Integrating CT-Based Lung Fibrosis and MRI-Derived Right Ventricular Function for the Detection of Pulmonary Hypertension in Interstitial Lung Disease

Abstract

Background/Objectives: Interstitial lung disease (ILD) is frequently complicated by pulmonary hypertension (PH), which is associated with reduced exercise capacity and poor prognosis. Early and accurate non-invasive detection of PH remains a clinical challenge. This study evaluated whether combining quantitative CT analysis of lung fibrosis with cardiac MRI-derived measures of right ventricular (RV) function improves the diagnostic accuracy of PH in patients with ILD. Methods: We retrospectively analyzed 72 ILD patients who underwent chest CT, cardiac MRI, and right heart catheterization (RHC). Lung fibrosis was quantified using a Gaussian Histogram Normalized Correlation (GHNC) software that computed the proportions of diseased lung, ground-glass opacity (GGO), honeycombing, reticulation, consolidation, and emphysema. MRI was used to assess RV end-systolic volume (RVESV), ejection fraction, and RV longitudinal strain. PH was defined as a mean pulmonary arterial pressure (mPAP) ≥ 20 mmHg and pulmonary vascular resistance ≥ 3 Wood units on RHC. Results: Compared to patients without PH, those with PH (n = 21) showed significantly reduced RV strain (-13.4 ± 5.1% vs. -16.4 ± 5.2%, p = 0.026) and elevated RVESV (74.2 ± 18.3 mL vs. 59.5 ± 14.2 mL, p = 0.003). CT-derived indices also differed significantly: diseased lung area (56.4 ± 17.2% vs. 38.4 ± 12.5%, p < 0.001), GGO (11.8 ± 3.6% vs. 8.65 ± 4.3%, p = 0.005), and honeycombing (17.7 ± 4.9% vs. 12.8 ± 6.4%, p = 0.0027) were all more prominent in the PH group. In receiver operating characteristic curve analysis, diseased lung area demonstrated an area under the curve of 0.778 for detecting PH. This increased to 0.847 with the addition of RVESV, and further to 0.854 when RV strain was included. Combined models showed significant improvement in risk reclassification: net reclassification improvement was 0.700 (p = 0.002) with RVESV and 0.684 (p = 0.004) with RV strain; corresponding IDI values were 0.0887 (p = 0.03) and 0.1222 (p = 0.01), respectively. Conclusions: Combining CT-based fibrosis quantification with cardiac MRI-derived RV functional assessment enhances the non-invasive diagnosis of PH in ILD patients. This integrated imaging approach significantly improves diagnostic precision and may facilitate earlier, more targeted interventions in the management of ILD-associated PH.

Keywords: computed tomography; interstitial pneumonia; magnetic resonance imaging; pulmonary fibrosis; right ventricular function.

Figure 1

A case of ILD with mild pulmonary fibrosis and preserved right ventricular (RV) strain. (A) Quantitative CT-based 3D reconstruction of lung parenchymal patterns. Normal lung parenchyma (pink), emphysema (dark blue), ground-glass opacity (GGO, orange), reticulation (light blue), and honeycombing (yellow) are segmented. (B) Cardiac MRI showing right ventricular strain analysis using feature tracking. (C) Strain–time curve of the right ventricle showing preserved peak longitudinal strain (−23.0%).

Figure 2

A case of ILD with advanced pulmonary fibrosis and impaired right ventricular (RV) strain. (A) Quantitative CT-based 3D reconstruction of lung parenchymal patterns. Normal lung parenchyma (pink), emphysema (dark blue), ground-glass opacity (GGO, orange), reticulation (light blue), and honeycombing (yellow) are segmented, with extensive honeycombing and emphysema observed. (B) Cardiac MRI showing RV strain analysis with feature tracking, indicating reduced myocardial deformation predominantly in the free wall. (C) Strain–time curve of the right ventricle showing severely reduced peak longitudinal strain (−7.6%).

Figure 3

Boxplots of CT and MRI parameters comparing patients with and without pulmonary hypertension (PH). Patients with PH [PH(+), blue] demonstrated significantly impaired right ventricular (RV) function, as indicated by lower RV longitudinal strain (p = 0.026) and higher RV end-systolic volume (RVESV, p = 0.003) compared to those without PH [PH(−), red]. On CT analysis, PH(+) patients exhibited significantly greater diseased lung area (p < 0.001), ground-glass opacity (GGO, p = 0.005), and honeycombing (p = 0.0027). No significant differences were observed for emphysema (p = 0.12), reticulation (p = 0.23), or consolidation (p = 0.27).

Figure 4

Receiver operating characteristic (ROC) curves of CT and MRI parameters for the diagnosis of pulmonary hypertension (PH). ROC curves compare the diagnostic performance of cardiac MRI-derived parameters (RV strain, RV end-systolic volume [RVESV]) and CT-derived parameters (diseased lung area, ground-glass opacity [GGO], honeycombing, reticulation, consolidation, and emphysema) for detecting PH in patients with interstitial lung disease. The area under the curve (AUC) was highest for diseased lung area (AUC = 0.778), followed by RVESV (AUC = 0.754), honeycombing (AUC = 0.738), and GGO (AUC = 0.724). RV strain showed moderate diagnostic utility (AUC = 0.665), while emphysema, reticulation, and consolidation showed lower diagnostic value.

Figure 5

Quadrant plot of RV longitudinal strain and diseased lung area with pulmonary hypertension (PH) prevalence. The plot divides the cohort into four quadrants based on threshold values determined by ROC analysis using the Youden index: −12.5% for RV longitudinal strain (x-axis) and 53.6% for diseased lung area (y-axis). Each dot represents a subject, color-coded by PH status (red = PH present, blue = PH absent). Dashed lines indicate the threshold values. In each quadrant, the percentage and count of patients with PH (n/N) are shown. The prevalence of PH was highest in the upper right quadrant (85.7%, 6/7) and lowest in the lower left quadrant (7.9%, 3/38), demonstrating the combined impact of reduced RV strain and increased lung fibrosis on PH presence.