Non‐Contrast‐Enhanced Functional Lung MRI to Evaluate Treatment Response of Allergic Bronchopulmonary Aspergillosis in Patients With Cystic Fibrosis: A Pilot Study

Allergic bronchopulmonary aspergillosis (ABPA) in cystic fibrosis (CF) patients is associated with severe lung damage and requires specific therapeutic management. Repeated imaging is recommended to both diagnose and follow‐up response to treatment of ABPA in CF. However, high risk of cumulative radiation exposure requires evaluation of free‐radiation techniques in the follow‐up of CF patients with ABPA.

C ystic fibrosis (CF) is one of the most frequent genetic disorders in the Caucasian population. 1 It is caused by mutations in the cystic fibrosis transmembrane conductance regulator (CFTR) gene, resulting in diverse pathologic manifestations including bronchiectasis, nasal polyposis, pancreatic insufficiency, and infertility. 2ung involvement (i.e., destruction of lung parenchyma and decline in pulmonary function) is responsible for most life-limiting complications. 1 Mucus composition and thickness in CF patients provide a favorable environment for fungal respiratory infections. 3Indeed, Aspergillus fumigatus has been found in the sputum of up to 57% of patients with CF. 4 The growth of A. fumigatus hyphae within the bronchial lumen may be responsible for an immunoglobulin E (IgE)-mediated hypersensitivity response, known as allergic bronchopulmonary aspergillosis (ABPA), leading to bronchial inflammation and airway destruction. 5Its prevalence in CF patients ranges between 3% and 25%, and is higher than that in asthmatic patients. 6In CF patients, ABPA, which is associated with severe lung damage and requires specific therapeutic management, may be suspected due to worsening of pulmonary function and evidence of new infiltrates on chest radiographs or computed tomography (CT). 5,68][9] Another finding of ABPA is highattenuation mucus (HAM) plugging on CT images, which has been considered highly specific, if not pathognomonic of the disease. 10One of the important aspects in the management of ABPA in CF patients is repeating imaging after treatment to evaluate the effectiveness of the chosen therapeutics. 11,12However, repeating CT examinations should be limited due to the risk associated with high cumulative radiation exposure. 13,14Recently, lung MRI has been evaluated to assess structural alterations in CF. 15,16 Indeed, lung morphological modifications in CF assessed using ultrashort echo time (UTE)-MRI have been found to be similar to those using CT. 15,17Furthermore, inverted mucus impaction signal on MRI (IMIS), which is characterized by a high-signal intensity on T1-weighted images and low-signal intensity on T2-weighted images, can be considered as the counterpart of HAM and has been found to be 100% specific to detect ABPA in CF patients. 18Moreover, lung MRI can evaluate pulmonary functional changes in a regional fashion. 19on-contrast techniques of pulmonary ventilation-perfusion assessment in CF patients (eg, Fourier decomposition [FD] MRI) have been reported. 20,21These have been validated against photon emission computed tomography (SPECT), dynamic contrast-enhanced MRI, and hyperpolarized 3He, which is considered to be the reference standard in assessing lung function using imaging. 22Nonetheless, longitudinal evaluations before and after treatment of ABPA in CF patients using morphological and functional lung MRI are lacking.We hypothesized that FD-MRI could detect response to treatment in CF patients with ABPA.
Thus, the main aim of this study was to evaluate the response to treatment of ABPA in CF patients based on FD-MRI functional changes.Our secondary aims were to: 1) evaluate morphological changes before and after treatment of ABPA using UTE-MRI; 2) investigate correlations between both morphological and functional alterations and pulmonary function test (PFT) parameters; and 3) assess the structurefunction relationships in CF patients with ABPA.

Study Population
This retrospective longitudinal pilot study was performed in a single center and approved by the local ethics committee.The requirement for written consent was waived.All consecutive CF patients referred to our institution, a tertiary CF center, were screened between August 2018 and September 2021.Inclusion criteria were: 1) diagnosis of CF proven by sweat chloride and/or genetic testing, 2) age older than 6 years, 3) diagnosis of ABPA established by multidisciplinary sessions involving pediatricians, pneumologists, physicians, immunologists, physiologists, and mycologists with full knowledge of the patient's medical history and according to the CFFC criteria. 5All included patients underwent lung MRI with UTE and FD sequences before and after treatment of ABPA using an association of oral corticosteroids and Itraconazole according to the standard of care (prednisolone 1-2 mg/kg per day for at least 8 weeks and Itraconazole 10 mg/kg per day for 4-6 months). 23edian interval between the two examinations was 5 months (minimum of 3 months and maximum of 9 months).In addition, total IgE and anti-A.fumigatus specific IgE levels were measured before and after treatment as well as PFTs within 1 week from the lung MRI examinations.A drop of at least 25% of total IgE was considered as positive response to ABPA treatments. 11mong 109 CF patients referred to our institution between 2018 and 2021, 29 patients showed an exacerbation attributed to ABPA according to the multidisciplinary care meeting and CFFC criteria.Seventeen patients were not included (nine did not undergo lung MRI before treatment and eight did not have lung MRI after treatment within a maximum interval of 1 week from the IgE measurements).Twelve patients were therefore included in this study, each having had lung MRI, total IgE and specific IgE measurements, and PFTs before and after treatment of ABPA (Fig. 1).All included patients showed a positive response to specific treatment (i.e., oral corticosteroids and itraconazole) with a drop of at least 25% of total IgE.

MRI Examinations
MRI examinations were completed on a 1.5 Tesla Siemens Aera scanner (Siemens, Erlangen, Germany).Images of a stack of spirals spoiled ultra-short gradient echo sequence (3D-UTE) were acquired in the supine position using the following parameters: TR/TE/flip angle = 4.3 msec/0.05msec/5 and a voxel size of 1 mm 3 .Respiratory synchronization at end normal expiration was achieved by automated self-navigated respiratory gating.Scan duration was between 6 and 8 minutes.Time resolved images of a 2D balanced steadystate free precession (bSSFP) sequence were acquired in two coronal planes: at the level of the carina and posteriorly at the level of the descending aorta.The main pulse sequence parameters were: TR/TE/flip angle = 1.74 msec/0.71msec/47 , pixel size of (1.2 Â 1.2 mm) 2 and slice thickness = 12 mm.Each bSSFP series consisted of 200 images.Scan duration was around 1 minute per slice for a total of two coronal slices acquired in each patient.In addition, for IMIS identification, a T2-weighted radial fast spin-echo (RFSE) sequence was acquired using the following parameters: TR/TEs/flip angle = 2350 msec/20 msec-150 msec/145 ; pixel size of (1.6 Â 1.6) mm 2 and slice thickness = 1.6 mm.A T1-weighted VIBE sequence was also acquired using the following parameters: TR/TE/flip angle = 3.37 msec/1.28msec/8 ; pixel size of (1.2 Â 1.2) mm 2 and slice thickness = 3 mm.

FD Image Processing Workflow
The image processing workflow was fully implemented in Python (version 3.9).After discarding the first 20 bSSFP images, where the steady state condition was not fulfilled, 20 the acquired image series was first elastically registered to a reference image in mid-position between full inspiration and full expiration which was determined automatically using average lung signal variation.The deformable registration was performed using the improved optical flow-based motion estimation strategy detailed in Zachiu et al, 26 which is based on a similarity term robust to local gray level intensity variations not attributed to motion and on a diffeomorphic transformation that ensures the preservation of the image topology (the image registration algorithm is included in a larger motion estimation library which can be downloaded from http://bsenneville.free.fr/RealTITracker/).Based on this reference image, a manual segmentation of the whole lung was performed and a region of interest (ROI) drawn within the aorta by two radiologists (I.B. and G.D. with 5 years and 10 years of experience in thoracic imaging, respectively).Each lung mask was automatically subdivised into three equal regions (superior, median, and inferior) based on its height.The sampling times were used to calculate the fast Fourier Transform per pixel in the segmented lung.Ventilation weighted (V) and perfusion weighted (Q) maps were then generated by taking the maximum magnitude of the peaks corresponding to the breathing frequency and the heart rate in the Fourier spectrum respectively for each pixel.In order to determine the breathing frequency, the maximal signal magnitude of the Fourier Transform of the mean signal in the lung ROI was considered.For the heart rate, the maximum in the Fourier spectrum of the mean signal within the aorta ROI was chosen.
Finally, the signal within the V-map and the Q-map was normalized using the mean signal in the aorta ROI of the reference image for each coronal slice acquired during the pre-treatment and post-treatment scans.
A muscular ROI of 5 mm of diameter was also drawn within the subscapular muscle perfectly visible on the reference image of each coronal slice acquired during the pre-and post-treatment scans to check if the overall signal varied between the two examinations before and after treatment.

Functional Analysis
QUALITATIVE EVALUATION.Defect extents on V and Q maps were independently assessed using a semi quantitative score (0 = absence/negligible, 1 = <50%, 2 = >50%) regarding the right/ left lungs, superior, median or inferior regions for each slice.The score was then averaged between the two acquired slices and ranged between 0 and 12. QUANTITATIVE EVALUATION.Mean and coefficient of variation (CV) of the signal intensity (SI) of both ventilation and perfusion maps were calculated (i.e., VSI_mean, VSI_cv, QSI_mean, and QSI_cv).

Morphological Analysis
A widely used structural alterations scoring system in CF patients was implemented 27 using UTE MR images as detailed in previously reported studies 15,28 (MR-Bhalla score; Table S1 in the Supplemental Material).Detection of IMIS was also part of the evaluation. 18hree radiologists with 5, 10, and 30 years' experience in thoracic imaging (I.B., G.D., and F.L.) performed morphological and functional evaluations.All the measurements performed in this study were averaged between the three readers.

Correlations with PFTs
Correlations of FEV1%p, FEV1/FVC, and FEF 25-75 %p with both morphological and functional lung MRI measurements were assessed before and after treatment.In addition, correlations of the changes of PFT parameters with the changes of lung MRI measurements were evaluated.Δmeasurement = measurement after À measurement before .

Structure-Function Evaluation
For the structure-function relationships evaluation, a matched 3D UTE-MRI coronal slice was assessed in terms of structural changes (i.e., bronchiectasis without mucus plugging, bronchiectasis filled with mucus plugging, and consolidation) alongside the V and Q maps.Each V and Q maps were divided into six regions (left and right lungs divided into superior, median, and inferior region) with two slices per patient and two examinations (before and after treatment) the total number of regions to classify was 576 in the 12 patients.Regions were then classified as defect regions or no defect regions and the predominant structural alteration identified.Three independent readings were carried out for these evaluations and scrolling through the adjacent slices of the 3D UTE images was allowed for identification of structural changes.A region was considered as defect region if at least two readers have classified it as defect region.If at least two readers have identified a structural alteration within regions, then it was considered as present.

Statistical Analysis
Statistical analyses were performed using MedCalc software (Version 20.216).Distribution normality was assessed using the Shapiro-wilk test.Data were expressed as medians with [minimum to maximum range] for continuous variables and absolute numbers for categorical variables.Comparisons of paired medians were performed using the Wilcoxon signed-rank test and paired percentages using the McNemar test; correlations were assessed using the Spearman test (correlation coefficient, rho).A P-value <0.05 was considered statistically significant.Reproducibility was assessed using intraclass correlation coefficients (ICCs), with mixed model analysis and absolute agreement option.ICC values were classified as null (=0), slight (>0 and <0.20), fair (≥0.20 and <0.40), moderate (≥0.40 and <0.60), good (≥0.60 and <0.80), very good (≥0.80 and <0.95), and almost perfect (≥0.95). 29Structure-function relationships were evaluated using Chi 2 tests.

Study Population
Patients' characteristics are summarized in Table 1.Median age of the included patients before treatment was 14 years with seven males and five females.The medians of total IgE and specific IgE before treatment were 667.5 and 10.5 UI/mL respectively with a significant reduction to 467.5 and 8.36 UI/mL respectively after treatment.FEV1%p showed a significant improvement after ABPA treatment with the median FEV1%p increasing from 61% to 67%.

Comparison of Functional FD-MRI Ventilation and Perfusion Before and After ABPA Treatment
The extent of the ventilation and perfusion defects decreased significantly after ABPA treatment reflecting a significant improvement of the lung function evaluated using FD-MRI (Table 2; Figs. 2 and 3).
Using the semi-automated quantification of lung ventilation, the median of VSI_mean was significantly increased after ABPA treatment (Table 2; Fig. 3) reflecting a global improvement of the lung ventilation whereas QSI_mean showed only a non-significant trend for increase after treatment (P = 0.85) (Table 2; Fig. 3).Regarding the heterogeneity of V map and Q map, the VSI_cv and QSI_cv did not demonstrate significant changes after ABPA treatment (P = 0.33 and P = 0.62, respectively) (Table 2).
No statistically significant difference was found between the mean and the standard deviation of the signal in the muscular ROIs before and after treatment (238 vs. 245 and 9.5 vs. 8.5 with P = 0.42 and P = 0.85 respectively for mean and SD).

Comparison of Morphological Bhalla Score Before and After ABPA Treatment
The median of the MR-Bhalla score was significantly increased after ABPA treatment reflecting a decrease of structural alterations' severity assessed using 3D UTE morphological MRI.In addition, six patients showed an IMIS on lung MRI before treatment which were not present on the followup MRI (P = 0.03) (Table 2).

Correlations With PFTs
Before treatment, no significant correlation was found between PFT parameters and lung MRI functional or morphological measurements (P > 0.44 for morphological measurements and P > 0.05 for functional measurements) (Table S2 in the Supplemental Material).After treatment, ventilation defects extent assessed visually was inversely correlated to PFT obstructive parameters (Table S2 in the Supplemental Material).Using quantitative imaging, QSI_mean was significantly correlated to FEV1%p and QSI_cv showed a significant inverse correlation with FEF25-75%p (Table S2 in the Supplemental Material).The MR-Bhalla score also showed a significant correlation with PFTs' obstructive parameters after treatment (Table S2 in the Supplemental Material).
However, only changes in quantitative functional imaging parameters before and after treatment (i.e., ΔVSI_cv and ΔQSI_cv) showed significant correlations with changes of PFT obstructive parameters before and after treatment (ΔFEV1%p and ΔFEF25-75%p) (Fig. 4; Table S3 in the Supplemental Material).

Assessment of Reproducibility
Very good inter-observer reproducibility was obtained for functional and morphological qualitative scores (ICC > 0.90).Reproducibility of quantitative measurements was almost perfect, with ICCs > 0.99 (Table 3).

Structure-Function Relationships
A total of 576 regions were evaluated.Three hundred and twenty-two (322) regions had no defect and 254 had defects (129 with Q defects and 125 with V defects).Regarding the defect extent, there were 118 defects with less than 50% extent and 136 with more than 50% extent.According to structural alterations evaluated in this study (Table 4), the presence of bronchiectasis without mucus plugging was found in 15 regions without any defects, and no bronchiectasis without mucus plugging was found in regions with V or Q defects.Bronchiectasis with mucus plugging was found in 38 regions without any defect and in 232 regions with V or Q defects (Fig. 5).Consolidations were found only in regions with V or Q defects (18 regions).Only four regions showed functional defects (one with V defects and three with Q defects) without any visible structural alteration.Presence of structural alterations was not different between regions with V defects and regions with Q defects (P = 0.33).The presence of functional defects (V or Q) was significantly different between regions with structural alterations and regions with no visible structural alterations.When these structural alterations were taking separately, significant difference was found between regions showing V or Q defects with mucus plugging and regions without any defects exhibiting mucus plugging.We reported also significant difference between regions with consolidation showing V or Q defects and regions without any defect exhibiting consolidation.However, no significant difference was found between regions showing V or Q defects with Bronchiectasis without mucus plugging and regions without any defects exhibiting Bronchiectasis without mucus plugging (P = 0.63).In addition, the presence of functional defects (V or Q) differed significantly between regions showing Bronchiectasis without mucus plugging and regions showing mucus plugging.The same result was found between regions showing Bronchiectasis without mucus plugging and regions showing consolidation.However, there was no significant difference between regions showing mucus plugging and regions showing consolidation (P = 0.09).Only five regions showed a mismatch between V and Q defects (P = 0.94).Moreover, no correlations were found between the MR-Bhalla morphological score and both qualitative and quantitative evaluations of ventilation and perfusion (P > 0.14 and P > 0.07 respectively) (Table S4 in the Supplemental Material).

Discussion
We demonstrated in this study that lung functional FD-MRI is able to assess ventilation and perfusion defects in CF patients with ABPA.We also showed that automatic quantification of lung V and Q is feasible in CF patients with ABPA using FD-MRI.Our findings showed that the extent of V and Q defects decreased after positive response to treatment of ABPA in CF patients.Moreover, we found an increase of VSI_mean after treatment.Our qualitative and semiautomatic quantitative methods showed a very good to almost perfect inter-reader reproducibility.Interestingly, only the  pre-and post-ABPA treatment changes in VSI_cv and QSI_cv, reflecting more homogenous V and Q maps following treatment, were found to correlate with pre-and post-ABPA treatment changes of FEV1%p.Finally, we showed that functional defects are associated with structural alterations, especially with mucus plugging and consolidations.
Regarding morphological lung MRI, our findings are in line with previously reported results which showed a complete resolution of IMIS in the follow-up. 18Moreover, a recent study 30 has shown that free breathing phase resolved functional lung MRI imaging could detect improvement following exacerbation treatments using antibiotics out of the context of ABPA.Furthermore, recent studies have demonstrated the benefit of using other MRI techniques (i.e., diffusion weighted imaging and T2 weighted imaging) in monitoring CF exacerbations. 31,32n this study, functional defects were found to be associated with ABPA exacerbation and showed a significant decrease after specific treatment.Indeed, CF patients with ABPA are prone to more severe exacerbations and require a specific treatment and repeated imaging.However, iterative radiation exposure raises concerns in young CF patients with a need of a radiation-free alternative.Moreover, FD-MRI is performed without any contrast agent administration, and thus avoids long-term effects of MRI contrast deposition. 33n this study, automatic quantification of VSI and QSI also showed an improvement in global respiratory function after specific ABPA treatment.We used a simple normalization technique to account for SI bias in longitudinal examinations and we also confirmed that the signal in a muscular ROI was not different between the pre-and post-treatment examinations.Interestingly, the pre-and post-treatment changes of VSI_cv and QSI_cv, which represent the homogeneity of V and Q maps, were correlated to the pre-and posttreatment change of FEV1%p.Thus, the decrease in the number and the extent of functional defects was associated with more homogenous maps and improved PFTs obstructive parameters.Insights into local and regional respiratory function allowed by the FD-MRI has potential to bring complementary information to the global evaluation of the respiratory function using PFTs in addition to the morphological information available from 3D UTE-MRI.
][36][37] In our study, we found that functional defects were associated with structural lung alterations without any distinction between V defects and Q defects.However, bronchiectasis without mucus plugging was not associated with functional defects assessed by using FD-MRI.Indeed, airway obstruction is more likely responsible for matched ventilation-perfusion defects.However, since we did not assess the relationship between the severity of the bronchiectasis and functional defects, our findings need to be interpreted with caution.

Limitations
First, this was a retrospective pilot study in a small number of patients.However, the patients included had a high confidence diagnosis of ABPA and also had IgE serological measurements performed before and after treatment close to the MRI examinations.However, large prospective cohorts to validate our findings are needed.Second, in this study, we did not correlate FD-MRI findings to dynamic contrast enhanced (DCE) MRI or to hyperpolarized gas MRI.However, functional FD-MRI has already been validated using these techniques. 22,38Third, CT was not available for the morphological assessment of structural changes before and after treatment.However, recent experts' recommendations state that lung MRI can be used as a surrogate marker for disease severity and response to treatment during short-term follow-up in CF patients. 39Fourth, although our qualitative evaluation of functional defects showed a high inter-reader reproducibility, it could be worth using an automatic relative quantification of defects based on threshold techniques. 40inally, a further comparison of different non-rigid image registration algorithms may provide information about the robustness of these findings regarding preprocessing techniques.

Conclusion
Functional changes in CF patients with ABPA may be reproducibly assessed using FD lung MRI and correlate with PFT changes after specific treatment of ABPA.Therefore, noncontrast enhanced functional FD lung MRI has potential be used to monitor treatment of ABPA in CF patients.Bordeaux Université-Fonds Délorme Broussin and IdEx Bordeaux LMU-Bordeaux Research Cooperation Program.

FIGURE 2 :
FIGURE 2: Coronal 2D balanced steady-state free precession image of the lungs of a 16-year-old male cystic fibrosis patient before (a) and after (d) allergic bronchopulmonary aspergillosis treatment.Note the matched defect of ventilation (b) and perfusion (c) before specific treatment (white arrows) with the decrease in the extent of the defects after treatment on both ventilation map (e) and perfusion map (f).

FIGURE 4 :
FIGURE 4: Spearman correlations between the change of FEV1%p and the change of both ventilation (a) and perfusion (b) coefficients of variation before and after treatment (ΔVSI_cv and ΔQSI_cv).

FIGURE 5 :
FIGURE 5: Matched coronal slice of the lung UTE-MRI image and Fourier decomposition ventilation map of a 16-year-old male cystic fibrosis patient before (a, b) and after (c, d) allergic bronchopulmonary aspergillosis treatment.Note the ventilation defect indicated by the solid arrows before (b) and after (d) treatment and the structural alterations (hollow arrows) seen on the UTE-MRI images showing mucus plugging (a) decreasing after treatment (c).
Data Conclusion: Non-contrast-enhanced FD lung MRI has potential to reproducibly assess response to treatment of ABPA in CF patients and correlates with PFT obstructive parameters.Evidence Level: 4 Technical Efficacy: Stage 3 J.MAGN.RESON.IMAGING 2024;59:909-919.

TABLE 1 .
Clinical and Functional Characteristics of Study Participants Data are median with [minimum-maximum range] for continuous variables and absolute number for categorical variables.ABPA = allergic bronchopulmonary aspergillosis; BMI = body mass index; IgE = immunoglobulin E; FEV1 = forced expiratory volume in 1 second; FVC = forced volume capacity; FEF25-75 = forced expiratory flow at 25%-75%; %p = percentage of predicted value.

TABLE 2 .
Imaging Characteristics of Study Participants Data are median with [minimum-maximum range] for continuous variables and absolute number for categorical variables.IMIS = inverted mucus signal intensity; V = ventilation; Q = perfusion; VSI = ventilation signal intensity; QSI = perfusion signal intensity; au = arbitrary unit; CV = coefficient of variation.

TABLE 4 .
Structure-Function Evaluation in 576 Lung Regions Data are absolute numbers.V = ventilation; Q = perfusion; MP = mucus plugging.