**1. Introduction**

Pulmonary arteriovenous malformations (PAVM) are vascular malformations frequently encountered in patients with hereditary hemorrhagic telangiectasia (HHT), also known as Rendu–Osler–Weber disease. In this type of vascular malformation, the normal capillary bed is absent and there is a direct connection between the pulmonary artery and vein. As a result, unfiltered and unsaturated blood enters the systemic circulation, which may cause hypoxemia and forms a potential conduit for paradoxical emboli. The latter may lead to serious clinical complications; the most feared are cerebral infarctions or cerebral abscesses [1,2]. Hence, it is important that these vascular malformations are detected and treated before they cause irreversible damage. In the 2011 International Guidelines for the Diagnosis and Management of HHT, it is recommended that all PAVMs with a feeding artery diameter (FAD) of 3 mm and, if feasible, >2 mm are treated [3]. Computed tomography (CT) is considered the imaging modality of choice to diagnose these PAVMs [3,4]. The drawback of this technique is that it requires ionizing radiation. This is important to realize because in this relatively young population CT investigations are used during long-term follow-up to assess PAVM persistence after embolotherapy and growth of untreated PAVMs [5–10]. Hanneman et al., reported an average of four chest CT scans (range 0–20) per patient and that 11% of their study cohort received a cumulative effective dose of >100 mSv at which patients are considered to be at risk for radiation induced harm [11]. Advancements in low-dose and ultra-low-dose (submillisievert) CT imaging have significantly lowered radiation burden [12]. This facilitates a more liberate use of CT in PAVM management, probably without increasing the radiation risks. However, the 'as low as reasonably achievable' principle still applies in radiology and technical developments in magnetic resonance imaging (MRI) offer the possibility of dynamic and functional imaging of PAVMs without radiation burden. It is therefore, interesting to investigate the role of MRI in PAVM management. A first prerequisite is that MRI can reliably detect PAVMs. Therefore, the aim of this study was to investigate the sensitivity of contrast-enhanced MR angiography (CE-MRA) compared to CT in the detection of PAVMs with a FAD > 2 mm.

#### **2. Materials and Methods**

All subjects gave their informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of Helsinki and was approved by the institutional review board of our hospital (MEC-U NL39415.100.12).

#### *2.1. Patient Population*

Consecutive patients presenting at the HHT outpatient clinic (a tertiary referral center for HHT), suspected having untreated PAVMs based on a moderate or severe shunt (grade 2 or 3) diagnosed by transthoracic contrast echocardiography (TTCE), were asked to participate in the study. As part of our protocol, all patients with a moderate or severe shunt on TTCE undergo a chest CT scan to evaluate the presence of PAVMs. After signing informed consent, they also underwent a CE-MRA of the pulmonary arteries.

#### *2.2. Trans Thoracic Contrast Echocardiography Protocol*

A detailed description of the TTCE protocol can be found in the paper by Velthuis et al. [13]. The contrast agent, consisted of 8 mL of saline 0.9%, 1 mL of room air and 1 mL of blood, was administration through the right antecubital vein. The number of microbubbles in the left side of the heart was counted on one still frame. A moderate (grade 2) or large (grade 3) shunt was present if 30–100 or >100 microbubbles appeared in the left ventricle, respectively. All shunts visualized in the left atrium through a pulmonary vein or after at least four cardiac cycles were classified as a pulmonary shunt.

#### *2.3. CT Acquisition Protocol*

All CT scans were performed on a 256-slice scanner (Philips Brilliance iCT, Best, The Netherlands). Scan direction was from the lung bases to apices and in a single breath hold. Scan parameters were 100 kV, reference mAs 120, collimation 128 × 0.625, pitch 0.68. Images were reconstructed at 1 mm slice thickness. In case of a previous embolization procedure a contrast enhanced CT scan was performed. Scan parameters were identical to nonenhanced CT scans. Scan delay after administration of 80 mL Iobitridol (Xenetix ® 300, Guerbet Laboratories, Roissy, France) was 30 s. A 40 mL saline flush was administered at the beginning of the scan.

#### *2.4. CE-MRA Acquisition Protocol*

CE-MRA was performed at 1.5 T (Philips Achieva, Best, The Netherlands) using a five-element cardiac coil. Time-resolved CE-MRA acquisition was with a T1 weighted fast field echo (FFE), repetition/echo time of 2.9/1.5 and a flip angle of 30 degrees. Field of view was 320 × 345 with a 200 × 200 matrix, 230 slices with slice oversampling factor of 1.1, slice thickness reconstruction 1.5 mm and 0.57 × 0.57 × 1.5 voxel. The resulting sagittal 3D scan was repeated three times: noncontrast, arterial phase and venous phase. Contrast agen<sup>t</sup> used is Dotarem (Guerbet 0.5 mmol/m; 30 mL flow 2.5 mL/sec followed by saline flush of 20 mL).

#### *2.5. CT and CE-MRA Assessment*

There were two readers in this study. Reader 1 (R1) with >10 years of experience in chest and interventional radiology was also involved in the development of the MRA protocol. Reader 2 (R2) was a resident in radiology and reviewed five learning cases in a session with R1 prior to study assessment. First, all CT scans were reviewed by R1. The CE-MRAs were independently and randomly reviewed by R1 and R2. To minimize recall bias, the time interval between the CT and the CE-MRA reviewing sessions was > four weeks for R1. All scans were assessed for the presence and location of PAVMs. Both readers scored the location per segmen<sup>t</sup> on CE-MRA to enable adequate correlation between the CT and MRA. For this the Jackson and Huber classification was used [14]. R1 measured the feeding artery diameter on CT within 1 cm of the sac. The 2011 guidelines recommend embolization of PAVMs with feeding artery diameter of 3 mm and if technically feasible also of PAVMs with feeding artery diameters of 2 mm [3]. In the current study, all PAVMs with a FAD of >2 mm were assessed. For the assessment of the CE-MRA the sagittal source images as well as axial 1 mm reconstructions and the sagittal and axial maximum intensity projections were used (Figure 1).

**Figure 1.** (**A**) Computed tomography (CT) scan 1.5 mm slice thickness with an inconspicuous PAVM (pulmonary arteriovenous malformation) in the medial segmen<sup>t</sup> of the middle lobe (arrow). (**B**,**C**) CE-MRA (contrast enhanced magnetic resonance angiography) axial reconstruction and axial MIP (maximum intensity projection) demonstrating the same PAVM.

## *2.6. Statistical Analysis*

CT was considered the reference standard and the CE-MRA the index test. Per-patient and per-PAVM analysis were performed for both readers. The per-patient analysis was limited to distinguishing between the presence or absence of PAVMs with a feeding artery diameter of ≥2 mm in an individual. The per-PAVM analysis was focused on finding and correlating all PAVMs per segment. Sensitivity, specificity and negative predictive values (NPV) and positive predictive value (PPV) were calculated for the per-patient analysis. For the per-PAVM analysis the same test statistics were calculated taking into account that for each patient 18 segments were screened. With this approach it was also possible to determine the specificity and NPV for the per-PAVM analysis. Statistical analysis was performed with R version 3.6.3 [15].
