Chronic heart failure (CHF) patients often present with heterogeneous patterns of cardiac dyssynchrony. Although QRS prolongation (>150 ms) and left bundle branch block (LBBB) are classical markers of electrical dyssynchrony, their direct association with mechanical dyssynchrony remains controversial. This study aimed to identify key determinants of left ventricular (LV) synchrony using machine learning and explainable artificial intelligence techniques. A cohort of 412 CHF patients was stratified using integrated echocardiographic and electrocardiographic criteria: synchronous group (SG, n = 239) with standard deviation of time to peak longitudinal strain in 18 LV segments (SD18STE) ≤ 33 ms, QRS duration ≤ 120 ms, and left ventricular end-diastolic volume (LVEDV) ≤ 150 mL; asynchronous group (AG, n = 173) with SD18STE > 33 gt; 33 ms typically accompanied by QRS duration > 120 gt; 120 ms and/or LVEDV > 150 gt; 150 mL. All patients underwent speckle tracking echocardiography (STE) to assess LV and left atrial function, along with electrocardiographic evaluation of QRS duration. Key parameters included electrical dyssynchrony markers (QRS duration, LBBB status) and mechanical dyssynchrony markers (LV ejection fraction [EF], end-diastolic volume [EDV], end-systolic volume [ESV], and segmental strain timing). A random forest model was used to identify predictors of mechanical dyssynchrony, and SHapley Additive exPlanations (SHAP) analysis was employed to quantify feature contributions. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, and F1-score. Compared with the SG group, AG patients had significantly higher QRS duration (132 vs. 116 ms, p < 0.001), B-type natriuretic peptide (BNP) levels (2,520 vs. 1,820 pg/mL, p = 0.034), EDV (211 vs. 142 mL, p < 0.001), and ESV (171 vs. 97 mL, p < 0.001), as well as lower EF (21% vs. 34%, p < 0.001). Machine learning identified EF, ESV, and posterior wall segments as the primary predictors of dyssynchrony. SHAP analysis revealed that EF < 40% and ESV > 100 gt; 100 mL increased the probability of dyssynchrony. Posterior wall delays were strongly associated with dyssynchrony. LBBB presence increased the likelihood of dyssynchrony 3-fold. The model demonstrated excellent performance (AUC = 0.925, accuracy = 85.5%, F1-score = 0.878), outperforming traditional dyssynchrony indices. Mechanical dyssynchrony indicators such as EF, ESV, and EDV are superior to electrical markers in predicting LV synchrony. Dysfunction in posterior wall segments significantly contributes to mechanical asynchrony. These findings provide new insights into CHF pathophysiology and support the use of personalized criteria for cardiac resynchronization therapy candidate selection.
Cardiac dyssynchrony refers to the lack of coordinated contraction either within the ventricles (intraventricular) or between them (interventricular). It mainly manifests as electrical dyssynchrony and mechanical dyssynchrony. When different cardiac regions contract at different times, the pumping efficiency of the heart is significantly reduced, worsening symptoms and negatively affecting quality of life and prognosis. Electrical and mechanical abnormalities interact with and reinforce each other, forming a pathological cycle. Currently, cardiac mechanical dyssynchrony is assessed through regional contraction delays, discrepancies in time to peak contraction, and increased ESV/EDV ratios. ESV, an essential measure of contractile efficiency and ejection capacity, is particularly valuable for evaluating mechanical dyssynchrony. EF, a fundamental indicator of systolic function, decreases with impaired myocardial contraction and ventricular wall motion coordination. Elevated BNP levels reflect increased cardiac load and dysfunction, correlating positively with the degree of mechanical dyssynchrony. Electrical dyssynchrony is assessed by QRS duration, presence of conduction blocks, and electrocardiographic patterns. The relationship between electrical and mechanical dyssynchrony is complex. Electrical conduction delays may lead to mechanical contraction abnormalities, while mechanical dysfunction may further exacerbate electrical instability, perpetuating a vicious cycle. Compared to electrical dyssynchrony, structural mechanical dyssynchrony more significantly impacts cardiac function, especially when systolic indicators are impaired. Among various measures of mechanical dyssynchrony, the time-to-peak systolic contraction of regional myocardial segments is considered a crucial parameter. Studies have shown that different cardiac regions contribute differently to overall synchrony. In particular, posterior wall segments—especially the basal, mid, and apical posterior regions—play a key role in the development of dyssynchrony. The unique myocardial fiber orientation of the posterior basal segment results in distinct mechanical stress during contraction, making its time-to-peak contraction a critical determinant of global mechanical synchrony. This study aimed to investigate the mechanisms of cardiac dyssynchrony, especially the interaction between electrical and mechanical components. It further evaluated the effects of clinical parameters (eg, ESV, EF, BNP) on cardiac synchrony and assessed the contribution of specific regions, particularly posterior wall segments, to global function. These findings may provide more accurate and individualized indicators for predicting and assessing dyssynchrony.
Methods
Study population
A total of 412 patients diagnosed with chronic heart failure between October 2021 and March 2023 at the First Affiliated Hospital of Xinjiang Medical University were enrolled. Their ages ranged from 43 to 76 years (mean 55 ± 11 years). All patients were in sinus rhythm.
Based on combined echocardiographic and clinical criteria, patients were stratified into 2 groups: The synchronous group (SG, n = 239) with standard deviation of time to peak systolic velocity in 18 left ventricular segments (SD18STE) ≤ 33 ms was defined by concordant electrical and mechanical parameters, specifically QRS duration ≤ 120 ms and/or left ventricular end-diastolic volume (LVEDV) ≤ 150 mL. The asynchronous group (AG, n = 173) with SD18STE > 33 gt; 33 ms was characterized by prolonged QRS duration (> 120 gt; 120 ms) and/or increased LVEDV (> 150 gt; 150 mL), which typically coincided with reduced ejection fraction (EF).
This study was approved by the hospital’s Ethics Committee, and written informed consent was obtained from all participants.
Inclusion criteria: (1) All patients met the guidelines for cardiac resynchronization therapy. (2) Diagnosed with cardiac insufficiency. (3) New York Heart Association (NYHA) functional class II or higher.
Exclusion criteria: (1) Clinical conditions: Acute heart failure; systemic infectious diseases, infective endocarditis, or sepsis; hemorrhagic diseases or bleeding tendency; severe hepatic or renal insufficiency; malignant tumors; history within the previous 3 months of unstable angina pectoris, acute myocardial infarction, coronary artery bypass grafting, percutaneous coronary intervention, or cerebrovascular accident. (2) Medication: Patients taking medications known to affect QRS complex duration. (3) Technical limitations: Poor image quality or inability to visualize the endocardium clearly, precluding speckle-tracking strain analysis.
Echocardiographic image acquisition and data measurement
Echocardiographic images were acquired using a Philips CVx color ultrasound diagnostic system equipped with the proprietary QLab quantitative analysis software. By delineating regions of interest (ROIs), myocardial motion curves were obtained for the basal and mid segments of the 6 walls of the left ventricle (LV) (totaling 18 segments). For both the LV synchrony and dyssynchrony groups, the time from QRS onset to peak systolic velocity (Ts) was measured for each segment. Subsequently, the mean Ts value across all 18 LV segments (Ts-18-LV) and its standard deviation (SD18STE) were calculated.
Statistical methods
All data in this study were non-normally distributed. While most ensemble learning algorithms offer excellent predictive performance, they often lack interpretability and struggle to identify key factors influencing disease progression, which is crucial for assisting physicians in diagnosis and treatment. Therefore, this study employed machine learning based on clinical data using a random forest model combined with SHapley Additive exPlanations (SHAP) analysis. Data visualization techniques were utilized to illustrate the risk factors associated with cardiac dyssynchrony and their impact on the prediction outcomes.
In addition to the overall analysis of factors influencing the prediction of cardiac dyssynchrony, this study also conducted individual sample analysis for the prediction of cardiac dyssynchrony risk. This dual approach aimed to comprehensively analyze the influence of various factors on cardiac dyssynchrony, spanning from the population level to the individual level. Finally, the performance of the random forest model was evaluated using the following metrics: accuracy, precision, recall, F1-score, and the area under the receiver operating characteristic curve (AUC).
Results
Comparative analysis of synchronized (SG) and asynchronized (AG) baseline data
Among the 412 chronic heart failure patients involved in the study, 218 (52.9%) had known underlying etiologies, while the remaining 194 (47.1%) had unknown etiologies. Of the 218 patients with underlying diseases, 47 (11.4%) were diagnosed with coronary heart disease, and 5 (1.2%) were diagnosed with other types of cardiomyopathy.
Of the 412 chronic heart failure patients included, 239 were in the SG and 173 in the AG. The comparison between the 2 groups showed no statistically significant differences in age, body surface area, and body mass index (p > 0.05, see Table 1 ). In the analysis of electrical dyssynchrony indices, the QRS complex and LBBB demonstrated statistically significant differences (p < 0.05). Among the mechanical dyssynchrony parameters, excluding B-type natriuretic peptide (BNP), the end diastolic volume (EDV), end-systolic volume (ESV), and ejection fraction (EF) exhibited statistically significant variations (p < 0.05). Compared with the synchronized group, the asynchronous group showed significant increases in QRS (median 132 vs. 116 ms, difference of 16 ms), BNP (median 2,520 vs. 1,820 pg/mL, difference of 700 pg/mL), EDV (median 211 vs. 142 mL, difference of 69 mL), and ESV (median 171 vs. 97 mL, difference of 74 mL), as well as a decrease in EF (median 21% vs. 34%, decrease of 13%), with all differences being statistically significant (p < 0.05, see Table 1 ). The prevalence of LBBB was significantly higher in the AG group (15.03%) compared with the SG group (3.35%), a difference of 11.68%.
Table 1
Comparison of baseline data between synchronized and unsynchronized groups.
| Variable | Total (n = 412) | Synchronous Group (n = 239) | Asynchronous Group(n = 173) | Z/χ² | P |
|---|---|---|---|---|---|
| Age, years | 60 (53-69) | 60 (54-67) | 60 (52-69) | -0.31 | 0.759 |
| Body surface area, m² | 1.85 (1.71, 1.96) | 1.85 (1.71, 1.95) | 1.85 (1.71, 1.98) | -0.69 | 0.488 |
| Body mass index, kg/m² | 25.4 (23.0-28.0) | 25.5 (23.4-28.0) | 25.0 (23.0-28.0) | -0.73 | 0.466 |
| Heart rate, bpm | 76 (67-86) | 76 (65-87) | 77 (68-86) | -0.65 | 0.514 |
| QRS duration, ms | 122 (103-142) | 116 (100-133) | 132 (111-153) | -5.41 | <0.001 |
| BNP, pg/mL | 2,210 (601-4,900) | 1,820 (518-4,045) | 2,520 (859-5,290)0) | -2.11 | 0.034 |
| EDV, mL | 167 (120-232) | 142 (112-189) | 211 (157-284) | -7.69 | <0.001 |
| ESV, mL | 119 (77-181) | 97 (63-134) | 171 (110-229) | -9.02 | <0.001 |
| EF, % | 28 (20-39) | 34 (27-42) | 21 (15-29) | 9.85 | <0.001 |
| VVD, ms | 32 (20-68) | 27 (19-57) | 40 (23-85) | -4.12 | <0.001 |
| A-AntSept, M (Q₁, Q₃) | 314.00 (272.75, 356.00) | 306.00 (273.50, 343.00) | 320.00 (271.00, 376.00) | -1.66 | 0.097 |
| M-AntSept, M (Q₁,Q₃) | 321.00 (279.00, 377.00) | 320.00 (282.50, 365.50) | 327.00 (274.00, 404.00) | -1.28 | 0.2 |
| B-AntSept, M (Q₁, Q₃) | 320.00 (275.00, 376.00) | 320.00 (281.00, 368.00) | 320.00 (256.00, 383.00) | -0.33 | 0.74 |
| A-InfSept, M (Q₁, Q₃) | 320.00 (277.00, 359.25) | 320.00 (277.50, 356.00) | 322.00 (273.00, 368.00) | -0.14 | 0.891 |
| M-InfSept, M (Q₁, Q₃) | 317.00 (274.00, 367.00) | 317.00 (277.00, 360.50) | 317.00 (265.00, 376.00) | -0.21 | 0.832 |
| B-InfSept, M (Q₁, Q₃) | 309.50 (264.50, 368.00) | 313.00 (276.50, 367.50) | 300.00 (220.00, 369.00) | -2.01 | 0.045 |
| A-Ant, M (Q₁, Q₃) | 302.00 (261.75, 349.25) | 302.00 (276.00, 345.00) | 304.00 (244.00, 363.00) | -0.48 | 0.633 |
| M-Ant, M (Q₁, Q₃) | 322.00 (277.75, 375.00) | 320.00 (285.00, 361.00) | 330.00 (257.00, 381.00) | -0.07 | 0.943 |
| B-An, M (Q₁, Q₃) | 331.00 (281.75, 387.00) | 320.00 (281.50, 376.00) | 344.00 (283.00, 400.00) | -1.5 | 0.133 |
| A-Lat, M (Q₁, Q₃) | 303.50 (261.75, 344.25) | 309.00 (268.50, 341.00) | 295.00 (245.00, 354.00) | -0.89 | 0.373 |
| M-Lat, M (Q₁, Q₃) | 321.00 (279.00, 372.50) | 320.00 (287.00, 363.50) | 321.00 (240.00, 390.00) | -0.68 | 0.496 |
| B-Lat, M (Q₁, Q₃) | 331.00 (284.00, 383.00) | 321.00 (288.00, 375.50) | 339.00 (274.00, 394.00) | -0.76 | 0.446 |
| A-Post, M (Q₁, Q₃) | 299.00 (252.50, 355.25) | 300.00 (272.00, 341.50) | 293.00 (191.00, 369.00) | -1.29 | 0.196 |
| M-Post, M (Q₁, Q₃) | 319.00 (272.00, 370.25) | 319.00 (284.50, 362.50) | 315.00 (210.00, 381.00) | -1.36 | 0.175 |
| B-Post, M (Q₁, Q₃) | 316.00 (266.00, 366.25) | 316.00 (277.50, 361.00) | 316.00 (237.00, 373.00) | -1.38 | 0.168 |
| A-Inf, M (Q₁, Q₃) | 312.00 (272.00, 362.25) | 311.00 (276.50, 350.00) | 315.00 (262.00, 373.00) | -0.1 | 0.92 |
| M-Inf, M (Q₁, Q₃) | 314.50 (273.00, 363.00) | 316.00 (281.00, 357.00) | 310.00 (244.00, 369.00) | -1.68 | 0.092 |
| B-Inf, M (Q₁, Q₃) | 317.00 (271.00, 371.00) | 317.00 (278.00, 367.50) | 315.00 (236.00, 371.00) | -1.43 | 0.152 |
| Ethnicity, n (%) | 3.16 | 0.076 | |||
| Han Chinese | 240 (58.25) | 148 (61.92) | 92 (53.18) | ||
| Other | 172 (41.75) | 91 (38.08) | 81 (46.82) | ||
| Gender, n (%) | 1.95 | 0.163 | |||
| Male | 302 (73.30) | 169 (70.71) | 133 (76.88) | ||
| Female | 110 (26.70) | 70 (29.29) | 40 (23.12) | ||
| LBBB, n(%) | 18.09 | <0.001 | |||
| No | 378 (91.75) | 231 (96.65) | 147 (84.97) | ||
| Yes | 34 (8.25) | 8 (3.35) | 26 (15.03) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
