Artificial intelligence-based assessment of central aortic haemodynamics using non-invasive pulse wave analysis in constrictive pericarditis

Abstract

Background

Constrictive pericarditis (CP) is a rare but significant pericardial disease resulting in impaired ventricular filling and heart failure symptoms, often following cardiac surgery. Its clinical presentation complicates diagnosis, mimicking other causes of heart failure. Recent technological advances, including artificial intelligence (AI)-based non-invasive pulse wave analysis (AI-PWA), have the potential for improved haemodynamic assessment and clinical decision-making.

Objectives

This study evaluates the clinical utility of AI-PWA in assessing central aortic blood pressure (CABP), arterial stiffness and cardiac function in CP.

Methods

This prospective case-control study enrolled 12 adult CP patients and 12 age- and sex-matched healthy controls. CABP and peripheral blood pressure (PBP) were measured using the VascAssist2. Haemodynamic parameters, including pulse wave velocity (PWV), augmentation index@75 (AIx@75), cardiac index, stroke volume and dP/dtmax, were assessed and compared between groups.

Results

CP patients showed significantly lower mean CABP than systolic PBP (101.8±23.4 mm Hg vs 112.3±22.9 mm Hg). PWV showed elevated values (>9 m/s) in nnn (42%) of cases, indicating increased arterial stiffness (8.88±1.94 m/s). AIx@75 was higher in CP patients (22.55±8.36%) compared with controls (16.38±6.53%), reflecting increased wave reflection, increased systemic vascular resistance or enhanced aortic compliance. Cardiac performance was notably impaired in the CP group, with reduced stroke volume (64.8±18.8 mL vs 94.9±25.0 mL, p=0.003) and dP/dtmax (724.9±228.2 mm Hg/s vs 1055.3±203.2 mmHg/s, p=0.0011), indicating impaired ventricular function. The heart failure index was significantly higher in CP patients (31.8±18.3% vs . 6.4±6.5%, p<0.001), indicating substantial functional compromise.

Conclusion

AI-PWA provides clinically relevant insights into central haemodynamics and arterial stiffness in CP patients. This non-invasive approach may enhance diagnosis and management of CP and should be considered for integration into routine cardiologic evaluation protocols.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• This study is the first to use artificial intelligence-based pulse wave analysis (AI-PWA) to evaluate central aortic blood pressure and cardiac function in patients with CP. The results demonstrate that conventional PBP may miss key abnormalities, while AI-PWA reveals significant alterations in central haemodynamics, including reduced systolic pressure, increased arterial stiffness and impaired contractility. Metrics such as stroke volume, dP/dtmax and a composite heart failure score clearly differentiated CP patients from healthy controls, offering new insights into disease severity.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• These findings highlight the potential role of AI-driven PWA in improving the diagnosis and management of CP. By providing a more comprehensive, non-invasive assessment of cardiovascular function, AI-PWA may help clinicians detect functional impairment earlier and tailor treatment more precisely. If validated in larger studies, this technology could be integrated into clinical protocols for CP and possibly extended to other cardiac diseases, such as complex heart failure.

Introduction

Constrictive pericarditis (CP) remains a significant, but often under-recognised clinical entity, even in modern cardiology. CP is characterised by the fibrotic thickening and, in some cases, calcification of the pericardium, leading to severe diastolic dysfunction and heart failure. While CP it is relatively rare, its clinical importance is particularly pronounced in post-cardiac surgery patients.

Historically, CP was primarily associated with bacterial infection, such as tuberculosis, or different viral infectious diseases. However, today CP is more prevalent as a complication after cardiac surgical procedures. This shift underscores the growing challenge in the post-operative setting, where its diagnosis and management are often complex.

The prevalence of CP is difficult to quantify exactly, largely due to its often subtle and nonspecific clinical presentation. Estimates suggest a prevalence of 0.2% and 2% among patients who have undergone cardiac surgery, though these figures may vary depending on the population studied and the diagnostic criteria employed.¹ The rarity of CP, coupled with its often indolent course, contributes to underdiagnosis, further complicating efforts to determine its true prevalence.

Diagnosing CP can be particularly challenging because it may mimic other forms of diastolic or systolic heart failure or restrictive cardiomyopathy, resulting in diagnostic delays. Advanced imaging techniques, such as echocardiography, chest X-ray, cardiac MRI, CT and cardiac catheterisation, are essential tools in the detection of CP, though they may not always be conclusive. Differentiating CP from other conditions with similar haemodynamic profiles requires a high level of clinical suspicion and expertise.

Recent diagnostic advances, including the innovative, artificial intelligence (AI)-based, non-invasive pulse wave analysis (AI-PWA) of the central aortic blood pressure (CABP), may provide a deeper insight into the overall haemodynamic status and aortic pressure characteristics in patients with CP. Therefore, AI-PWA might be used to quantify disease severity more accurately.

Accordingly, the aim of our study was to use AI-PWA for assessment of the CABP profile and related parameters in patients with CP. Our goal is to provide better guidance for the follow-up and for therapeutic decisions in this demanding patient cohort.

Study cohort

This prospective case-control study includes patients seen at the Department for Cardiac Surgery at Friedrich-Alexander-University Erlangen-Nürnberg, Erlangen, Germany.

Subjects and measurements

This study enrolled patients diagnosed with proven CP before or after pericardiectomy, admitted between June 2023 and July 2024. All patients were included consecutively based on their presentation at our department, with no preselection process.

The inclusion criteria were a confirmed diagnosis of CP and age 18 years or older. Exclusion criteria included the presence of implanted cardiac devices (such as pacemakers or automatic implantable cardioverter defibrillators), lack of cognitive ability to provide informed consent, and refusal to participate in the study.

Patient medical records were reviewed for demographic information and clinical data. A comprehensive data form was used, documenting clinical diagnoses of CP, as well as anthropometric and clinical parameters, including age, sex, weight, height, body mass index (BMI), medication regimen and history of previous cardiac surgery.

Fourteen age- and sex-matched volunteers without overt cardiovascular disease were recruited as a baseline control group.

Sample size Rational

Given the rarity of CP, the sample size was determined pragmatically based on the expected number of eligible patients during the recruitment period rather than on a formal a priori power calculation. Comparable exploratory studies in CP and related haemodynamic disorders have typically enrolled between 10 and 20 participants per group.

At the time of study planning, no published data were available to reliably estimate effect sizes for AI-PWA parameters in CP.

Blood pressure measurement and PWA

Peripheral brachial and ankle blood pressures (PBPs) were assessed using the VascAssist2 device (VA) (inmediQ GmbH, Butzbach, Germany) after the patients had rested in a supine position for at least 5 min.

For analysis, brachial systolic and diastolic blood pressure (pSDBP and pDBP) values were used from the arm with the higher value.

Pneumatic brachial cuffs of adequate size were applied to both upper arm and both ankles. These cuffs were inflated to supra-systolic pressure and then gradually deflated until a flow signal was detected.

CABP was also determined by the VA using the methodology described by Halder et al.² The algorithm underlying the CABP measurement is based on a software modification of the Westerhof model,^{3 4} which was further refined by Schumacher et al in 2018, significantly improving computational accuracy by increasing the number of segments from 121 to 711.⁴

Validation of AI-PWA parameters

The VA system (inmediQ, Germany) applies the Cardiovascular Twin Model, which reconstructs central aortic pressure and related indices using non-invasive oscillometric waveforms. The algorithm has been validated in independent studies against invasive aortic pressure recordings, showing high correlations for central systolic and diastolic pressure, pulse pressure and augmentation index (Aix). Although no invasive validation was performed within the present CP cohort, the same standardised measurement and calibration protocol were applied. Therefore, the previously established accuracy is assumed to extend to this population, within the acknowledged limitations of disease-specific haemodynamic alterations.

For blood pressure assessment, appropriately sized pneumatic cuffs, ranging from 21 to 29 cm and 28–37 cm, were applied to both arms. A smaller cuff of 16–22 cm was used for wrist measurement. After a fully automated bilateral cuff calibration and a 2-min rest, brachial blood pressure was measured on both sides. This was followed by radial blood pressure and pulse contour recordings, taken in triplicate with 30-s intervals between measurements. The entire examination took approximately 25 min.

CABP was calculated offline using proprietary software (VascViewer 2, 3.0, inmediQ GmbH, Butzbach, Germany) based on Cardiovascular Twin Technology.^{4 5}

Overall, the following parameters were evaluated: pSBP and pDBP, pulse pressure (pPP), central arterial blood pressure (cSBP, cDBP, cPP), arterial stiffness indices (Aix, Augmentation Index@75 (Aix@75), aortic pulse wave velocity (PWV) and haemodynamic parameters (cardiac index (CI); stroke volume (SV), ejection duration (ED, left ventricular ejection time (LVET)), dp/dt max radial, left ventricular plateau time (LVPT), Dicrotic ratio, heart failure score).

Pulse wave analysis (PWA)

Figure 1 explains in detail parameters, measurement and calculations of PWA.

Figure 1. Pulse wave recording: The method first measures peripheral pressure waveforms and then accurately calculates central pressure waves. After the start of the pulse wave, T0 indicates the onset of ejection. The pressure wave then rises to an initial systolic peak (P1). The second peak (P2) represents the peak of the arterial reflection wave. The end of ejection is the point of closure of the aortic valve and the end of systole. Augmentation pressure (AP) is the contribution of the reflected wave to the systolic arterial pressure and is calculated as the difference (mm Hg) between P2 and P1. The augmentation index (AIx) is calculated as the difference between the peak systolic blood pressure (at P2) and the peak pressure at the inflection point (at P1), divided by the central pulse pressure (PP). It is expressed as a percentage of the central pulse pressure. Pb, backward or reflection wave; Pf, forward or ejection wave.

Control for confounding Factors

Demographic characteristics, comorbidities and medication regimens were collected for all participants. To reduce confounding, patients and controls were matched 1:1 for age and sex, and baseline characteristics were compared using appropriate statistical tests (table 1). Because of the limited sample size, formal multivariate regression modelling was not performed to avoid model overfitting.

Table 1. Baseline characteristics of the one-to-one matched study participants.

Parameter	Overall (n=24)	Cases (n=12)	Control (n=12)	p-Value
Age (years)	57.96±13.17 (36–71)	58.00±13.48 (36–71)	57.92±13.46 (36–71)	1.0
Sex (f:m)	2 (8.0%):22 (92%)	1 (8.3%):11 (91.2%)	1 (8.3%):11 (91.2%)	1.0
Height (cm)	179.71±10.42 (159–198)	181.83±10.41 (165–195)	177.58±10.44 (159–198)	0.3288
Weight (kg)	83.52±15.75 (59–112)	88.64±14.37 (63–104.9)	78.40±15.97 (59–112)	0.1127
BMI (kg/m2)	25.89±4.89 (21.80–38.75)	26.91±4.71 (21.80–34.67)	24.88±5.05 (21.94–38.75)	0.3122
BSA (m2)	2.03±0.22 (1.61–2.34)	2.11±0.20 (1.72–2.34)	1.96±0.22 (1.61–2.30)	0.0971

BMI, body mass index; BSA, body surface area; f, female; m, male.

Statistical analysis

The data analysis was performed using Python (Python 3.11.5). All statistical evaluations of the data were pseudonymised and not person related.

Quantitative variables were described using the mean, median and range (minimum and maximum). To compare the means of independent variables, the student’s t-test and the Mann-Whitney U test were applied.

The distribution of continuous variables was additionally illustrated using boxplots, from which the quartiles (25th, 50th and 75th percentiles) and the range can be derived. Outliers were defined as values lying more than 1.5 to 3 times the IQR above the upper or below the lower quartile and are indicated by ‘o’. Extreme values, exceeding three times the IQR from the upper or lower quartile, are marked with ‘*’.

Receiver operating characteristic (ROC) analyses were performed to evaluate predictive parameters. The area under the ROC curve (AUC) was used as a measure of diagnostic accuracy. AUC values greater than 0.5 indicate a non-random explanatory capacity of the respective quantitative variable; values closer to the theoretical maximum of 1 reflect better overall predictive performance. A clinically relevant threshold for sensitivity was defined as ≥80%.

All statistical tests were conducted at a 5% significance level.

Study sample and demographic characteristics

Enrolled in this study were 24 participants, 12 patients and 12 controls (online supplemental table 1).

A total of 12 patients had the confirmed diagnosis of CP, either prior to or following pericardiectomy.

The underlying aetiology of CP was idiopathic in six patients, postoperative following heart transplantation in one patient, post-perimyocarditis in two patients and postradiation therapy in three patients (radiation only, 1; radiation+cardiac surgery, 1; radiation+thoracic surgery: 1).

Among the 12 participants, four had previously undergone surgical pericardiectomy. In these cases, persistent or recurrent constriction may have resulted from incomplete pericardial resection, fibrotic scarring, underlying diastolic dysfunction or extension of pericardial calcification into the myocardial tissue.

To assess central haemodynamic parameters and vascular stiffness derived from AI-PWA, a comparative evaluation was conducted using a control cohort matched for age and anthropometric characteristics.

The demographic data of patients and controls are given in table 1.

Sex distribution was balanced between the case and control groups. Among the 24 participants, two were female (8.0%), and 22 were male (92%). Each group (n=12) included one female and eleven male participants. Using Fisher’s exact Test, there was no statistically significant (p=1.0) difference in sex distribution between the two groups.

The demographic characteristics of the study population are presented in table 1.

The mean age in both groups was almost identical (58.0±13.48 vs 57.92±13.46 years; p=1.0), with no statistically significant differences in height, weight, BMI or body surface area.

Blood pressure and AI-PWA: PBP and CABP

Results of AI-PWA

Using AI-PWA, the central (aortic) systolic blood pressure in CP was, consistent with physiological expectations, lower than the corresponding PBP values in all patients, with a mean of 101.83±23.35 mm Hg (range: 78.0–147.0 mm Hg). In nine out of 12 cases, central systolic pressure was hypotensive (<110 mm Hg), while it exceeded the normative threshold in only two cases. The mean central diastolic aortic blood pressure in CP was 68.92±16.10 mm Hg (range: 45.0–102.0 mm Hg).

The calculated aortic pulse pressure was significantly reduced in the CP patients (32.58±12.09 mmHg vs 42.33±5.30 mm Hg; p=0.0301) (table 2, figure 2).

Table 2. Peripheral (brachial), central (aortic) blood pressure (BP) values and AI-PW parameters.

Parameter	Overall (n=24)	Cases (n=12)	Control (n=12)	p-Value
Heart rate (beats per minute)	75.58±14.45 (55.0–111.0)	84.33±12.84 (66.0–111.0)	66.83±10.23 (55.0–86.0)	0.0013**
Brachial BP systolic, arm (mm Hg)	119.33±18.52 (90.0–150.0)	112.25±22.87 (90.0–150.0)	126.42±9.18 (108.0–139.0)	0.1119
Brachial BP diastolic, arm (mm Hg)	69.50±12.98 (44.0–100.0)	68.83±16.15 (44.0–100.0)	70.17±9.52 (52.0–83.0)	0.8077
Brachial pulse pressure, arm (SBP-DBP) (mm Hg)	49.83±12.63 (27.0–72.0)	43.42±13.73 (27.0–66.0)	56.25±7.45 (42.0–72.0)	0.0112*
Brachial side difference arms (mm Hg)	7.46±7.13 (1.0–31.0)	10.17±9.01 (1.0–31.0)	4.75±3.02 (2.0–12.0)	0.0626
MAP, brachial	84.11±14.06 (60.4–117.7)	82.55±18.26 (60.4–117.7)	85.67±8.63 (71.0–98.0)	0.5984
Central/aortic BP systolic, (mm Hg)	107.67±18.16 (78.0–147.0)	101.83±23.35 (78.0–147.0)	113.50±8.37 (96.0–125.0)	0.1259
Central/aortic BP diastolic, (mm Hg)	69.50±13.16 (45.0–102.0)	68.92±16.10 (45.0–102.0)	70.08±10.11 (54.0–84.0)	0.8337
Aortic pulse pressure	37.46±10.40 (22.0–57.0)	32.58±12.09 (22.0–57.0)	42.33±5.30 (31.0–51.0)	0.0301*
Aortic MAP	82.22±14.14 (56.0–115.0)	79.89±17.94 (56.0–115.0)	84.56±9.17 (68.0–97.67)	0.4309
Aortic PWV	8.40±1.70 (4.9–13.0)	8.88±1.94 (4.9–13.0)	7.92±1.33 (5.8–9.8)	0.1755
Aix	19.30±9.32 (1.8–38.6)	18.25±12.27 (1.8–38.6)	20.35±5.37 (13.0–33.9)	0.4187
Aix@75 (normalised at 75 beats per minute)	19.46±7.99 (5.8–36.6)	22.55±8.36 (9.5–36.6)	16.38±6.53 (5.8–30.8)	0.0561

* p < 0.05; ** p < 0.01; *** p < 0.001

AI-PW, artificial intelligence-based pulse wave ; Aix, augmentation index; DBP, diastolic blood pressure; MAP, mean arterial pressure; PWV, pulse wave velocity; SBP, systolic blood pressure.

Figure 2. Boxplots illustrate group differences between patients with constrictive pericarditis (CP) and healthy controls. The central line represents the median; box edges correspond to the 25th (Q1) and 75th (Q3) percentiles (IQR). Whiskers extend to the most extreme values within 1.5× IQR from the quartiles. Data points beyond this range are shown as outliers. Asterisks indicate statistically significant differences.

PBP in CP, assessed at the upper extremity with the higher reading, showed a mean systolic pressure of 112.25±22.87 mm Hg (range: 90.0–150.0 mm Hg). Notably, seven out of 12 patients were hypotensive (systolic pressure <110 mmHg).

There was no statistically significant difference in brachial systolic blood pressure between the CP cohort (112.25±22.87 mm Hg) and the control group (126.42±9.18 mmHg; p=0.1119).

The brachial pulse pressure, which is proportional to the SV and inversely related to aortic compliance, averaged 43.42±13.73 mm Hg (range: 27.0–66.0 mm Hg) (p=0.0112*) (table 2). It was reduced in patients and elevated in 4.

However, brachial pulse pressure was significantly lower in the patient’s group (43.42±13.73 mm Hg vs 56.25±7.45 mmHg; p=0.0112).

AI-PWV, as a gold-standard marker of arterial stiffness, with a normal range of 7 to 9 m/sec, was measured as an indicator of aortic stiffness, with lower values indicating greater vessel elasticity.

PWV was elevated in the patient group compared with the control group (8.88±1.94 m/s vs 7.92±1.33 m/s), although this difference did not reach statistical significance (p=0.1755) (table 2).

In CP, values above 9 m/sec were found in five of 12 patients, with values ranging from 9.1 to 13 m/sec, suggestive of increased arterial stiffness. Of these patients, 4 had a status post pericardectomy as the final procedure.

A low PWV—interpreted as indicative of a low CI or reduced myocardial performance—was observed in only one case.

The Aix adjusted to a heart rate of 75 /min (AIx_ao75), as an indirect measure of aortic stiffness, reflective of the strain imposed on the heart by stiff arteries, averaged in CP 22.55±8.36 (range: 9.50–36.60). Elevated values (above normal reference) were present in six out of 12 patients, while patients exhibited subnormal values (<15).

AIx@75 failed to achieve statistical significance. Nevertheless, there is a trend towards elevated AIx@75 in the patient group (22.55±8.36% vs 16.38±6.53%; p=0.0561).

Notably, the heart rate was significantly elevated in the CP cohort (84.33±12.84 beats per minute vs 66.83±10.23 beats per minute; p=0.0013*) (figure 2).

Cardiac function Parameters in PWA

Indices of myocardial contractility differed significantly between the case and control groups and were markedly impaired in CP, indicative of impaired myocardial performance.

Both CI and SV were lower in the patient group (SV (64.83±18.83 mL vs 94.92±25.01 mL, p=0.003); CI (2.68±0.72 vs 3.46±1.21 l/min/m², p=0.072)).

A reduced CI (<2.0 L/min/m²), consistent with impaired cardiac output, was identified in CP patients.

The SV (<70 mL in men, <60 mL in women) was diminished in nine out of 12 patients. SV exhibited a distinct separation between groups (AUC 0.85), underscoring its relevance in assessing ventricular function (table 3).

Table 3. Haemodynamic profiles in study versus control subjects.

Parameter	Overall (n=24)	Cases (n=12)	Control (n=12)	p Value
dp/dt max (mm Hg/s)	890.08±270.40 (329.0–1410.0)	724.92±228.16 (329.0–1085.0)	1055.25±203.22 (684.0–1410.0)	0.0011*
Dicrotic Notch ratio (%)	8.28±7.96 (0.0–25.4)	11.62±9.84 (0.0–25.4)	4.94±3.37 (1.3–13.0)	0.0437*
SV (ml)	79.88±26.55 (34.0–136.0)	64.83±18.83 (34.0–105.0)	94.92±25.01 (65.0–136.0)	0.003**
CO (l/min)	6.14±1.91 (3.5–10.3)	5.62±1.53 (3.5–8.1)	6.67±2.17 (4.1–10.3)	0.1844
CI (l/min/m2)	3.07±1.05 (1.53–5.46)	2.68±0.72 (1.53–3.81)	3.46±1.21 (1.81–5.46)	0.072
LVPT (ms)	49.00±38.75 (4.0–143.0)	65.58±47.68 (4.0–143.0)	32.42±16.33 (15.0–77.0)	0.1188
LVET (ms)	245.92±33.66 (188.0–316.0)	244.67±43.59 (188.0–316.0)	247.17±21.57 (217.0–289.0)	0.8603
HI-Score (%)	19.12±18.69 (0.0–72.0)	31.83±18.34 (11.0–72.0)	6.42±6.46 (0.0–18.0)	0.0003***

* p < 0.05; ** p < 0.01; *** p < 0.001.

CI, cardiac index; CO, cardiac output; HI-Score, heart insufficiency score; LVET, left ventricular ejection time; LVPT, left ventricular plateau time; N/n, absolute number; SV, stroke volume.

The contractility Parameter dp/dt_max, a marker for global ventricular function, was <800 mm Hg/s in seven cases, reflecting compromised contractile reserve. dP/dt_max was significantly reduced in CP patients compared with the controls (724.92±228.16 mm Hg/s vs 1055.25±203.22 mm Hg/s; p=0.0011*). This finding aligns with the strong diagnostic performance observed in the corresponding ROC analysis (AUC 0.86) (figure 3).

Figure 3. The receiver operating characteristic (ROC) curve depicts the relationship between sensitivity (true-positive rate) and 1 − specificity (false-positive rate) across varying classification thresholds. The area under the curve (AUC) provides a summary measure of classification performance, with an AUC of 1.0 indicating perfect discrimination and 0.5 representing chance-level performance.

Shortened ED (< 270 msec), suggestive of reduced systolic performance or tachycardia, was observed in patients.

The duration of the LVPT, indicative of systolic function and afterload, was not shortened in any case, but was prolonged in patients, suggesting increased afterload or ventricular-arterial mismatch (table 3).

The dicrotic ratio, reflecting both the timing and magnitude of reflected pressure waves, was significantly elevated in the CP group (11.62±9.84 vs 4.94±3.37; p=0.0437*). This may signal pathological aortic compliance, contributing to increased afterload and inefficient ventricular-arterial coupling.

Notably, the heart failure index (heart insufficiency score (HI-Score)) was significantly elevated in the study cohort (31.83±18.34 vs 6.42±6.46; p=0.0003*), underscoring impaired cardiovascular reserve and subclinical myocardial dysfunction in this population.

Discussion

Emerging technologies such as AI-PWA may complement traditional diagnostic approaches by enabling more nuanced detection of subtle changes in cardiovascular dynamics.

To our knowledge, this is the first comprehensive study using a refined, sophisticated PWA into the clinical diagnosis and follow-up of patients with CP.

This study employs AI-PWA, which represents a novel approach differing from conventional PWA methods.

These measurements are based on the patented Cardiovascular Twin Technology by Schumacher et al.⁴ This method advances a 1969 virtual simulation model by Westerhof and Noordergraaf,³ which divided the arterial tree into 121 segments. Using AI, this model is now expanded to 711 segments, creating a non-invasive, patient-specific arterial profile.

This tool holds promise not only for initial diagnosis but also for longitudinal monitoring of therapeutic response in patients with CP.

AI-PWA provides a novel, non-invasive method for assessing central haemodynamics. The inclusion of dP/dt as a parameter offers valuable insight into cardiac contractility, serving as an important marker for evaluating myocardial function in patients with CP. Even in modern cardiology, CP remains a significant, yet often under-recognised, clinical entity, which can lead to severe diastolic dysfunction of the right and/or left ventricle and heart failure. CP is an uncommon and pathophysiologically heterogeneous disorder characterised by chronic fibrotic remodelling of the pericardium that impairs ventricular filling and ultimately compromises cardiac output. Despite advances in diagnostic imaging and therapeutics, the timely and accurate diagnosis of CP remains a significant clinical challenge, particularly in complex or atypical cases.

In high-income countries, the predominant aetiologies have shifted from infectious causes to idiopathic origins and iatrogenic factors, most commonly prior cardiac surgery, thoracic irradiation or pericardial injury following interventional procedures.^{1 6} Estimates suggest that the incidence ranges between 0.2% and 2% among patients who have undergone cardiac surgery, though these figures may vary depending on the population studied and the diagnostic criteria employed.¹

In contrast, historically, tuberculosis was the leading cause. However, widespread implementation of effective antituberculous therapy has markedly reduced its incidence in developed regions. However, mycobacterium tuberculosis remains a leading cause of CP in low- and middle-income regions, where access to early diagnosis and effective anti-tuberculous therapy may be limited.⁷

Its pathophysiology is defined by dense fibrotic thickening, calcification and adhesions between the visceral and parietal pericardial layers. These alterations lead to a non-compliant pericardium that restricts diastolic ventricular filling, producing a characteristic haemodynamic profile.⁸

Clinical presentation of CP consists mainly of signs and symptoms of right heart failure. In more advanced stages, a reduction in cardiac output may also be present.

The exact diagnosis is pivotal yet remains challenging due to symptom overlap with restrictive cardiomyopathy and other causes of heart failure with preserved ejection fraction.

The diagnostic process is multimodal. The diagnosis is usually suspected clinically and confirmed by echocardiography, chest X-ray, MRI, CT and finally a cardiac catheterisation.9,11 In equivocal cases, invasive haemodynamic assessment via cardiac catheterisation remains the gold standard, offering definitive documentation of hallmark findings such as elevated and equalised diastolic pressures in both ventricles, a prominent ‘dip-and-plateau’ (or ‘square root’) sign and an M- or W-shaped right atrial pressure tracing.¹² Although SV is typically reduced, cardiac output may be maintained at rest via compensatory tachycardia until late in the disease course.

Surgical pericardiectomy remains the only definitive treatment. A complete pericardiectomy, defined as anterior and posterior pericardial resection from phrenic to phrenic nerve including the diaphragmatic surface, is associated with superior long-term survival, symptom relief and functional recovery compared with partial resection.^{13 14}

The diagnosis of recurrent or persistent constriction following partial pericardiectomy remains challenging, particularly in patients after partial pericardiectomy. In these cases, residual non-constricted ventricular segments may attenuate classic Doppler echocardiographic findings.¹⁵ Moreover, aggressive diuretic therapy may offer transient symptomatic relief but often masks clinical signs of elevated filling pressures and delays definitive diagnosis.

This study was performed to better guide diagnostic and therapeutic decisions in patients with CP by assessing central aortic haemodynamics before or after cardiac surgery using AI-PWA.

Our findings demonstrate a distinct haemodynamic profile in patients with CP, as assessed by AI-PWA. CABP was consistently lower than peripheral readings, with values falling into the hypotensive range in the majority of patients. This pattern aligns with the expected pathophysiology of CP, where impaired ventricular filling limits effective SV generation. The observed reduction in central pulse pressure further supports this notion and may indicate compromised myocardial performance or increased aortic compliance in the setting of pericardial constraint.

Although peripheral (brachial) SBP did not differ significantly from controls, a trend towards hypotension was evident, with more than half of the cohort presenting with values below 110 mm Hg. Moreover, the significantly lower brachial pulse pressure in CP patients points towards reduced SV, reinforcing the hypothesis of limited diastolic filling due to pericardial constriction.

The analysis of PWV, a surrogate marker for arterial stiffness, revealed higher average values in CP compared with controls, although statistical significance was not reached. Nonetheless, individual elevated PWV values exceeding 9 m/sec in a substantial subset of patients, particularly those post-pericardectomy, may reflect either intrinsic aortic stiffening or altered ventricular-vascular coupling. Elevated sympathetic tone or systemic vascular resistance, as compensatory responses to limited cardiac output, may also contribute to this finding.

Interestingly, the AIx@75, while not significantly different between groups, showed a trend towards elevation. This may indicate increased wave reflection or higher vascular load in CP, again suggesting impaired ventricular-arterial interaction. Conversely, subnormal AIx values in some patients might reflect preserved vascular compliance or reduced wave reflection in early or compensated stages.

Moreover, the significantly elevated heart rate observed in the CP cohort likely represents a compensatory tachycardia, potentially driven by sympathetic overactivation to maintain cardiac output under conditions of restricted preload.

Beyond vascular dynamics, AI-PWA also enabled detailed evaluation of myocardial performance in CP. Indices of cardiac function were consistently impaired in the CP cohort. SV was significantly reduced in the majority of patients, with values falling below established sex-specific thresholds in nine out of 12 cases. The marked separation of SV between patients and controls, reflected by an AUC of 0.85, underscores its diagnostic relevance in assessing functional impairment due to pericardial constriction.

Similarly, the CI showed a trend towards reduction, with values <2.0 L/min/m² in one-third of the CP group, which is suggestive of compromised forward cardiac output despite often preserved ejection fraction. These findings are consistent with the known haemodynamic hallmark of CP, where the systolic function is preserved at rest but impaired during diastolic filling and reduced SV.

Contractility, as measured by dP/dtmax, a surrogate for global left ventricular systolic performance, was significantly impaired in CP patients. Values <800 mm Hg/s were observed in over half the cohort, and the mean dP/dtmax was markedly lower than in controls. The high diagnostic accuracy (AUC 0.86) supports its potential role as a non-invasive contractility marker in CP. Additionally, shortened ED in most patients may reflect reduced systolic performance or compensatory tachycardia, while prolongation of the LVPT in some cases might indicate elevated afterload or disrupted ventricular-arterial interaction.

Further, the significantly increased dicrotic ratio observed in CP suggests altered pressure wave reflection and pathological aortic compliance. This, in turn, contributes to the suboptimal ventricular-arterial coupling and increased afterload. Perhaps most notably, the HI-Score, a composite marker of cardiovascular stress and reserve, was substantially elevated in the CP cohort, reinforcing the presence of subclinical myocardial dysfunction even in the absence of overt systolic heart failure.

Altogether, AI-based PWA reveals a coherent pathophysiological profile of PC characterised by reduced pulse pressures, impaired contractility, vascular stiffening and compensatory tachycardia. By capturing both central haemodynamics and myocardial performance, AI-PWA offers a non-invasive, multidimensional tool for diagnosis, functional assessment, and monitoring, especially when conventional imaging is inconclusive.

The results of the study emphasise the need for further research into the value of modern AI-PWA in determining haemodynamics in patients with CP, as this innovative method shows promise in enhancing the detection and monitoring of CP. Moreover, it could facilitate earlier diagnosis, risk stratification and personalised surgical planning.

Limitations

Several factors may limit the interpretation and generalisability of the present findings. The study cohort exhibited a heterogeneous mix of CP aetiologies, which may influence the generalisability of findings.

Moreover, the use of a single-centre, tertiary care setting may introduce a selection bias towards patients with more complex disease profiles, limiting the applicability of results to broader, community-based populations.

Also, the limited sample size restricts the statistical power of the study and may hinder the detection of subtle yet clinically relevant associations. However, CP remains a rare condition, and the study was designed as an exploratory, single-centre feasibility investigation. Confirmation in larger, multicentre and longitudinal cohorts will be essential to establish robust reference values and validate diagnostic cut-offs.

While AI-PWA offers a novel, non-invasive approach to assessing central haemodynamics, it relies on proprietary algorithms and theoretical cardiovascular models. This introduces a degree of uncertainty in interpreting the haemodynamic indices obtained.

The accuracy of derived parameters, especially those reflecting myocardial contractility, has not yet been thoroughly validated against invasive gold-standard techniques in the CP population. Regarding Clinical Threshold Interpretation, the present AI-PWA-derived indices should be interpreted cautiously given the limited sample size and absence of external validation.

At this stage, the data support the conceptual feasibility of using AI-PWA to capture the characteristic haemodynamic signature of constrictive physiology rather than providing actionable diagnostic cut-offs. Future multicentre studies with larger cohorts should define robust, disease-specific reference ranges and validate clinically meaningful thresholds for AI-PWA parameters.

Therefore, a further limitation of the study is the absence of direct invasive reference measurements in our CP cohort. While this was not ethically or practically feasible, the AI-PWA parameters used here have been validated against invasive aortic pressure recordings in prior studies, demonstrating strong agreement across diverse cardiovascular conditions.

Given that the same algorithm and calibration procedure were applied, these validation results can be reasonably extrapolated to CP, although disease-specific pathophysiology may influence individual waveform morphology.

Future studies integrating simultaneous invasive and non-invasive recordings in CP patients are warranted to further substantiate the accuracy of AI-PWA in this specific setting.

Despite matching controls for age and anthropometric characteristics, the potential for residual confounding due to unmeasured variables such as comorbidities, pharmacologic treatment or lifestyle factors remains. Furthermore, the reference values used for key metrics like PWV and AIx@75 are based on general population data and may not adequately reflect the pathophysiological nuances of patients with CP.

The cross-sectional design of the study inherently limits causal inferences and precludes any evaluation of longitudinal changes in cardiovascular function. To fully establish the diagnostic and prognostic utility of AI-PWA in CP, future studies should aim to incorporate larger patient cohorts, multicentre data and long-term follow-up.

Finally, the presented data are exclusively from patients living in Germany. Generalising the conclusions and applying them to patients living in other countries or cultural contexts is questionable.

In conclusion

In con his study introduces AI-PWA for the non-invasive assessment of central haemodynamics in patients with CP. The study results demonstrate the potential of this advanced technology to augment traditional diagnostic modalities and improve the monitoring of therapeutic response. Using a patient-specific arterial model based on Cardiovascular Twin Technology, AI-PWA enables detailed insight into cardiovascular function that may be particularly valuable in complex or equivocal CP cases.

Given the persistent diagnostic challenges and clinical heterogeneity of CP, AI-PWA represents a promising tool for earlier identification, individualised risk assessment and potentially more precise surgical planning.

To validate and expand on these preliminary results, future studies should involve larger patient populations and include longitudinal follow-up.

If confirmed, AI-PWA could emerge as a valuable addition to the diagnostic and prognostic toolkit for CP and potentially other forms of heart failure.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.