Cardiac magnetic resonance imaging-derived septum swing index detects pulmonary hypertension: A diagnostic study

Abstract Background and Objectives Because of pressure differences between the pulmonary artery and aorta, the ventricular septum moves in a swinging motion that is commonly observed on cardiac MR (CMR) cine sequences in patients with pulmonary hypertension (PH). We aimed to assess the use of septum swing index (SSI) derived by CMR for detecting PH. Methods We retrospectively identified consecutive patients with suspected PH who underwent right heart catheterization (RHC) and CMR at a PH referral center between July 2019 and December 2020. The diagnostic accuracy of SSI for identifying PH (mean pulmonary artery pressure [mPAP] ≥ 25 mmHg) was assessed by receiver operating characteristic curves, sensitivity, specificity, and positive and negative predictive values. Results A total of 105 patients (mean age: 47.8 ± 15.0 years; 68 females) were included in the final analysis. SSI and mPAP were negatively correlated in the total study population and patients with PH, but not in patients without PH. SSI was an independent predictor of PH (adjusted odds ratio: 12.9, 95% confidence interval: 3.6 to 45.5, P = 0.003). The area under the curve for SSI was 0.91, with a cut-off value of 0.9673 yielding the best balance of sensitivity (86.4%), specificity (88.2%), positive predictive value (97.4%), negative predictive value (55.6%), and accuracy (86.7%) for detecting PH. Conclusions Septum swing index was lower in patients with PH and is a simple, reliable method for detecting PH.


INTRODUCTION
Pulmonary hypertension (PH) is a severe clinical condition characterized by increased pulmonary vascular resistance and right ventricular remodeling, leading to right heart failure. [1,2]5][6] However, unlike the left ventricle (LV), the complex chamber geometry and suboptimal endocardial definition of the right ventricle (RV) have limited the application of noninvasive RV function assessment in clinical practice. [7]] In patients with PH, the circular shape of the LV on crosssectional images changes to a "D" shape, and the crescentic RV assumes a more circular shape. [10]Changes in ventricular geometry can be represented by the left ventricular eccentricity index.[16][17] Interventricular septum swing is frequently observed on CMR cine sequences of patients with PH.Pathophysiologic changes in PH, with alterations in the normal pressure difference between the pulmonary artery and aorta, lead to changes in the position of the interventricular septum (an elastic tissue between two pressure sources: the RV and LV), as well as deformation of the ventricular chambers.However, no simple quantitative method has been reported describing the link between the degree of ventricular chamber deformation detected by non-invasive MR imaging and pulmonary artery pressure values obtained by invasive, "gold standard" right heart catheterization (RHC).
In this study, we introduce a very simple CMR imagingbased deformation value, which we have called the septum swing index (SSI), and investigate the utility of SSI for assessing RV hemodynamic changes in PH.Specifically, we analyze the correlation between SSI and the mean pulmonary arterial pressure (mPAP) obtained by RHC.We hypothesize that SSI will be an accurate reflection of mPAP alteration and RV remodeling in patients with PH and may play an important role in the long-term monitoring of these patients.

MATERIALS AND METHODS
The protocol for this retrospective study was approved by our local ethics committee (L21-387), and informed consent was obtained from all patients.
We screened all patients with suspected PH who underwent RHC and CMR at a PH referral center between July 2019 and December 2020.All included patients met the following criteria: (1) World Health Organization (WHO) group 1 or 4 PH based on the 6th World Symposium on PH, [18] including idiopathic, heritable, associated with connective tissue disease, and chronic thromboembolic PH;

Hemodynamic measurement
RHC was performed as described previously, [19] The baseline hemodynamic variables evaluated included mPAP, pulmonary artery wedge pressure, cardiac output, cardiac index and pulmonary vascular resistance.

Cardiac magnetic resonance procedure and image analysis
All CMR images were analyzed by two radiologists who had at least 10 years' experience with interpreting CMR results and who were unaware of the RHC results (M.H. and J-Y.S.).Endocardial contours of the LV and RV and diastolic and systolic perimeter and area were obtained on CMR images.[22] Detailed procedures and analyses are presented in the supplementary text.All patients underwent RHC and CMR within 7 days.

Septum swing index calculation
The senior MRI data expert (F.S.) created the SSI calculation, which consists of three steps: Step 1: Calculate the SSI at end-diastole Select two consecutive slices that show the maximum LV chamber area in the short-axis cine sequence and draw a complete outline of the LV of each slice using any measurement software (e. g.Radiant [www.radiantviewer.com]) to obtain the slice area and corresponding slice perimeter.The diastolic SSI (SSI diastolic ) is calculated using this formula: Step 2: Calculate the SSI at end-systole Repeat the same calculation as step 1 for the end-systolic phase cine sequence to obtain the systolic SSI (SSI systolic ).
Step 3: Calculate the final SSI Based on the RHC-based mPAP formula ( ), the corresponding final SSI is calculated as follows: SSI =   + 2 ×   3 .

SSI numerical approximation
We approximated the interventricular septum movement as the swinging movement of an elastic rope connected to both ends of a semicircle with a fixed radius of R (Figure 1).The elastic rope (solid line in the figure) corresponds to the area of the LV with a semicircle formation area.During ventricular septum swinging, the pressure distribution from the left and right chambers is basically uniform, so it can be assumed that the elastic rope (septum) will remain approximately in the shape of an arc fixed at the endpoints.
In this swinging process, we find that there is an index involving LV area that is related only to the amount of ).
When the elastic rope is at the left of the circle center: When the elastic rope is at the middle of the circle center: When the elastic rope is at the right of the circle center: . When L= R, SSI = 0.

Statistical analysis
Data were expressed as number, percentage, median with interquartile range, or mean with standard deviation.Parameters between groups were compared using the Mann-Whitney U-test or unpaired Student t-test for continuous data and the Chi-square test or Fisher's exact test for categorical data. [23,24] luding collinearity and Bonferroni-type adjustments, binary univariate and logistic regression analyses were performed to evaluate SSI and conventional CMR predictors of PH.Receiver operating characteristics curve methodology was then used to assess the ability of SSI and conventional CMR parameters to detect PH.Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated.Present estimates suggest a PH prevalence of 1% of the global population. [25]After PH prevalence-adjusted, sensitivity, specificity, positive predictive value and negative predictive value are also calculated, respectively.Interobserver correlations and agreement were determined using Pearson's correlation coefficient and the intraclass correlation coefficient (single scoring, not adjusted).

Characteristics of the study population
We screened 109 consecutive patients with clinically suspected PH who underwent both RHC and CMR.A total of 105 patients were included in the final analyses (Figure 2).The included patients had a mean age of 47.8 ± 15.0 years, and 68 (64.8%) were female.Demographic information and hemodynamic measurements obtained at RHC are shown in Table 1.

Relationship between mPAP and quantitative CMR parameters
Conventional CMR-derived morphologic and functional parameters related to PH are shown in Table 2. Structure and function parameters of both the LV and RV differed significantly between patients with and without PH.Conventional parameters for detecting PH, such as main pulmonary artery diameter and VMI were significantly higher, whereas SSI was significantly lower in patients with PH.SSI and mPAP were negatively correlated in the whole study population and in patients with PH but not in patients without PH (Figure 3A).By contrast, VMI and mPAP were positively correlated in the whole study population and in patients with PH (Figure 3B).While main pulmonary artery diameter and mPAP were positively correlated in the whole study population (Spearman rank correlation coefficient = 0.307, P = 0.001), they were not significantly correlated in patients with PH or without PH (Figure 3C).Both interobserver and intra-observer agreement were excellent for SSI (intraclass correlation coefficient = 0.970 and 0.930, respectively, both P < 0.001).

Factors predicting PH
Table 3 shows the results of univariate binary logistic regression analysis, including unadjusted odds ratios for PH for each variable.After adjusting by Bonferroni P values, PH remained significantly associated with main pulmonary artery diameter, LV end-diastolic volume index, RV ejection   fraction, RV end-systolic mass, RV end-systolic volume index, RV end-systolic mass index, and SSI.After assessing collinearity of variables in linear regression analysis of the association between CMR with mPAP, we excluded LV end-diastolic volume index and RV end-systolic mass index.The remaining factors (main pulmonary artery diameter, RV ejection fraction, RV end-systolic mass, RV endsystolic volume index, and SSI) that were significant after Bonferroni adjustment were then entered into stepwise multivariable binary logistic regression to analyze predictors of PH.This revealed SSI as an independent predictor of PH (adjusted odds ratio: 12.9, 95% confidence interval: 3.6 to 45.5, P = 0.003).
The area under the receiver operating characteristics curve for SSI as a method of detecting PH was 0.91 (Figure 4

DISCUSSION
This study showed that the septum swing index determined by cardiac MRI provides a simple quantitative method that links the degree of ventricular chamber deformation and mean pulmonary arterial pressure and accurately detect PH.
By numerical simulation, we can obtain a correlation elastic curve between the amplitude of an elastic rope moving from side to side and SSI.The horizontal coordinates are the ratio of the swing amplitude of the middle point of the elastic rope (relative to the initial position) to the radius of the initial circumference, and the ordinate is SSI.This curve shows that SSI changes synchronously from 1 to 0 as the septum swings from side to side.Theoretically, there is a strong correlation between the swing amplitude and PAP, so we predict that SSI and PAP will also be correlated, and when PAP is higher, SSI changes will be more obvious.
SSI has several advantages for clinical use.First, it has good sensitivity and specificity.CMR imaging constitutes one of the most complete diagnostic modalities for diagnosing PH, as it evaluates both morphology and hemodynamics of the pulmonary artery and RV.Several cine steady-state freeprecession-derived parameters (RV end-diastolic volume index or RV stroke volume index) and phase-contrast regional area changes have been suggested as powerful biomarkers for use in prognosis and treatment.RV end- diastolic volume index ≥ 84 mL/m 2 , RV end-systolic volume index ≥ 70 mL/m 2 , RV stroke volume index < 25 mL/m 2 , LV end-diastolic volume index < 40 mL/m 2 , and RV mass index > 59 g/m 2 have all been associated with worse prognosis.Previous studies have reported that pulmonary flow artifact can be used to predict PH, with a sensitivity of 86%, specificity of 85%, and positive predictive value of 95%.Other reports found that VMI had a sensitivity of 81%-98% and specificity of 69%-89%, with an optimal cut-off value of 0.45, [26] and pulmonary artery mean velocity > 11.7 cm/sec had a sensitivity of 92.9% and specificity of 82.4%. [27]By comparison, SSI has a high sensitivity (86.4%), specificity (88.2%), and positive predictive value (97.4%) for detecting PH, with an optimal cut-off value of 0.9673.
Secondly, calculating SSI is a simple procedure.SSI values are determined by measuring the maximum or minimum area and diameter during diastole and systole on shortaxis cine images and then inserting these measurements into a fixed formula.Measuring diastolic and systolic area and diameter is the most basic CMR technique, with low technical difficulty.Conversely, CMR image processing techniques based on steady state free preceesion (SSFP) sequences generally require independent and complex software.
Another advantage of SSI is that it reflects cardiac remodeling, combining right and left heart interactions.The right heart affects left heart function through the interventricular septum.Finally, SSI has broad clinical potential for  predicting prognosis and assessing the severity of right heart dysfunction. [28]It may be combined with PH risk grades for prognosis prediction and may be used to explore correlations with other pulmonary vascular remodeling parameters, such as pulmonary vascular resistance, cardiac output, and mixed venous oxygen saturation.
Based on its numerical approximation, SSI is a function related to only L/R, representing the swing distance divided by the LV radius.The rationale for not measuring these two lengths directly is because of measuring reproducibility.When measuring a distance, two exact points must be chosen.These points are difficult to define or reproduce when the target region is the chamber of a cardiac ventricle, which is not uniform in shape and contains chordae tendineae.By contrast, drawing a smooth contour inscribing the LV chamber is intuitively easy to accomplish.
Our results showed that SSI values were determined consistently between reviewers.
Although averaged over two slices, SSI determination is based mostly on 2-dimensional information, while the actual septum moves in three dimensions.In the future, we plan to conduct studies including long-axis cine images to cover more LV mass or studies using a multi-slice weighted average algorithm, which may correlate even more strongly with mPAP.To reduce the radiologist's workload and increase data repeatability, automatic LV chamber fitting and an automated algorithm for SSI calculation may also be developed.
Our numerical approximation also showed that SSI is positively correlated with the L/R ratio but not in a linear manner.When the septum swung from the left to the middle, SSI only changed from 1 to 0.745, but when the septum swung from the middle to the right, it decreased more rapidly to 0. This phenomenon suggests that in addition to discriminating patients with PH from those without PH, SSI may be especially sensitive for detecting changes in patients with severe PH during treatment or long-term follow-up.
In the study, SSI was proved to be with satisfying AUC, sensitivity, specificity, PPV, NPV and accuracy for detecting PH.After PH prevalence -adjusted, sensitivity, specificity and AUC still remained.However, PPV had descended from 97.4% to 6.9%, and NPV had increased from 55.6% to 99.8%.Because PH is a rare disease, the prevalence rate is only 1%.The low prevalence rate has a significant impact on both PPV and NPV.For the general population, SSI has a high NPV, but a low PPV.However, at the hospital level, patients have symptoms to seek medical advice, so the probability of PH will also be greatly increased, so it is not suitable for screening of natural populations.For patients with suspected PH symptoms, SSI is still of great diagnostic value, because the sensitivity and specificity are
OF TRANSLATIONAL INTERNAL MEDICINE / OCT-DEC 2023 / VOL 11 | ISSUE 4 septum movement and not to the initial radius R of the left chamber (solid line).Thus, SSI = 4π ×   2 .The calculation process is as follows: Set the distance between the midpoint of the elastic rope and the center of the circle as L and d =   ,  =      1

Figure 1 :
Figure 1: Schematic diagram and diastolic and systolic cardiac MR images of SSI calculation in patients with various mPAP values.(A) Normal mPAP; (B) mPAP of 35 mm Hg; (C) mPAP of 56 mm Hg; and (D) mPAP of 81 mm Hg. mPAP: mean pulmonary artery pressure; SSI: septum swing index.Set the distance between the midpoint of the elastic rope and the center of the circle as L, R indicates the radius; d = L/R.

Figure 4 :
Figure 4: Receiver operating characteristics curve showing that SSI has good diagnostic accuracy for predicting pulmonary hypertension (mPAP ≥ 25 mm Hg).AUC: area under the receiver operating characteristics curve; mPAP: mean pulmonary arterial pressure; SSI: septum swing index.

Table 2 : Cardiac magnetic resonance imaging parameters in patients with or without pulmonary arterial hypertension
Data expressed as mean ± standard deviation or median (interquartile range).LV: left ventricle; RV: right ventricle; SSI: septum swing index; VMI: ventricular mass index.JOURNAL OF TRANSLATIONAL INTERNAL MEDICINE / OCT-DEC 2023 / VOL 11 | ISSUE 4

Table 3 : Univariable analyses of cardiac magnetic resonance imaging predictors of pulmonary arterial hypertension
Continuous variables were transformed into binary variables stratified by 0.6.‡ Variables were categorized according to interquartile ranges (25 th to 75 th ).LV: left ventricle; RV: right ventricle; SSI: septum swing index; VMI: ventricular mass index.

Table 4 : Comparisons of the area under the curve of cardiac magnetic imaging for detecting pulmonary hypertension
† Continuous variables were transformed into binary variables stratified by 0.6.SSI: septum swing index; VMI: ventricular mass index; AUC: area under the receiver operating characteristics curve; SE: standard error: CI: Confident interval.JOURNAL OF TRANSLATIONAL INTERNAL MEDICINE / OCT-DEC 2023 / VOL 11 | ISSUE 4