunteers and patients. Alldots are aligned according to volunteer (E) or patient (�) group.Horizontal larger lines indicate mean values. Error bars indi-cate SD. Normality test proved a normal distribution of OAP%values for both volunteers and patients. Nonparametric Mann-Whitney test was used to compare mean OAP% because anoutlier value was present in the volunteer group (OAP% �50%). Mean OAP% from the patient group was significantlylower than that from volunteers (36.7 � 16.6 vs. 81.7 � 7.1;P � 0.001).

Table 2Correlations Between the Percentage of Oxygen-Activated Pixels(OAP%) and the Pulmonary Functional Tests (PFTs)*

Pulmonary functional testSpearmancorrelation

P

Tlco (mmol/minute/kPa) 0.64 0.002Kco (mmol/minute/kPa/liter) 0.75 �0.001pO2 0.77 �0.001O2Sat (%) 0.70 �0.001

*Note that the correlations were calculated considering PFTs in acontinuous distribution from normal to pathologic values. Therefore,each correlation coefficient refers to all subjects (see also Fig. 4 forgraphical representation of data).pO2 � partial pressure of oxygen in peripheral blood (mmHg),O2Sat � oxygen saturation in peripheral blood.

Figure 4. Correlation between the percentage of oxygen-acti-vated pixels (OAP%) and the pulmonary functional tests(PFTs): Tlco (a), Kco (b), pO2 (c), and O2Sat (d). Circles indicatepatients. Crosses indicate volunteers. Note that the correla-tions were calculated considering patients and volunteers intothe same graph, as both PFTs and OAP% represent numericalvariables arranged in continuous scales (i.e., PFTs vary typi-cally from normal to pathologic values).

Double-Triggered Oxygen-Enhanced MRI in ILD 1527

Dietrich et al (8) have reported higher percentage oftriggering success using less advanced technical equip-ment (nonsimultaneous double triggering and mechan-ical pneumatic belt). There is a major difference fromthis study that could explain the lower performance ofour triggering routine and their better results. The sub-jects of their study were asked to suspend breathing inevery end-expiratory phase for two to 5.5 seconds. Dur-ing these end-expiratory periods, all imaging phaseswere allowed, including double triggering, the inversionpulse, and the readout, while diaphragm position waskept stable. In our study, we did not train the subjectsbecause our aim was to evaluate the navigator tech-nique in a full free-breathing fashion. Also, to assessthe correlations with the PFTs and the clinical feasibil-ity of navigator-triggered oxygen-enhanced MRI, pa-tients were included in our study. In this setting, con-tinuous voluntary adaptation of breathing during thewhole examination time would have been hardly feasi-ble, if not unrealistic. In contrast, Dietrich et al (8) onlyexamined healthy subjects. Similar to our study, theirtriggering method had a poor performance in one vol-unteer (66–83% of discarded images). This might alsobe explained by a poor compliance of that subject to theadaptation of breathing. Finally, their volunteer study(8) was performed for a different purpose, the prelimi-nary evaluation of a multislice technique.

For the assessment of lung enhancement, cross-cor-relation was used. This statistical method has beenpreviously applied in oxygen-enhanced MRI of the lung(18) and provides additional advantages over subtrac-tion imaging (3,13). In the latter, the multiple measure-ments of each ventilatory phase are averaged and thensubtracted (i.e., 100%-oxygen mean SI vs. room-airmean SI). For instantaneous physiological changes ofthe heart rate, signal fluctuations may occur in thetime-intensity curve of the lung. Mean signal intensitiesdo not reflect this signal variability. Conversely, usingcross-correlation the steadiness of the signal in eachventilatory phase is also assessed. Therefore, in thecorrelation maps of this study, a differentiation be-tween pixels with statistically-proven signal enhance-ment against pixels showing unreliable signal varia-tions was achieved.

Since molecular oxygen has a paramagnetic effect inthe lung after diffusing through the alveolar-capillarybarrier and dissolving into capillary blood (22), the cor-relation with the PFTs assessing alveolar-capillary gasexchange was expected. However, in this study, a newobservation has been made. The correlation was higherwith Kco than with Tlco. Considering the differencebetween those two parameters, the higher correlationwith Kco can be explained. In clinical practice, to assessthe diffusing capacity of the lung, both Tlco and Kco areused. However, several clinical conditions in whichthese two functional parameters might be discordanthave been described (14). This is because Kco evaluatesthe rate of carbon monoxide removal from alveolar gas,corrected for the alveolar volume. Patients with pneu-monectomy represent the typical example in which thedifference between Tlco and Kco is considerably large.In these patients, the reduction of Tlco by one-halfmight be interpreted as pathological, whereas Kco pro-

vides the real estimation of diffusing capacity of theresidual lung. Therefore, to selectively assess the dif-fusing capacity of the lung, Kco is commonly consid-ered a very sensitive parameter (14). In our study, thelower correlation with Tlco most probably expresses thelower accuracy of this functional parameter in assess-ing lung diffusion.

In two previous works (2,13), a higher correlationwith Tlco has been reported. To obtain reliable data, inour study, all lung parenchyma displayed in the coro-nal slice was processed using large ROIs. A pixel-by-pixel analysis, using the nominal spatial resolution ofindividual images and without filtering the data, wasperformed. Conversely, to assess lung enhancement inboth previous studies (2,13), only three small ROIs perlung were used. The parameters computed for the cor-relation were also different. Muller et al (2) calculatedthe relative mean SI enhancement, whereas Ohno et al(13) calculated the maximum mean relative enhance-ment ratio. Assuming proportionality between signalenhancement and Tlco, those quantitative parameterswere correlated with the latter. Considering the afore-mentioned limitation of using signal average, we pre-ferred to test signal enhancement with cross-correla-tion. Because the value of Tlco is dependent on theamount of lung tissue in which diffusion occurs, wecomputed the percentage of “oxygen-activated” lung pa-renchyma, confirmed by cross-correlation. Instead ofsignal enhancement, which might depend on severalfactors (value of the inversion time, etc.), that semi-quantitative parameter was correlated with PFTs. More-over, the population investigated by both studies (2,13)was different. Ohno et al (13) examined lung cancerpatients, whereas Muller et al (2) investigated patientswith various pulmonary diseases. Our study popula-tion included patients and volunteers. Finally, our as-sessment was extensively correlated with other PFTspertaining lung diffusion, whereas their analysis waslimited to Tlco. For all these differences, the results ofthese studies might not be fully comparable.

There are some limitations in this study. Image ac-quisition was performed in a single coronal plane. Toinvestigate the performance of the double-triggeringmethod, numerous measurements were performed inall subjects. To avoid excessive prolongation of the ex-periments, which would have implied less subject’s co-operation and poor image quality, the examination wasnot extended to other planes. Although in the patientsthe distribution of the disease was checked before oxy-gen-enhanced MRI, to cover more lung parenchyma, amiddle coronal location of the imaging plane was priv-ileged. In some patients, the correlation with the PFTsmight have been influenced by the incomplete lungcoverage. Most probably, interleaved schemes for mul-tislice acquisition (8) will increase the clinical feasibilityof oxygen-enhanced MRI of the lung. However, thesemultislice techniques have not been implemented incommercial MR sequences yet.

Similar to Dietrich et al (8), the performance of thedouble-triggering method was assessed by the percent-age of images discarded by retrospective image selec-tion. Because of the limitation of examining patientswithin a reasonable time, a comparison with different

1528 Molinari et al.

triggering methods was not performed. Therefore, anextensive evaluation of the efficiency of simultaneousdouble triggering based on navigator pulses in compar-ison with cardiac or respiratory triggering, with thelatter also obtained by pneumatic belt, needs to beassessed.

The mean age of the healthy subjects was lower thanthat of the patients. OAP% and PFTs may both vary withage. The age difference might have had some influenceon the correlation between those parameters.

Finally, in our triggering scheme, automatic algo-rithms for retrospective reacquisition of images display-ing diaphragm mismatch were not available. As antici-pated by Dietrich et al (8), using these algorithms,respiratory synchronization might be further improved.However, since the total imaging time might be sub-stantially increased (8), further research is required toprove the advantages and the clinical feasibility of thistechnique.

In conclusion, the combination of navigator and car-diac triggering for oxygen-enhanced MRI of the lung ispromising and may have a potential role in the quanti-tative assessment of lung function in patients with in-terstitial lung disease. Further studies are required todetermine the efficacy of this combined triggeringmethod.

REFERENCES1. Edelman RR, Hatabu H, Tadamura E, Li W, Prasad PV. Noninvasive

assessment of regional ventilation in the human lung using oxygen-enhanced magnetic resonance imaging. Nat Med 1996;2:1236–1239.

2. Muller CJ, Schwaiblmair M, Scheidler J, et al. Pulmonary diffusingcapacity: assessment with oxygen-enhanced lung MR imaging pre-liminary findings. Radiology 2002;222:499–506.

3. Ohno Y, Hatabu H, Higashino T, et al. Oxygen-enhanced MR imag-ing: correlation with postsurgical lung function in patients withlung cancer. Radiology 2005;236:704–711.

4. Chen Q, Levin DL, Kim D, et al. Pulmonary disorders: ventilation-perfusion MR imaging with animal models. Radiology 1999;213:871–879.

5. Loffler R, Muller CJ, Peller M, et al. Optimization and evaluation ofthe signal intensity change in multisection oxygen-enhanced MRlung imaging. Magn Reson Med 2000;43:860–866.

6. Mai VM, Chen Q, Li W, Hatabu H, Edelman RR. Effect of respiratoryphases on MR lung signal intensity and lung conspicuity using

segmented multiple inversion recovery turbo spin echo (MIR-TSE).Magn Reson Med 2000;43:760–763.

7. Vaninbroukx J, Bosmans H, Sunaert S, et al. The use of ECG andrespiratory triggering to improve the sensitivity of oxygen-enhancedproton MRI of lung ventilation. Eur Radiol 2003;13:1260–1265.

8. Dietrich O, Losert C, Attenberger U, et al. Fast oxygen-enhancedmultislice imaging of the lung using parallel acquisition tech-niques. Magn Reson Med 2005;53:1317–1325.

9. Oshinski JN, Hofland L, Mukundan SJr, Dixon WT, Parks WJ,Pettigrew RI. Two-dimensional coronary MR angiography withoutbreath holding. Radiology 1996;201:737–743.

10. Firmin D, Keegan J. Navigator echoes in cardiac magnetic reso-nance. J Cardiovasc Magn Reson 2001;3:183–193.

11. Plathow C, Zimmermann H, Fink C, et al. Influence of differentbreathing maneuvers on internal and external organ motion: use offiducial markers in dynamic MRI. Int J Radiat Oncol Biol Phys2005;62:238–245.

12. Gross TJ, Hunninghake GW. Idiopathic pulmonary fibrosis. N EnglJ Med 2001;345:517–525.

13. Ohno Y, Hatabu H, Takenaka D, Van Cauteren M, Fujii M, Sug-imura K. Dynamic oxygen-enhanced MRI reflects diffusing capacityof the lung. Magn Reson Med 2002;47:1139–1144.

14. Hughes JM. The single breath transfer factor (Tl,co) and the trans-fer coefficient (Kco): a window onto the pulmonary microcircula-tion. Clin Physiol Funct Imaging 2003;23:63–71.

15. Zech CJ, Herrmann KA, Huber A, et al. High-resolution MR-imag-ing of the liver with T2-weighted sequences using integrated par-allel imaging: comparison of prospective motion correction andrespiratory triggering. J Magn Reson Imaging 2004;20:443–450.

16. Klessen C, Asbach P, Kroencke TJ, et al. Magnetic resonance im-aging of the upper abdomen using a free-breathing T2-weightedturbo spin echo sequence with navigator triggered prospective ac-quisition correction. J Magn Reson Imaging 2005;21:576–582.

17. Chen Q, Jakob PM, Griswold MA, Levin DL, Hatabu H, EdelmanRR. Oxygen enhanced MR ventilation imaging of the lung. MAGMA1998;7:153–161.

18. Mai VM, Tutton S, Prasad PV, et al. Computing oxygen-enhancedventilation maps using correlation analysis. Magn Reson Med2003;49:591–594.

19. Lung function testing: selection of reference values and interpreta-tive strategies. American Thoracic Society. Am Rev Respir Dis1991;144:1202–1218.

20. Miller A. Lung function testing: selection of reference values andinterpretative strategies. Am Rev Respir Dis 1992;146:1368–1369.

21. American Thoracic Society. Single-breath carbon monoxide diffus-ing capacity (transfer factor). Recommendations for a standardtechnique—1995 update. Am J Respir Crit Care Med 1995;152:2185–2198.

22. Jakob PM, Wang T, Schultz G, Hebestreit H, Hebestreit A, Hahn D.Assessment of human pulmonary function using oxygen-enhancedT(1) imaging in patients with cystic fibrosis. Magn Reson Med2004;51:1009–1016.

Double-Triggered Oxygen-Enhanced MRI in ILD 1529