Quantification of 3-Dimensional Confluence-Atrial Morphology in Supracardiac Total Anomalous Pulmonary Venous Connection

Background Pulmonary vein stenosis (PVS) continues to be a major complication after surgical repair of total anomalous pulmonary venous connection (TAPVC). Recent studies suggest that the morphology of pulmonary venous confluence and the left atrium (LA) is associated with PVS. However, there are limited data on the prognostic value of integrating quantitative confluence-atrial morphology into risk stratification. Objectives This study sought to evaluate the prognostic impact of novel imaging metrics derived from 3-dimensional (3D) computed tomography angiography (CTA) modeling on postsurgical PVS (PPVS) in the supracardiac TAPVC (sTAPVC) setting. Methods Patients undergoing sTAPVC repair in 2017 to 2022 from 3 centers were retrospectively reviewed. Study investigators developed 3D CTA modeled geometric features to quantify confluence-atrial morphology that were analyzed with regard to PPVS. Results Of the 162 patients (median age 61 days; 55% having preoperative pulmonary venous obstruction [prePVO]) included, 47 (29%) with PPVS at a median of 1.5 months ([quartile 1-quartile 3: 1.5-3.0 months]). In the univariable analysis, the indexed total volume of the LA and confluence (iTVLC) and the ratio of the corresponding confluence length to the mean distance between the LA and confluence (CCL/mDBLC ratio) were significantly associated with PPVS. In a multivariable model adjusting for prePVO and age, the iTVLC and CCL/mDBLC ratio independently predicted PPVS (HR: 1.15; 95% CI: 1.06-1.25; and HR: 1.20; 95% CI: 1.08-1.35, respectively, all P < 0.01). Specifically, an iTVLC ≥20 cm3/m2 and a CCL/mDBLC ratio ≥7.7 were significantly associated with a reduced risk of PPVS. Conclusions Quantification of 3D confluence-atrial morphology appears to offer a deeper and better metric to predict PPVS in patients with sTAPVC.

D irect anastomosis between the left atrium (LA) and the pulmonary venous confluence (PVC) is the standard of care procedure for supracardiac total anomalous pulmonary venous connection (sTAPVC), which is the most common subtype of total anomalous pulmonary venous connection (TAPVC) (w40%). 1-38][9] Anatomical alignment between the PVC and the LA is crucial for avoiding geometric distortion or stretch of the individual pulmonary veins (PVs), thereby optimizing the flow pattern and lowering the risk of PPVS.Conversely, achieving a large and patent atriovenous anastomosis free of PV distortion can be challenging because it is usually complicated by morphologic heterogeneity with respect to the confluence position, size, or shape, and left atrial size. 1 Currently, morphologic assessment in this entity relies mostly on 2-dimensional (2D) echocardiography and computed tomography angiography (CTA) or cardiac magnetic resonance. 10However, plane angulation and slice selection may affect the cross-sectional 2D diameter measurements and contribute to interobserver or intraobserver variability.Additionally, the 2D measurements cannot fully describe the complex spatial information and geometric shape, thus creating an unmet need for a profound understanding of the involved 3-dimensional (3D) anatomical structure.
[13] In this study of quantifying confluence-atrial positional relationship and geometric features by using 3D computed tomography (CT) modeling in the sTAPVC setting, we noticed the association between confluence-atrial morphology and PPVS.We hypothesized that this observation could be of clinical relevance.Segmentation and 3D modeling.Details of the process were elaborated in our previous publication. 15Briefly, a convolutional neural network was used to segment the PVC and LA automatically from 3D CTA.We incorporated the attention mechanism 16 into the V-Net 17 by implementing the attention mechanism in spatial and channel attention blocks, which can guide the network to focus on the important spatial position and feature channels to extract more useful features for better segmentation performance.Thereafter, the expert investigator (Z.Y.M.) performed the manual revision to ensure the accuracy of the segmentation.

METHODS
Figure 1 shows the workflow of this process.
Quantification of the morphologic features.The centerline paths of the confluence, vertical vein (VV), and individual PVs were computed automatically using flux-driven centerline extraction. 18The confluence was marked from the plane where the right upper and inferior PVs joined the plane where the VV and left upper PV joined (Figure 2).The confluence was projected onto the left atrial surface to obtain the corresponding confluence on the basis of the stereo geometry; 15,19 thereafter, the corresponding confluence length (CCL) could be determined.The corresponding confluence was discretely sampled at 100 points (p i , I ¼ 1..100) at equal intervals.Accordingly, 100 corresponding projection points (q i , I ¼ 1..100) in the centerline of the confluence could be identified for each sampling point (p i ).The projection distance between the LA and the confluence (DBLC) could be measured, defined as the distance between p i and q i , and therefore the mean DBLC (mDBLC) could be calculated (Figure 2).Volume measurements of the confluence and LA were estimated by counting voxels in their segmentation areas (Figure 2).In addition, we used a restricted spline curve with 4 knots at the 5th, 35th, 65th, and 95th percentiles of the A time-dependent area under the receiveroperating characteristic (ROC) curve 20     DERIVATION COHORT.We assessed the associations of CTA-derived morphologic parameters with PPVS after adjusting for patient age and prePVO (Table 2).
The time-dependent ROC curves revealed that the discrimination of each CTA-derived morphologic parameter outperformed prePVO in predicting PPVS (Figure 3A).Interestingly, iTVLC showed better prognostic value than iPVC volume or iLA volume alone.

OUTCOME ANALYSES IN THE COMBINED COHORT.
We combined the derivation and validation cohorts to increase the sample size (N ¼ 162) further, to facilitate the follow-up analyses.In the univariate analysis for  1).
Furthermore, we categorized those patients who had an iTVLC >20 cm 3 /m 2 or a CCL/mDBLC ratio >7.7 into a low-risk group if there was no observed PPVS; otherwise, the patients were considered to be at moderate risk of PPVS (Figure 5).

DISCUSSION
This study demonstrates the critical concept for quantitative 3D morphologic analysis in sTAPVC.The key findings (Central Illustration) are as follows.First, the iTVLC and CCL/mDBLC ratio are 2 useful 3D CTAderived quantitative morphologic features to identify Unfortunately, studies focusing on this issue are limited, and most have been qualitative studies. 1,13e present work bridges this knowledge gap by analyzing the geometric features in sTAPVC on the basis of 3D CT modeling, in the hope of selecting patients at increased risk of PPVS.
We took the sum of volumes of confluence and LA as an indicator and found that it showed better prognostic value for predicting PPVS than LA or PVC volume alone after adjusting for patients' BSA.
Additionally, our data indicated that a smaller iTVLC had a strong association with prePVO.These findings merit special discussion.First, a study demonstrated that patients with obstructed pulmonary venous return had decreased left atrial reservoir and contractile functions. 23Moreover, such impairment in left atrial function exists even after TAPVC repair even though the small left side of the heart can increase postoperatively. 24There is a possibility that the mismatch reservoir function of the LA exposes the PV system to a high filling pressure postoperatively and drives the onset of vascular remodeling. 25This speculation is partially supported by the pathologic findings in the left atrial stenosis rat model, which revealed a progression of stenosis extending to the upstream PVs characterized by intimal fibrosis and medial hypertrophy, closely resembling the pathophysiology of PPVS. 26Second, there are some suggestions that a small confluence and LA add difficulty in achieving a large and patent atriovenous anastomosis that has a strong association with PPVS. 1,27lumetric measurement of the LA and PVC is necessary but not sufficient because reconstruction of the anomalous PV drainage to the LA requires comprehensive consideration of not only the involved structures per se but also their positional relationships.We used mDBLC and CCL to quantify the spatial relationship between the confluence and the LA.Further, we uniquely integrated these 2 parameters and used the CCL/mDBLC ratio as an indicator, and we observed that a smaller CCL/mDBLC ratio was associated with an increased risk of PPVS.A small CCL/mDBLC ratio reflects the following: 1) the confluence is relatively far from the LA; and 2) there is less alignment between the confluence and the LA preoperatively.These 2 morphologic components can be impediment to perfection of direct atriovenous anastomosis.These findings provide evidence to support the previous speculation that a greater distance from the confluence to the atrial mass may increase the risk of PPVS.
Compared with prePVO, integration of the 2 3D CT modeled morphologic parameters improves risk prediction for PPVS.There are several explanations.Several aspects warrant additional clarification.
First, the higher-velocity threshold (1. STUDY DESIGN.We conducted a retrospective, observational study involving 3 tertiary hospitals in China (Shanghai Children's Medical Center [SCMC], Guangdong General Hospital, and Hunan Children's Hospital).SCMC served as the data coordinating center and received the study data transmitted from the other 2 clinical sites.Institutional Review Board approval was received for each site, and written informed consent was obtained from all participants' guardians.This study complies with the Declaration of Helsinki.PATIENTS AND DATA COLLECTION.Patients undergoing surgical repair for sTAPVC between March 2017 and September 2022, with CTA performed shortly before surgery, were consecutively included.Participants from SCMC formed the derivation cohort, and participants from the other 2 hospitals formed the external validation cohort.Exclusion criteria were as follows: 1) patients aged >18 years; 2) patients having TAPVC in association with heterotaxy, atrial isomerism, or other cardiac anomalies except atrial septal defect (ASD) or patent ductus arteriosus; 3) patients without available preoperative CTA scans or with inadequate CTA image quality to allow reliable morphologic analysis; and 4) patients who did not undergo atriovenous anastomosis.Data collection included the following: patient baseline demographics (age at surgery, sex, body surface area [BSA], oxygen saturation); preoperative and surveillance imaging information, including echocardiographic and CTA data; clinical information on requirement of preoperative treatment (mechanical ventilation, infusion therapy); time of operation (emergency [#24 hours on presentation] vs nonemergency surgery [>24 hours on presentation]); procedural details; postoperative course during hospitalization; and follow-up status.Data were collected from the electronic medical records and were analyzed retrospectively.The echocardiographic data were reread and reanalyzed centrally in the echocardiographic core laboratory of SCMC.Preoperative pulmonary venous obstruction (prePVO) was defined on the basis of the combined evaluation of oxygen saturation (<90% at rest), echocardiographic reports (nonphasic Doppler velocity of >1.8 m/s within the pulmonary venous draining course and/or restrictive ASD 11 ), and/or CTA findings (minimal diameter/reference vessel diameter <50% 14 ).Specifically, Doppler velocity of >1.8 m/s has been consistently used as the diagnostic criterion A B B R E V I A T I O N S A N D A C R O N Y M S ASD = atrial septal defect AUC = area under the curve BSA = body surface area CCL = corresponding confluence length CT = computed tomography CTA = computed tomography angiography DBLC = distance between left atrium and confluence iLAvolume = body surface areaadjusted volume of left atrium iPVCvolume = body surface area-adjusted volume of the pulmonary venous confluence iTVLC = body surface areaadjusted total volume of left atrium and confluence LA = left atrium Confluence-Atrial Morphology in sTAPVC of pulmonary vein stenosis (PVS) since our center established the TAPVC program in 2016. 3,4,11,12CTA IMAGING ACQUISITION AND ANALYSIS.The CTA data (full DICOM [Digital Imaging and Communications in Medicine] data sets) were transmitted to the core laboratory of SCMC.For the purpose of this study, the CTA data were reread and reanalyzed by a single expert investigator (Y.Z.) with >20 years of experience in pediatric cardiac imaging who was blinded to the follow-up data.

FIGURE 1
FIGURE 1 Study Workflow

FIGURE 3
FIGURE 3 Time-Dependent Receiver-Operating Characteristic Analysis of PPVS

FIGURE 4
FIGURE 4 PPVS Analysis According to the Cutoff Value of Imaging Metrics

FIGURE 5
FIGURE 5 Risk Stratification on the Basis of the Imaging Metrics

First, prePVO consists
of external expression in an anomalous connector vein, hypoplastic confluence or PVs, and functional obstruction attributed to restrictive ASD.In reality, restrictive ASD or external compression influences the procedural complexity less and can be relieved with a benign prognosis, whereas venous size is potentially the important factor as a technical issue.More importantly, the confluence features an inherent variation in not only size but also shape and orientation, thus introducing patient-specific particularities, which cannot be completely represented by prePVO.Second, an echocardiogram is an important imaging metric for assessment of prePVO, but the standard definition remains elusive, with velocity thresholds varying from 1.2 to 2 m/s among different centers.28The angle of interrogation and decreased blood flow attributed to stenosis may cause inconsistence between Doppler velocity and the severity of obstruction, thereby miscategorizing patients at risk of PPVS.Our findings may produce clinically relevant benefits in optimizing care management for this subgroup of patients in terms of assisting surgical decision making and tailoring postsurgical surveillance.The quantitative descriptors of the confluence and the LA, as well as their spatial relationships, help surgeons better comprehend patient-specific anatomy.For patients with a small iTVLC and/or CCL/mDBLC ratio, technical modifications (ie, L-shaped incision with patch augmentation,29 U-shaped flap technique,30 or window surgery31 ) can be used to CENTRAL ILLUSTRATION Risk Stratification by Morphologic Quantification Using Advanced 3-Dimensional Computed Tomography Modeling Shi G, et al.JACC Asia.2024;4(8):594-606.Morphologic quantification derived from advanced 3-dimensional computed tomography modeling assists in individualized risk stratification.CTA ¼ computed tomography angiography.J A C C : A S I A , V O L . 4 , N O .8 , 2 0 2 4 Shi et al A U G U S T 2 0 2 4 : 5 9 4 -6 0 6 3-Dimensional Confluence-Atrial Morphology in sTAPVC achieve better alignment between the LA and the PVC as well as a large anastomosis.Additionally, increased predictive ability for PPVS by using the advanced imaging metrics can help to establish an individualized follow-up strategy.More regimented postoperative surveillance should be required for patients at higher risk for PPVS to ensure anticipatory intervention or targeted chemotherapeutic treatment in early disease to avoid aggressive propagation to upstream PV segments where treatment efficacy is limited. 9 Emerging evidence shows a relationship between the preoperative confluence-atrial morphology and PPVS in patients with TAPVC, thus creating an unmet need to characterize the confluence-atrial morphologic features more clearly.Because of the tremendous morphologic heterogeneity in these patients, single 2D size assessment cannot fully describe the confluence-atrial morphology.Our study suggests that quantifying the geometric features of the LA and PVC as well as the positional relationship between the 2 anatomical structures after accurate 3D CT modeling enhances risk evaluation in the setting of sTAPVC, and it appears to be a deeper and better metric than preoperative PVO alone to predict PPVS.Moreover, the study potentially provides implications for the clinical utility of morphologic quantification through advanced 3D CTA modeling in individualized patient care.TRANSLATIONAL OUTLOOK: Future prospective, large-scale validation studies are needed to confirm the value of presurgical quantification of the confluenceatrial morphology with advanced CTA imaging in patientspecific risk stratification and to evaluate the clinical relevance of surgical strategy selection on the basis of these morphologic parameters.J A C C : A S I A , V O L . 4 , N O .8 Confluence-Atrial Morphology in sTAPVC

TABLE 1
Patient Baseline Characteristics, Intraoperative and Postoperative Data, and CTA-Derived Morphologic Parameters volume of left atrium after indexing to BSA; iPVCvolume ¼ volume of pulmonary venous confluence after indexing to BSA; iTVLC ¼ BSA-adjusted total volume of the left atrium and confluence; mDBLC ¼ mean distance between left atrium and pulmonary venous confluence;; PPVS ¼ postsurgical pulmonary vein stenosis; PVO ¼ pulmonary venous obstruction; SpO2 ¼ percutaneous arterial oxygen saturation; sTAPVC ¼ supracardiac total anomalous pulmonary venous connection; VV ¼ vertical vein.

TABLE 2
Associations Between the CTA-Derived Metrics and Postsurgical Pulmonary operation (HR: 2.05; 95% CI: 1.14-3.69;P ¼ 0.017), and age (HR: 1.01; 95% CI: 1.00-1.02;P ¼ 0.003).Only prePVO (HR: 1.91; 95% CI: 1.04-3.53;P ¼ 0.038) was independently associated with PPVS in the multivariate analyses.We assessed the association between the CTA metrics (iTVLC, CCL/mDBLC ratio) after adjusting prePVO, and these 2 imaging ratio.Third, not all patients with sTAPVC during the study period were assessed, given the absence of preoperative CT scans or adequate image quality; accordingly, we cannot exclude a selection bias.Finally, because this is an observational study, we are unable to identify causative mechanisms, and as such, the results should be seen as hypothesis generating.Seale AN, Uemura H, Webber SA, et al.Total anomalous pulmonary venous connection: morphology and outcome from an international population-based study.Circulation.2010;122: 2718-2726.2. Tucker BL, Lindesmith GG, Stiles QR, Meyer BW.
R E F E R E N C E S 1.