If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
MR imaging of the chest has been traditionally challenging and difficult. Shortcomings for thoracic MR imaging are motion artifacts related to breathing and heart and vascular pulsation, susceptibility artifacts associated to air–tissue interfaces, and low proton density in both lungs creating low signal in all pulse sequences.
Improvements in MR imaging systems, including more powerful gradients and phased-array coils, development of fast imaging techniques, such as echo-planar sequences (EPI), and application of parallel imaging, have made it possible to increase the clinical applications of thoracic MR imaging, although it is still far from being a first-line imaging test in pulmonary and mediastinal pathology. In a similar manner, cardiovascular MR imaging has developed in the last few years, with more clinical impact than pulmonary MR imaging.
In the era of functional imaging, diffusion-weighted imaging (DWI) has been proposed as a cancer biomarker.
DWI allows the analysis of tissue characteristics based on the diffusivity of water molecules within the tissues. Although it was first used to detect acute cerebral ischemia, the use of DWI outside the brain has been possible in the last few years because of the previously mentioned technologic developments. Despite these advances, its use in the chest is still very challenging because of the high sensitivity of DWI to artifacts. Because of this, almost all the clinical studies of DWI in the chest have been performed in 1.5-T magnets. In addition, different methods of acquisition and quantification of DWI have been used. All of these facts have limited the clinical use of DWI in the thorax, with scarce clinical experience, mostly limited to detection and characterization of pulmonary nodules and mediastinal lymph nodes. Ongoing research with DWI in such areas as cardiac imaging and pulmonary ventilation makes DWI a potential clinical imaging tool in different areas and systems of the chest, which should be fully developed in the coming years.
This article clarifies which are the most appropriate sequences and technical adjustments for the different chest applications of DWI, including its use in 3-T magnets. Current realistic and potential clinical applications of DWI in the lungs, mediastinum, pleura, and heart are also analyzed.
Technical considerations
DWI is an MR imaging technique sensitive to the Brownian molecular motion of spins.
described that the presence of a magnetic field gradient during an MR imaging spin-echo (SE) experiment results in a signal attenuation because of the molecular diffusion of the spins. In 1965, Stejskal and Tanner
proposed an MR imaging sequence to quantify the diffusion coefficient (D) in an MR imaging experiment. In their experiment, a pair of additional gradient pulses was inserted into a pulse sequence, the so-called “Stejskal–Tanner diffusion gradients.” This pulse sequence is still in use today in most DWI experiments, although some modifications have been proposed for moving organs, such as the heart.
A new parameter (b, measured in seconds per square millimeter), is derived from the Stejskal–Tanner experiment, to control the image contrast in diffusion. This parameter is mainly controlled by the area under the two gradient lobes and the separation between them is used to weight diffusion. When higher b values are applied, the signal from the molecules that suffer a higher displacement is lost, with only the signal from those molecules with less displacement remaining.
The early DWI experiments were performed in stimulated-echo and SE pulse sequences.
However, these pulse sequences required very long acquisition times of several minutes to acquire a single multislice data set, because they filled the required raw-data line by line. Therefore, these slow sequences were very prone to motion artifacts, which limited their usefulness in clinical applications, mainly in moving organs, such as the chest. Nowadays, the most extended pulse sequence for DWI is the single-shot (SS) SE EPI sequence.
This sequence is relatively insensitive to macroscopic patient motion because of its very fast readout of the complete image data, within about 100 ms. It has become the standard technique for DWI and diffusion tensor imaging (DTI) not only for the brain but also for body applications.
Unfortunately, EPI images frequently suffer from gross geometric distortion in the presence of B0 inhomogeneities because of the accumulation of phase error during the long echo train length (ETL).
This error is accumulated in phase acquisition direction, limiting the achievable resolution to maintain the geometric distortion under control. These distortions are particularly important in regions prone to magnetic susceptibility, such as bone–soft tissue interfaces or those structures in contact with air-filled spaces, as occurs in the chest.
To avoid the artifacts associated to SS EPI acquisitions, different strategies have been proposed. The most sensible one is to segment the ETL of the EPI acquisition in different shots, reducing the phase error accumulated during different readouts. Although this approach has fewer geometric artifacts, the acquisition time increases proportionally to the number of EPI shots. Besides, these sequences are more prone to motion artifacts, making it necessary to apply motion correction techniques, such as navigation echoes.
This sequence organizes the segmented (PROPELLER) acquisition in a radial way around the center of k-space. This approach has the advantage that the results are less sensible to motion artifacts. A different strategy to reduce the ETL without increasing the acquisition time is to apply parallel imaging, where the phase encoding lines that are not acquired are recovered using the sensitivity profile of phased array coils.
Recent innovations in hardware and acquisition techniques have substantially improved the suitability of EPI for chest DWI. Improved gradient systems with reduced eddy-current effects have allowed faster EPI readout, which can decrease geometric distortions. Moreover, new gradient technology, reaching gradient strength of 80 mT/m, makes it feasible to acquire DWI with a b value up to 1000 s/mm2 and an echo time (TE) under 45 milliseconds, with an acquisition matrix of 128 × 128.
Another important aspect related to the combination of EPI readout with DWI is the intensity of fat signal for very high b values. Fat signal has a very low diffusion coefficient, which makes it very relevant for high b values. However, the difference in precession frequency between the water and the fat produces a water-fat shift of several voxels in the phase encoding direction of the EPI readout. Because of both factors, the fat signal usually overlaps on the studied anatomy making it mandatory to apply fat suppression techniques for more accurate apparent diffusion coefficient (ADC) estimation. When studying the chest, the short tau inversion recovery (STIR) approach has been most commonly used as a fat suppression technique in such sequences as DWI with background suppression (DWIBS). The main problem of sequences using STIR is the low signal to noise ratio (SNR) caused by water signal reduction after the inversion pulse. To overcome this problem, different spectral fat suppression techniques, such us spectral presaturation inversion recovery (SPIR) and spectral selection attenuated inversion recovery, have been proposed, because of their superior SNR to acquisitions using STIR (Fig. 1).
Fig. 1Differences in DWI using STIR and SPIR acquisition on a 3-T magnet. Two DWI images were acquired in the same patient affected by an epidermoid carcinoma (arrows) using the same b value (1000 s/mm2). (A) DWI with STIR (DWIBS) and (B) DWI with spectral fat suppression. Spectral fat suppression DWI has a higher signal-to-noise ratio compared with the DWIBS sequence.
To solve the lack of spatial resolution of DWI sequences, the use of higher field magnets as 3 T has been proposed for body applications. For example, a signal improvement of 50% has been reported in kidney studies when comparing 3 T with 1.5 T within the same acquisition time.
The increase of signal of 3-T magnets may be used to obtain higher resolution or to reduce scan time. The acquisition problems inherent to DWI increase in 3-T magnets, because of higher magnetic field variation and susceptibility artifacts, which produce image distortion, and SAR limitations, which make fat suppression diffcult.
These limitations can be overcome using appropriately higher strength of the gradient systems of 3-T scanners in combination with parallel imaging and advanced fat suppression sequences.
recently reported the first clinical series of DWI performed on a 3-T magnet, with satisfactory evaluation of 57 patients with malignant pleural mesothelioma (MPM), although ADC quantification could not be obtained in seven patients because of image distortion.
Fig. 2Pulmonary metastasis of renal carcinoma at 3-T magnet. (A) Respiratory-triggered SS EPI DWI sequence with spectral fat suppression and a b value of 900 s/mm2 nicely depicts a metastasis in the upper lobe of left lung. (B) Black-blood STIR TSE shows the lesion similarly to DWI.
The authors’ standard sequences for DWI of the chest at 1.5- and 3-T magnets are detailed in Table 1; their sequence recommendation when possible is as follows:
•
SS SE EPI
•
Phased array surface coil
•
b-values: several values between 0 and 100 s/mm2 until 1000 s/mm2
•
Field of view: 320–400
•
Parallel Imaging acceleration factor of 2
•
Pixel resolution 2.5 × 2.5 × 7 mm3
•
Spectral fat suppression
•
Number of slices, 24
•
TR: 5000 ms
•
TE: 53 milliseconds (shortest)
•
Respiratory triggered
•
Three orthogonal motion probing gradients.
Table 1DWI sequences performed at our centers at 1.5- and 3-T magnets
acquisition sequence, it can be derived that the signal attenuation caused by DWI has an exponential behavior, modulated by the control sequence parameter, b value, and the diffusion properties of the tissue. In the presence of single water compartment the diffusion signal can be expressed as:
where S(b) represents the acquired signal, S0 is the signal taking into account the T1 and T2 relaxation effects. When more than a single DWI is acquired the diffusion coefficient D can be estimated as
There are many parameters that can affect the in vivo measured diffusion coefficient, such as the presence of cell membranes and organelles or blood flow along the vessels. For these reasons, the diffusion coefficient is referred to as ADC.
To isolate the effect of the blood flow from the estimation of the diffusion coefficient, Le Bihan and colleagues
proposed the Intra Voxel Incoherent Motion (IVIM) model of the diffusion signal. This model separates the diffusion signal decay in two different diffusion compartments. For low b values, between 0 and 100 s/mm2, the diffusion signal experiments a fast decay because of the blood flow along the microvasculature, whereas for higher b values, over 100 s/mm2, the signal decay corresponds to the conventional diffusion of the tissue, following this equation:
where f represents the perfusion fraction, D is the perfusion free diffusion coefficient, and D* is the perfusion diffusion coefficient.
This model has been successfully evaluated in several pathologic conditions, such as brain tumors
compared the diffusion coefficient (D), estimated from an IVIM DWI sequence with 10 different b values (0, 10, 20, 30, 50, 80, 100, 200, 400, and 800 s/mm2), with the ADC value obtained from a separate DWI measurement with four b values (0, 200, 400, and 800 s/mm2). In this series, ADC values were significantly higher than D in both cirrhotic and noncirrothic patients. This difference between ADC and D values is probably secondary to the perfusion effect in the diffusion signal decay. Therefore, the IVIM approach allows one to avoid the perfusion effects (Fig. 3).
Fig. 3IVIM model applied to a renal carcinoma pulmonary metastasis (same case as Fig. 2). (A) Parametric D map shows a nodule in the upper left lobe with restricted diffusion. (B) Comparison of diffusion signal decay within the lesion using either the IVIM model estimation (solid line) or the conventional ADC estimation from the monoexponential model (dotted line). The effect of the perfusion contribution to the ADC estimation can be appreciated as fast signal decay in the lower b values caused by perfusion effect. The results of both models show a clear difference between the conventional ADC and the D measurements.
Another problem involving the diffusion signal is the macroscopic movement produced by the respiratory motion and heartbeat, which are critical in thoracic acquisitions. To avoid this movement, different strategies have been proposed, which have been carefully studied in the liver.
Comparison and reproducibility of ADC measurements in breathhold, respiratory triggered, and free-breathing diffusion-weighted MR imaging of the liver.
Respiratory-triggered versus breath-hold diffusion-weighted MRI of liver lesions: comparison of image quality and apparent diffusion coefficient values.
Comparison and reproducibility of ADC measurements in breathhold, respiratory triggered, and free-breathing diffusion-weighted MR imaging of the liver.
reported good agreement in the estimation of ADC value comparing breathhold and free-breathing sequences, whereas respiratory-triggered acquisitions systematically showed an overestimation in the ADC values. In contrast, Kandpal and colleagues
Respiratory-triggered versus breath-hold diffusion-weighted MRI of liver lesions: comparison of image quality and apparent diffusion coefficient values.
found good agreement in the ADC values acquired with respiratory-triggered and breathhold strategies for normal liver and focal lesions, although respiratory-triggered acquisitions showed higher SNR in normal liver and higher contrast-to-noise ratio between normal liver and focal lesion than with breathhold sequences. Finally, in another report, Kwee and colleagues
also studied the effect of the heart motion on DWI of the liver, showing a strong degradation of those images acquired during the heart systole because of the effect of the heart movement. Although the effect of cardiac movement in the ADC estimation was not studied in this paper, the authors suggest that the signal loss in DWI images should affect the ADC estimation. Similar approaches have been applied in DWI of the chest with current lack of consensus. Most of the time the use of respiratory trigger improves the quality of DWI sequences compared with those using breathholding, according to our experience. A cardiac trigger is also useful to avoid pulsation artifacts, but it is not always necessary except in the case of lesions located immediately around the heart or in dedicated cardiac acquisitions, because it is time consuming (Fig. 4).
Fig. 4Synchronization on chest DWI. Five different approaches under different strategies of motion compensation of the same DWI sequence are shown, using the same b value (800 s/mm2) at a 3-T magnet, in a patient with small cell lung cancer (SCLC). (A) Free-breathing. (B) Breathhold. (C) Breathhold and cardiac trigger. (D) Respiratory trigger. (E) Respiratory and cardiac trigger. Higher signal of the mediastinal mass is shown in acquisitions with cardiac and respiratory control (C, E), caused by reduction of the signal loss on DWI related to respiratory and cardiac movement. On the contrary, in acquisitions without cardiac synchronization (A, B, and D), a loss of signal within the tumor is evident because of the cardiac movement effect over the DWI signal.
As in other organs, there is a lack of standardization in region of interest analysis, which is prone to errors because it is operator-dependent. The number and size of region of interest varies from series to series. There is also no consensus as to whether or not it is more appropriate to use the mean or minimal ADC value. Areas of necrosis and those with susceptibility artifacts should be avoided. Misregistration of the trace images may cause variations of the ADCs, which may be partially solved using coregistration software.
Clinical applications
Detection of Pulmonary Nodules
Pulmonary MR imaging with conventional sequences, including STIR, has demonstrated good results in the detection of pulmonary nodules.
MRI of the lung: value of different turbo spin-echo, single-shot turbo spin-echo, and 3D gradient-echo pulse sequences for the detection of pulmonary metastases.
In comparison with multidetector CT (MDCT), STIR sequences have achieved better results with sensitivities superior to 90% for nodules measuring 3 mm in size or larger.
There are scarce reports of the capability of DWI to detect pulmonary nodules, which are most commonly included in whole-body acquisitions for tumor staging or detection. In most of the series comparing the detection of pulmonary metastasis in either positron emission tomography (PET) or integrated PET-CT cameras with whole-body (WB) DWI MR imaging, the accuracy and sensitivity of both methods are similar, with similar rates of false-positive lesions.
2-[Fluorine-18]-fluoro-2-deoxy-D-glucose positron emissiontomography/computed tomography versus whole-body diffusion-weighted MRI for detection of malignant lesions: initial experience.
Detection of metastatic lesions from malignant pheochromocytoma and paraganglioma with diffusion-weighted magnetic resonance imaging: comparison with 18F-FDG positron emission tomography and 123I-MIBG scintigraphy.
WB-DWI MR imaging missed three of five pulmonary metastasis of non–small cell lung cancer (NSCLC), with a size of less than 10 mm. Koyama and colleagues
Comparison of STIR turbo SE imaging and diffusion-weighted imaging of the lung: capability for detection and subtype classification of pulmonary adenocarcinomas.
recently showed that the detection rate of pulmonary adenocarcinomas was significantly lower for a dedicated chest DWI sequence than that of STIR. In another series, small metastases and nonsolid adenocarcinomas showed low signal intensity on DWI sequence with high b value (1000 s/mm2), which makes them very difficult to be detected (Fig. 5).
where seven metastasis and two moderately differentiated adenocarcinomas demonstrated only moderate hyperintensity on DWI with a b value of 500 s/mm2.
Fig. 5Pulmonary metastases of papillary thyroid carcinoma. (A) Axial black-blood STIR image shows several bilateral millimetric pulmonary metastases and enlarged prevascular and bilateral hilar lymphadenopathies, probably representing lymph node metastases. Respiratory-triggered SS EPI DWI sequence with SPIR on a 1.5-T magnet with b values of 150 s/mm2 (B) and 500 s/mm2 (C) demonstrate a lesser number of pulmonary metastases than STIR. The enlarged lymph nodes are also less evident.
In most of the reports evaluating pulmonary nodules with a dedicated chest DWI sequence, the size of the studied nodules is usually superior to 1 cm, which supposes a limitation to find out the real potential of DWI in the detection of lung lesions, making further research with smaller nodules necessary. Therefore, with the available data, DWI may be inferior to STIR in the detection of pulmonary nodules, and similar to PET and PET-CT. Larger series are necessary to define the behavior of well-differentiated adenocarcinomas and lung metastasis.
Pulmonary Nodules and Lung Cancer Characterization
The characterization of pulmonary nodules is still a common clinical dilemma. The probability of malignancy increases with the nodule’s size. Only 20% of nodules larger than 20 mm are benign. For instance, the prevalence of malignancy for nodules larger than 20 mm ranges from 64% to 82%.
To avoid unnecessary surgical resection of benign nodules, it is important to get as precise as possible noninvasive characterization. CT is the most used test in pulmonary nodule evaluation, but still is based on morphologic criteria, showing obvious limitations, mainly in areas of altered pulmonary anatomy. Dynamic enhanced CT shows an excellent sensitivity but a limited specificity, because there is some overlap in enhancing nodules between active granulomas, hypervascular benign nodules, and malignant nodules.
PET has demonstrated usefulness in this task, but it is also limited in the detection of adenocarcinomas and shows an important false-positive rate caused by inflammation.
Conventional MR imaging has been proposed for the evaluation of solitary pulmonary nodules according to their relaxation times with significant overlap between benign and malignant tumors.
More recently, dynamic enhanced MR imaging has shown better specificity and accuracy than multidetector CT and coregistered PET-CT in the differentiation between benign and malignant nodules,
Dynamic MRI, dynamic multidetector-row computed tomography (MDCT), and coregistered 2-[fluorine-18]-fluoro-2-deoxy-D-glucose-positron emission tomography (FDG-PET)/CT: comparative study of capability for management of pulmonary nodules.
Based on the concept that malignant lesions demonstrate increased cellularity, higher tissue disorganization, and increased extracellular space tortuosity compared with benign lesions, diffusion of interstitial water should be restricted in cases of lung cancer (Fig. 6). Several series have been published in the last 5 years exploring the capabilities of DWI in the characterization of pulmonary lesions (Table 2).
Fig. 6Poorly differentiated adenocarcinoma. (A) Free breathe SS EPI DWI sequence with SPIR and a b value of 800 s/mm2 performed in a 1-T magnet shows restricted diffusion of the lesion, which is confirmed in the ADC map (B). The mass demonstrated an ADC value of 1.2 × 10−3 mm2/s. Notice the presence in A of a metastatic right hilar lymphadenopathy, which is detectable on the DWI sequence (arrow), which was not evident on T2 weighted (not shown).
Comparison of STIR turbo SE imaging and diffusion-weighted imaging of the lung: capability for detection and subtype classification of pulmonary adenocarcinomas.
evaluated 54 nodules larger than 5 mm. They could accurately differentiate benign from malignant nodules, with an area under the curve of 0.80. Small metastasis and some nonsolid adenocarcinomas were predominantly hypointense on DWI with high b value. Granulomas, active inflammatory, and fibrous nodules occasionally showed high signal intensity in a similar fashion to malignant lesions (Fig. 7).
Fig. 7Pulmonary abscess and exudative pleural effusion. (A, B) Free breathe SS EPI DWI sequence with SPIR on a 1-T magnet with b values of 0 s/mm2 (A) and 800 s/mm2 (B) show a hyperintense mass located in the left lung along with an exudative pleural effusion. (C) The ADC map confirms the restriction of the mass, because it shows a minimal ADC value of 1 × 10−3 mm2/s.
analyzed 66 pulmonary lesions and did not find significant differences in the signal intensity between benign and malignant nodules using a DWI sequence with a maximum b value of 500 s/mm2. However, in this report, a threshold ADC value of 1.4 × 10−3 mm2/s allowed the distinction between benign and malignant lesions with a sensitivity of 83% and a specificity of 74%. Furthermore, small cell lung cancer (SCLC) demonstrated statistically significant lower ADC values than NSCLC.
Different results were obtained in a series by Uto and colleagues.
They evaluated 28 pulmonary lesions larger than 1 cm with a DWI sequence with a maximum b value of 1000 s/mm2. They compared a semiquantitative approach, measuring the signal intensity of the lesions and the spinal cord, with the ADC values to differentiate between benign and malignant nodules. In the receiver operating characteristic curve analysis, the semiquantitative approach had a higher area under the curve compared with ADC values (0.911 vs 0.600).
achieved better results in the distinction of benign from malignant pulmonary nodules with DWI with a b value of 1000 s/mm2 compared with PET. Using a cut-off ADC value of 1.1 × 10−3 mm2/s, they obtained a sensitivity of 70% and specificity of 97%, compared with the 72% and 79% obtained by PET, respectively. DWI also reduced the rate of false-positive lesions compared with PET.
DWI has been defined as an in vivo biomarker of tumoral grade and differentiation in oncologic lesions in other organs, because more aggressive lesions are more hypercellular than well-differentiated lesions.
performed a split acquisition of fast SE signals for diffusion imaging with two b values (68.46 and 577.05 s/mm2) in 30 patients with lung carcinoma, demonstrating an adequate correlation between the ADC values of lung cancer and tumor cellularity. Although there was an overlap between the ADC values of the different types of lung carcinoma, they demonstrated that well-differentiated adenocarcinomas showed higher ADC values than those of more aggressive adenocarcinomas and epidermoid carcinomas in a significant manner, because well-differentiated adenocarcinomas showed lesser tumor cellularity and cellular differentiation than the other types of pulmonary carcinoma.
studied 46 peripheral adenocarcinomas lesser than 3 cm, with DWI with a higher b value of 1000 s/mm2. Adenocarcinomas were histologically classified as BAC, advanced-BAC, mixed subtype, and non-BAC. The first one shows a favorable prognosis and the last two subtypes are invasive cancers. They did not use ADC quantification, if not a visual assessment of the signal intensity on DWI of the nodules compared with spinal cord. They were able to significantly differentiate invasive adenocarcinomas from BAC, because invasive adenocarcinomas usually showed higher signal intensity (see Fig. 6, Fig. 8).
Fig. 8Bronchioalveolar carcinoma. (A) Fluorodeoxyglucose PET shows ill-defined uptake of a peripheral lesion in right upper lobe (arrow), which was not considered suspicious for malignancy. (B) IVIM-DWI sequence at 3-T magnet with multiple b values (only shown 9) depicts the nodule as moderately hyperintense with high b values. (C) Parametric map of D confirms the lesional restricted diffusion. (D) Comparison of signal decay within the lesion using either the IVIM model estimation (black line) or the conventional ADC estimation from the monoexponential model (gray line). The ADC value is that of 1.85 × 10−3 mm2/s, in the range of a benign lesion. However, if one applies the bicompartimental model of DWI, one can calculate the lesional value of D, 1.23 × 10−3 mm2/s, in the range of a malignant lesion. In this case, the effect of the perfusion contribution to the ADC estimation may cause a false-positive, as was the PET.
Comparison of STIR turbo SE imaging and diffusion-weighted imaging of the lung: capability for detection and subtype classification of pulmonary adenocarcinomas.
evaluated 33 adenocarcinomas with DWI and a higher b value of 1000 s/mm2. The ADC values were not useful to differentiate the subtypes of adenocarcinoma.
studied 41 patients with clinical stage IA NSCLC who had undergone curative resection with DWI and PET-CT. Using a visual qualitative analysis, DWI was found to be an independent predictive factor to detect patients with invasive cancer with a sensitivity of 90%, a specificity of 81%, a positive predictive value of 60%, and a negative predictive value of 96%. In another report, DWI was equivalent to PET in distinguishing NSCLC from benign pulmonary nodules. PET was able to predict tumoral aggressiveness, showing significant differences between pathologic stages IA versus IB or more advanced stages and between well-differentiated and moderately or poorly differentiated adenocarcinomas. However, DWI using ADC quantification was of no value in the prediction of tumoral invasion.
Is diffusion-weighted magnetic resonance imaging superior to positron emission tomography with fludeoxyglucose F 18 in imaging non-small cell lung cancer?.
It is difficult to compare results from the previously referred series because of the different DWI sequences performed, and different qualitative and quantitative assessment methods. As a general rule, DWI of pulmonary nodules achieves good results in the differentiation between benign and malignant nodules (Fig. 9). Limitations of the technique are the presence of a significant number of false-positives related mainly to benign inflammatory lesions and potential false-negatives of low-grade adenocarcinomas and metastasis. In the limited series available comparing DWI with PET, both perform equivalently in pulmonary lesion characterization, with similar limitations, although DWI tends to have fewer false-positives.
Fig. 9Benign pulmonary nodule. Respiratory-triggered SS EPI DWI sequence with SPIR on a 1.5-T magnet with b values of 0 (A), 150 (B), and 600 s/mm2 (C) demonstrates a spiculated nodule in right lung (arrows) with progressive loss of signal while increasing the diffusion-weighting. An ADC value of 1.9 × 10−3 mm2/s suggests a benign lesions, as confirmed clinically in 6 years of imaging follow-up.
the authors’ experience indicates that with current state-of-the-art magnets, a b value of 1000 s/mm2 can be performed perfectly without significant image quality loss or increase of susceptibility artifacts, allowing a better differentiation between benign and malignant lesions. With regard to the most appropriate assessment of pulmonary DWI, Uto and colleagues
correctly stated that ADC calculations of pulmonary lesions were significantly affected by perfusion phenomena. They proposed to increase the higher b value over 1000 s/mm2 to prevent perfusion effects, because DWI sequences obtained with higher b values are more sensitive to diffusion. Additionally, ADC measurements obtained with higher b values are generally smaller than those obtained using lower b values. In addition, the IVIM model of diffusion signal decay has demonstrated that microvascular perfusion is detected at low b values (under 100 s/mm2), allowing one to calculate the perfusion-free diffusion parameter (D) in several organs, such as brain, abdominal organs, or muscle.
Differentiation of pancreas carcinoma from healthy pancreatic tissue using multiple b-values: comparison of apparent diffusion coefficient and intravoxel incoherent motion derived parameters.
is to avoid b values under 100 in the ADC quantification, to partially avoid perfusion effects. In the authors’ experience, the IVIM model is feasible in the thorax, allowing differentiation of the diffusion and perfusion in pulmonary nodules, although it remains to be proved whether this approach improves lesional detection and characterization (see Figs. 3 and 8).
In centrally located lung cancers, DWI has been shown to be able to accurately differentiate postobstructive consolidation from central lung carcinoma, which is important in the planification of radiotheraphy.
In another series, the differentiation of central lung cancer from postobstructive lobar collapse was superior with DWI compared with either T2-weighted sequences or enhanced CT (Fig. 10).
Fig. 10Central bronchogenic carcinoma with postobstructive consolidation. Coronal fusion image of a T2 TSE image and a SS EPI DWI sequence with a b value of 1000 s/mm2 allows a good depiction of the epidermoid carcinoma as an area of restricted diffusion (asterisk) surrounded by postobstructive consolidation, which does not demonstrate hyperintensity on DWI.
Other potential applications of DWI in lung cancer, still to be fully explored, are monitoring treatment response after chemotherapy or radiation, distinguishing posttherapeutic changes from residual active tumor, and the detection of recurrent cancer (Fig. 11). DWI has also been used in other organs to predict response to treatment of cancer before and soon after therapy, which has still to be investigated for lung cancer.
evaluated prospectively 17 patients with 20 malignant lung lesions that underwent CT-guided radiofrequency ablation. DWI with ADC calculation was performed immediately before and 3 days after treatment. The posttreatment ADC of the lesions without local progression was significantly higher than that of the lesions with local progression. However, this difference could not be demonstrated for the pretreatment ADC quantification.
Fig. 11Recurrent poorly differentiated adenocarcinoma. A 64 year-old man with antecedent of NSCLC, treated 2 years before, and in clinical complete response. (A) Axial postcontrast THRIVE shows a spiculated lesion with heterogeneous enhancement. (B) IVIM-DWI sequence with a b value of 900 mm2/s depicts the nodule with focal areas of hyperintensity (arrows). (C) Parametric map of D confirms the restricted diffusion. Lesional D value, at the place where the ROI is positioned, was that of 1.5 × 10−3 mm2/s, consistent with recurrent lesion.
Currently, PET and PET-CT are considered the most accurate noninvasive techniques in the N-staging of NSCLC, although they lack specificity because of concurrent inflammatory lymphadenitis.
Although STIR turbo-SE sequence has shown is ability to distinguish between benign and malignant mediastinal lymph nodes, its use is still limited in daily clinical practice.
Metastases in mediastinal and hilar lymph nodes in patients with non-small cell lung cancer: quantitative and qualitative assessment with STIR turbo spin-echo MR imaging.
Diffusion-weighted magnetic resonance imaging can be used in place of positron emission tomography for N staging of non-small cell lung cancer with fewer false-positive results.
demonstrated that DWI was significantly more accurate than PET in the N-staging of NSLC, because of less overstaging and fewer false-positives in the former (Fig. 12). The detectable size of node metastases for both methods was 4 mm. They used a maximum b value of 1000 s/mm2 and a threshold ADC value of 1.6 × 10−3 mm2/s. In this series, inflammatory lymphadenitis usually showed increased fluorodeoxyglucose uptake but not restricted diffusion, which justifies the difference in false-positive results between both techniques (Fig. 13). Most of the false-positive results on DWI were caused by lymph nodes with granulation tissue of tuberculosis or nontuberculosis origin, which are also a cause of false-positives on PET-CT.
Fig. 12Metastatic adenopathies of poorly differentiated lung adenocarcinoma. (A, B) Free breathe SS EPI DWI sequence with SPIR on a 1-T magnet with a b value of 800 s/mm2 at two different levels demonstrates a mass with restricted diffusion in the superior segment of the right inferior lobe corresponding to a pulmonary adenocarcinoma (red arrow) (A) and metastatic right hilar (white arrow on A) and right paratracheal adenopathies (white arrow on B).
Fig. 13Lymphadenitis in a patient with bronchioalveolar carcinoma (same case as Fig. 8). Respiratory-triggered SS EPI DWI sequence with SPIR on a 3-T magnet with b values of 300 s/mm2 (A) and 900 s/mm2 (B) demonstrates absence of restricted diffusion of a small right hilar lymph node (arrows on both images), which was confirmed as benign lymphadenitis on pathologic analysis. Notice the presence of fat signal overlap in both images, although consistent fat suppression was reached.
Metastases in mediastinal and hilar lymph nodes in patients with non-small cell lung cancer: quantitative assessment with diffusion-weighted magnetic resonance imaging and apparent diffusion coefficient.
In all of these series, the presence of susceptibility and chemical shift artifacts was not a limitation to obtaining acceptable ADC maps. A potential problem when using DWI for mediastinal imaging may be to correctly locate the lymph nodes, because of its intrinsic low spatial resolution of DWI sequences. The use of fusion software allows the overlay of anatomic and DWI sequences, partially solving this problem.
The detectable size of metastatic thoracic lymph node with the current available technology is around 4 to 5 mm for both DWI and PET-CT. Therefore, lymph node dissection may not be reduced for patients with N0 stage diagnosed by DWI or PET-CT, because node metastases inferior to this size are not uncommon. Table 3 summarizes the technical characteristics of the DWI sequences used in the evaluation of mediastinal lymph nodes.
Table 3Resume of technical parameters of DWI sequences used in the published series evaluating mediastinal lymph nodes
Diffusion-weighted magnetic resonance imaging can be used in place of positron emission tomography for N staging of non-small cell lung cancer with fewer false-positive results.
Metastases in mediastinal and hilar lymph nodes in patients with non-small cell lung cancer: quantitative assessment with diffusion-weighted magnetic resonance imaging and apparent diffusion coefficient.
In contrast to these results, in series using WB-DWI MR imaging approach with a DWIBS sequence, the accuracy of the N-staging of lung cancer has not been as favorable as PET-CT. Lichy and colleagues
evaluated the performance of WB-DWI MR imaging for tumor detection compared with PET-CT. They included in their series three patients with lung cancer and 11 metastatic thoracic lymph nodes detected on PET-CT, of which WB-DWI MR imaging could only detect one. More recently, Chen and colleagues
analyzed in 56 patients the performance of WB-DWI MR imaging and PET-CT in the N- and M-staging of NSCLC. They obtained significant differences in the accuracy of lymph node metastases detection, favoring PET-CT. In the evaluation of hilar and mediastinal node metastases, they had two false-negative cases not detected by WB-DWI MR imaging that corresponded to lymph nodes with a size less of 10 mm, and only one with PET-CT. In the same series, WB-DWI MR imaging missed six node metastases in the neck and presented four false-positive cases in the same region. In contrast, PET-CT had only one false-negative and another false-positive in the same area. The evaluation of the neck region was problematic because of parallel girdle-like artifacts, which frequently obscured the metastatic lymph nodes.
similar results were found in the M-staging of NSCLC for WB-DWI MR imaging and PET-CT, although better detection rates were achieved with the last technique. Ohno and colleagues
stated that WB-DWI MR imaging should be used alone with morphological whole-body sequences to improve the diagnostic accuracy of this technique, because WB-DWI MR imaging, including only a DWIBS sequence, showed a significantly worse specificity and accuracy for M-stage assessment of NSCLC with the inclusion of brain metastases than either morphological WB imaging, with or without DWIBS sequence or PET-CT (Fig. 14). Most of the false-positives and false-negatives with both techniques corresponded to brain and pulmonary lesions. Chen and colleagues
did not used conventional MR imaging sequences in their series, which may justify the differences in results compared with the report by Ohno and colleagues.
Neither performed ADC quantifications, which may also leave room for further improvements in the accuracy of WB-DWI MR imaging.
Fig. 14Staging of lung cancer with WB-DWI MR imaging. Coronal fusion image of a T2-weighted TSE sequence and a DWIBS acquisition (b value of 1000 s/mm2) reveals a huge mass in the inferior lobe of the right lung corresponding to a SCLC (asterisk). Metastasis in L2 vertebral body (green arrow) and right supraclavicular lymph node metastasis (yellow arrow) are depicted as areas of restricted diffusion.
demonstrated that DWI and ADC measurements allowed one to distinguish between benign and metastatic lymph nodes of SCLC and NSCLC. In a similar manner, Sakurada and colleagues
demonstrated that the ADC values of node metastases of esophageal cancer unexpectedly were significantly higher than that of nonmetastatic lymph nodes, although there was an overlap in the ADCs of both groups. The higher ADC values of node metastases may be related to areas of microscopic necrosis. Average patient-based sensitivity and specificity for the detection of node metastasis was 77.8% and 55.6%, respectively. They also investigated the role of DWIBS in the detection of thoracic esophageal cancer, with a poor detection rate of 49.4%, the depiction of early tumors being especially problematic (Fig. 15).
Fig. 15Esophageal cancer. (A) Respiratory-triggered SS EPI DWI sequence with SPIR on a 1.5-T magnet with a b value of 1000 s/mm2 demonstrates a nodular area of restricted diffusion (arrow) corresponding to an esophageal cancer with T2 stage, which was not detectable on the axial T2 TSE image at the same level (B). The mural thickening of distal esophagus (arrows) is confirmed in an oblique sagittal dynamic balanced field echo acquisition after water swallowing (C).
have recently explored the capabilities of DWI to further characterize mediastinal masses using free-breathe SS EPI MR imaging with b factors of 0, 300, and 600 s/mm2 and ADC quantification. They evaluated 45 patients with mediastinal tumors, excluding the purely cystic ones. Using a cut-off ADC value of 1.56 × 10−3 mm2/s, they obtained an accuracy of 95%, sensitivity of 96%, specificity of 94%, positive predictive value of 94%, negative predictive value of 96%, and area under the curve of 0.938 in the differentiation between benign and malignant tumors. Moreover, there was a significant difference in the ADC value between poorly and well-differentiated malignant tumors of the mediastinum. In this series, the lowest ADC value was that of lymphoma cases, although there was an overlap with ADC of other malignancies, such as thymoma and bronchogenic carcinoma. These results were similar to previous reports using a DWIBS sequence for WB imaging.
WB-DWI MR imaging has also been demonstrated to be a feasible technique in the initial staging of lymphoma, including mediastinal involvement, with results as accurate as CT or PET-CT.
Whole-body diffusion-weighted magnetic resonance imaging with apparent diffusion coefficient mapping for staging patients with diffuse large B-cell lymphoma.
being especially interesting in pediatric and pregnant patients (Fig. 16).
Fig. 16Staging of Hodgkin's lymphoma in a 24-weeks pregnant woman. (A) Coronal TSE T2-weighted image and (B) coronal maximum intensity projection (MIP) of a DWIBS sequence with a b value of 1000 s/mm2 show disease limited to mediastinum and left laterocervical lymph nodes (arrows).
could accurately differentiate between exudative and transudative pleural effusion using a DWIBS sequence with the body coil and a maximum b value of 1000 s/mm2. They proposed a cutoff ADC value of 3.38 × 10−3 mm2/s to obtain a sensitivity of 90.6% and specificity of 85% (see Fig. 7, Fig. 17). In a more recent report by Inan,
the ADCs of the exudative lesions were also significantly lower than those of transudative ones. In this series, the DWI sequence was performed with a four-element phased-array coil using spectral fat saturation with inversion recovery technique and obtaining b values of 0, 500, and 1000 s/mm2. The signal intensity of transudative effusions tend to be isointense and exudative effusions hyperintense compared with muscle.
Fig. 17Transudative pleural effusion in a patient with chronic renal failure. Respiratory-triggered SS EPI DWI sequence with SPIR on a 1.5-T magnet with b values of 0 s/mm2 (A) and 800 s/mm2 (B) demonstrates a bilateral pleural effusion, which does not show restriction of diffusion.
None of the three distinct histologic subtypes of MPM (epithelial, sarcomatoid, and biphasic) can be distinguished from each other by current imaging modalities.
This differentiation is important because there is a significant difference in prognosis between epithelioid and nonepithelioid (biphasic and sarcomatoid) MPM.
DWI has recently demonstrated the ability to differentiate epithelial and sarcomatoid subtypes of MPM in a group of 57 patients, using a 3-T magnet. The authors used a free-breathing SS SE EPI DWI sequence with spectral fat saturation and the generalized autocalibrating partially parallel acquisition (GRAPPA) technique was the parallel imaging technique. They obtained 3 b values (250, 500, and 750 s/mm2) to posterior calculate ADC maps. The sarcomatoid subtype showed significantly lower ADC values than the epithelial subtype. The ADC values of biphasic MPM had a wide range of overlap with the ADC values of other subtypes. In the same series, two cases of benign pleural plaque were included, demonstrating a lower ADC value than any type of MPM.
Future developments
Cardiac DWI and DTI
Another possibility of DWI is to acquire signal diffusion information from the heart. In DWI of the myocardium, it is a difficult task to completely avoid the macroscopic movement signal, originating in the heartbeat and respiratory motion, from the microscopic movement information provided by DWI and DTI. A deep explanation of the diffusion sequences and DTI reconstruction methods for heart applications is beyond the scope of this article, but an excellent review can be found in Sosnovik and colleagues.
None of those sequences are normally available in commercial scanners. Therefore, to get DWI information of the heart, it is necessary to tune the conventional SE Stejskal–Tanner sequence to the ECG signal by means of synchronization. Conventional DWI sequences for cardiac applications are most commonly synchronized with the systolic part of the heart cycle to improve the reproducibility and to benefit from the increase in thickness of the myocardium in this phase. If a conventional SE Stejskal–Tanner approach is performed, the bulk motion of the heart during systole completely destroys the diffusion signal. For this reason, when using the conventional SE approach, it is desirable to synchronize the DWI acquisition with the diastole reducing the effect of the left ventricle movement. Besides, it is also preferable to use the maximum gradient strength in the diffusion-weighted part of the sequence, to make this step as short as possible, enabling one to consider the heart almost completely quiet during this preparation phase. The maximum applied b value is normally around 300 s/mm2, making it feasible to acquire images with TE of about 40 milliseconds. Finally, to remove breathing artifacts it is recommended to acquire the images with breathholding or respiratory triggered.
In the short existent experience with cardiac DWI, it has been demonstrated to be a feasible method to detect myocardial edema in patients with recent myocardial infarction (MI), because areas of increased signal caused by restricted diffusion, show ADC values in the edematous area lower than in normal myocardium.
DWI differentiated necrotic from viable myocardium, whereas DWI was normal in chronic infarcts. In our experience, DWI is also able to detect myocardial edema in cases of myocarditis (Fig. 18), as Laissy and colleagues anticipated in their paper, and may help to further characterize cardiac and paracardiac masses (Fig. 19).
Fig. 18Detection of myocardial edema in acute myocarditis. (A) Short-axis double-inversion black-blood STIR demonstrates an area of myocardial edema in the anteroseptal wall (arrow). Respiratory- and cardiac-triggered SS EPI DWI sequences with SPIR on a 1.5-T magnet with b values of 150 s/mm2 (B) and 300 s/mm2 (C) also clearly depict the edematous myocardium as an area of focal hyperintensity (arrows). (D) On the corresponding ADC map the area of edema demonstrated restricted diffusion (arrow).
Fig. 19Bronchogenic cyst. (A) Sagittal TSE T2-weighted sequence shows a hyperintense well-defined mass in the posterior mediastinum, which is cystic-appearing (asterisk). Respiratory-triggered SS EPI DWI sequence with SPIR on a 1.5-T magnet with b values of 0 s/mm2 (B) and 800 s/mm2 (C) demonstrate absence of restricted diffusion within the mass (asterisks), which helps to characterize it as cystic.
There is a growing interest in the literature to study the microstructure of the heart to separate the different layers of the myocardium using DTI experiments. To acquire this information, it is necessary to perform DWI in at least six different diffusion directions that permit the building of tensor information. In Sosnovik and colleagues
a good overview of the reconstruction algorithms can be found. DTI has allowed a better understanding of the three-dimensional organization of myocardial fibers, which is determinant of cardiac torsion, strain, and stress (Fig. 20). Nowadays, cardiac DTI is feasible in vivo for animals and humans, although it is still far from being ready for the clinical arena. In vivo DTI of MI has recently revealed a significant increase in trace ADC (mean diffusivity) and a decrease in fractional anisotropy, related to altered myocardial structure, in patients with previous MI at a median interval of 26 days.
Diffusion tensor magnetic resonance imaging mapping the fiber architecture remodeling in human myocardium after infarction: correlation with viability and wall motion.
In the same study, the alteration of tissue integrity and fiber architecture measured by DTI demonstrated a significant correlation with viable myocardium and regional wall function. Posteriorly, the same researches have shown that DTI may monitor the sequential changes that occur in the transition from recent to chronic MI, with an association between sequential zonal improvement of tissue integrity and fiber architecture remodeling with sequential recovery of zonal wall thickening of the infarcted area.
Sequential changes of myocardial microstructure in patients postmyocardial infarction by diffusion-tensor cardiac MR: correlation with left ventricular structure and function.
The role of cardiac DTI and DWI in the detection of hyperacute MI and in the monitorization of postinfarction remodeling needs further research.
Fig. 20Ex vivo DTI of a pig heart, using a conventional SE DWI sequence with a b value of 800 s/mm2. On the left row, short-axis source and fractional anisotropy images of the heart are presented. On the right, a three-dimensional DTI reconstruction reveals the organization and pathways of the heart fibers.
Functional MR imaging of the lungs with hyperpolarized gases has revolutionized the potential for in vivo lung function measurement in health and disease.
Gases, such as 129Xe and more frequently 3He, may be hyperpolarized to be administered into the lungs and imaged with MR imaging, before they return to the thermal equilibrium conditions dictated by the body temperature and the local magnetic field. With the hyperpolarization process of these gases, a very important increase in MR imaging signal is achieved that allows the overcoming of low concentration of gas molecules in airways and may be used for higher spatial or temporal resolution imaging.
Their use is still limited to research centers, because the access to hyperpolarization methods and hardware MR imaging requirements limit their spread. The research on imaging of ventilation with 3He is more extended than that with 129Xe, because of higher levels of polarization and higher gyromagnetic ratio, and because it is easily available.
In contrast, 129Xe permits a more comprehensive assessment of lung function, because measurement of perfusion and gas exchange is possible because of its solubility in blood and tissues, which are not properties of 3He.
Besides the quantitative MR imaging-derived ventilation, DWI of inhaled hyperpolarized gases has been shown to give information of microstructural pulmonary changes in various diseases, such as asthma or chronic obstructive pulmonary disease (COPD), through measurements of their ADC (Fig. 21).
Assessment of the lung m in patients with asthma using hyperpolarized 3He diffusion MRI at two time scales: comparison with healthy subjects and patients with COPD.
The MR imaging sequences to monitoring the diffusion information of lung are normally based in a gradient echo or fast gradient echo acquisition where two bipolar gradients, with some separation between them, are included to sensitize DWI. To reduce the T2* decay as much as possible, centric radial or spiral acquisitions are needed.
Fig. 21Hyperpolarized 3He pulmonary diffusion in a rat model of induced emphysema. (A) ADC map of a rat lung with emphysema-like disease induced using elastase in left lobe (right-side image in the figure). The emphysema can be considered as mild. (B) Graphic of mean ADC value of the entire pulmonary region comparing elastase-treated (left bars) with normal (right bars) rats, demonstrating higher ADC values in lungs with induced emphysema. The ADC was obtained using four b values between 0 and 2.4 s/cm2, and a bipolar sinusoidal diffusion gradient of 1.5-ms diffusion time. The data were accumulated at the end of the expiratory volume after three prewashes with pure 3He and 15 mbar inspiration.
(Courtesy of Angelos Kyriazis and Jesus Ruiz-Cabello, Research Center of Respiratory Diseases, Complutense University, Madrid, Spain.)
Studies using 3He have demonstrated increases in ADC in animals and patients with both emphysema and COPD, resulting from the enlargement of the airspaces. Furthermore, increases in ADC have been described in healthy smokers, as an early marker of alveolar breakdown, and also regional differences in ADC values within the same subject indicate different grades of lung destruction.
Assessment of the lung m in patients with asthma using hyperpolarized 3He diffusion MRI at two time scales: comparison with healthy subjects and patients with COPD.
Similar results may be theoretically obtained with 129Xe (Fig. 22), although larger studies are needed to validate and extend these data.
Fig. 22Hyperpolarized 129Xe pulmonary ventilation and diffusion on a healthy volunteer (top row) and subject with COPD (bottom row). Multiple ventilation defects and an increase in ADC values are detected in the COPD patient compared with the volunteer. The diffusion alterations are mainly located in both upper lobes.
(Courtesy of S. Kaushik and B. Driehuys, Center for In Vivo Microscopy, Department of Radiology, Duke University Medical Center, Durham, NC.)
DWI of the chest is technically feasible with state-of-the-art magnets, although it is technically complex. It has a clinical role in lung nodule characterization, lung cancer staging, and the evaluation of pleural and mediastinal pathology, although larger series are needed to validate the existing preliminary data. The evaluation of the IVIM approach in chest pathology may also improve the existing results. Promising applications to be developed are DWI and DTI of the heart and diffusion of hyperpolarized gases.
Comparison and reproducibility of ADC measurements in breathhold, respiratory triggered, and free-breathing diffusion-weighted MR imaging of the liver.
Respiratory-triggered versus breath-hold diffusion-weighted MRI of liver lesions: comparison of image quality and apparent diffusion coefficient values.
MRI of the lung: value of different turbo spin-echo, single-shot turbo spin-echo, and 3D gradient-echo pulse sequences for the detection of pulmonary metastases.
2-[Fluorine-18]-fluoro-2-deoxy-D-glucose positron emissiontomography/computed tomography versus whole-body diffusion-weighted MRI for detection of malignant lesions: initial experience.
Detection of metastatic lesions from malignant pheochromocytoma and paraganglioma with diffusion-weighted magnetic resonance imaging: comparison with 18F-FDG positron emission tomography and 123I-MIBG scintigraphy.
Comparison of STIR turbo SE imaging and diffusion-weighted imaging of the lung: capability for detection and subtype classification of pulmonary adenocarcinomas.
Dynamic MRI, dynamic multidetector-row computed tomography (MDCT), and coregistered 2-[fluorine-18]-fluoro-2-deoxy-D-glucose-positron emission tomography (FDG-PET)/CT: comparative study of capability for management of pulmonary nodules.
Is diffusion-weighted magnetic resonance imaging superior to positron emission tomography with fludeoxyglucose F 18 in imaging non-small cell lung cancer?.
Differentiation of pancreas carcinoma from healthy pancreatic tissue using multiple b-values: comparison of apparent diffusion coefficient and intravoxel incoherent motion derived parameters.
Metastases in mediastinal and hilar lymph nodes in patients with non-small cell lung cancer: quantitative and qualitative assessment with STIR turbo spin-echo MR imaging.
Diffusion-weighted magnetic resonance imaging can be used in place of positron emission tomography for N staging of non-small cell lung cancer with fewer false-positive results.
Metastases in mediastinal and hilar lymph nodes in patients with non-small cell lung cancer: quantitative assessment with diffusion-weighted magnetic resonance imaging and apparent diffusion coefficient.
Whole-body diffusion-weighted magnetic resonance imaging with apparent diffusion coefficient mapping for staging patients with diffuse large B-cell lymphoma.
Diffusion tensor magnetic resonance imaging mapping the fiber architecture remodeling in human myocardium after infarction: correlation with viability and wall motion.
Sequential changes of myocardial microstructure in patients postmyocardial infarction by diffusion-tensor cardiac MR: correlation with left ventricular structure and function.
Assessment of the lung m in patients with asthma using hyperpolarized 3He diffusion MRI at two time scales: comparison with healthy subjects and patients with COPD.
00069-3&id=gr1.jpg){kind=link}
00069-3&id=gr2.jpg){kind=link}
00069-3&id=gr3.jpg){kind=link}
00069-3&id=gr4.jpg){kind=link}
00069-3&id=gr5.jpg){kind=link}
00069-3&id=gr6.jpg){kind=link}
00069-3&id=gr7.jpg){kind=link}
00069-3&id=gr8.jpg){kind=link}
00069-3&id=gr9.jpg){kind=link}
00069-3&id=gr10.jpg){kind=link}
00069-3&id=gr11.jpg){kind=link}
00069-3&id=gr12.jpg){kind=link}
00069-3&id=gr13.jpg){kind=link}
00069-3&id=gr14.jpg){kind=link}
00069-3&id=gr15.jpg){kind=link}
00069-3&id=gr16.jpg){kind=link}
00069-3&id=gr17.jpg){kind=link}
00069-3&id=gr18.jpg){kind=link}
00069-3&id=gr19.jpg){kind=link}
00069-3&id=gr20.jpg){kind=link}
00069-3&id=gr21.jpg){kind=link}
00069-3&id=gr22.jpg){kind=link}