ARYA Atherosclerosis Journal

From association to causation: Integrating stable, subgroup-aware SHAP explanations with causal inference for trustworthy AI in coronary bifurcation PCI

Document Type : Original Article

Authors

Alireza Khosravi Farsani 1
Iman Zand 2
Ehsan Shirvani 2
Bashir Najafabdian 3
Mohaddeseh Behjati 4

1 Hypertension Research Center, Cardiovascular Research Institute, Isfahan University of Medical Sciences, Isfahan, Iran

2 Interventional Cardiology Research Center, Cardiovascular Research Institute, Isfahan University of Medical Sciences, Isfahan, Iran

3 Department of Biomedical Engineering, Science and Research Branch, Islamic Azad University, Tehran, Iran

4 Cardiac Rehabilitation Research Center, Cardiovascular Research Institute, Isfahan University of Medical Sciences, Isfahan, Iran

10.48305/arya.2025.45815.3117
Abstract
BACKGROUND: Coronary bifurcation lesions (CBLs) represent 15-20% of percutaneous coronary interventions (PCIs) and are associated with increased procedural complexity and adverse outcomes. Although machine learning (ML) models show promise in stratifying lesion complexity, their clinical adoption requires not only high accuracy but also robust, equitable, and causally interpretable explanations.
Objectives: To develop high-performance ML models for classifying CBL complexity, to rigorously evaluate the stability and subgroup heterogeneity of SHAP-based explanations, and to extend interpretability from associative feature attribution to causal effect estimation using a unified counterfactual framework.
METHODS: In a prospective registry of 500 consecutive PCI patients at Chamran and Askariyeh Hospitals, Isfahan, Iran, from March 2023 to January 2025, lesions were stratified into three classes: non-bifurcation (Class 0), simple bifurcation (Class 1), and complex bifurcation (Class 2). Ten key predictors spanning demographics, angiographic morphology, and procedural variables were selected. Four ML models were optimized via cross-validation. SHAP values were used to quantify feature contributions. Interpretability stability was assessed over 1000 bootstrap iterations using the SHAP Stability Index (SSI) and Kendall’s τ. Subgroup analyses were stratified by sex, diabetes, and hypertension. A causal inference pipeline, including a directed acyclic graph (DAG), propensity-stratified average treatment effects (ATE), doubly robust augmented inverse-propensity weighting (AIPW), mediation decomposition, and counterfactual simulation, was implemented.
RESULTS: The Random Forest model achieved near-perfect classification accuracy with high interpretability stability. Side-branch stenosis, heavy calcification, and bifurcation angle were the dominant global predictors. Subgroup analyses revealed clinically meaningful heterogeneity, with diabetic and female patients assigned substantially greater explanatory weight to calcification. Causal analysis confirmed that dual-stent use was associated with increased procedural risk, largely mediated through procedural intensity. Counterfactual simulation demonstrated that reducing side-branch stenosis meaningfully lowered predicted risk.
CONCLUSION: Integrating stable, subgroup-aware SHAP explanations with formal causal inference transforms ML from a predictive tool into a clinically actionable decision-support system. This dual framework enables clinicians to distinguish drivers of treatment selection from true determinants of outcome, supports equitable interpretation across patient subgroups, and quantifies the policy impact of hypothetical interventions, thereby advancing trustworthy, transparent, and individualized care in high-stakes interventional cardiology.

Keywords

Coronary Bifurcation
Machine learning
SHAP Analysis
Percutaneous Coronary Intervention
Lesion Complexity
Explainable AI
Model Stability
Subgroup Heterogeneity
Multimodal Visualization
Clinical Decision-Making
1. Lugo-Gavidia LM, Alcocer-Gamba MA, Martinez-Cervantes A. Challenges and Advances in Interventional Cardiology for Coronary Artery Disease Management. Medicina (Kaunas). 2024 Aug 15;60(8):1323. https://doi.org/10.3390/medicina60081323
2. Panayotov P, Mileva N, Vassilev D. Current Challenges in Coronary Bifurcation Interventions. Medicina (Kaunas). 2024 Sep 3;60(9):1439. https://doi.org/10.3390/medicina60091439
3. Ford TJ, McCartney P, Corcoran D, Collison D, Hennigan B, McEntegart M, et al. Single- Versus 2-Stent Strategies for Coronary Bifurcation Lesions: A Systematic Review and Meta-Analysis of Randomized Trials With Long-Term Follow-up. J Am Heart Assoc. 2018 May 25;7(11):e008730. https://doi.org/10.1161/JAHA.118.008730
4. Lassen JF, Holm NR, Banning A, Burzotta F, Lefèvre T, Chieffo A, et al. Percutaneous coronary intervention for coronary bifurcation disease: 11th consensus document from the European Bifurcation Club. EuroIntervention. 2016 May 17;12(1):38-46. https://doi.org/10.4244/EIJV12I1A7
5. Chen SL, Ye F, Zhang JJ, Xu T, Tian NL, Liu ZZ, et al. Randomized comparison of FFR-guided and angiography-guided provisional stenting of true coronary bifurcation lesions: the DKCRUSH-VI trial (double kissing crush versus provisional stenting technique for treatment of coronary bifurcation lesions VI). JACC Cardiovasc Interv. 2015;8(4):536-46. https://doi.org/10.1016/j.jcin.2014.12.221
6. Mohamed MO, Lamellas P, Roguin A, Oemrawsingh RM, Ijsselmuiden AJJ, Routledge H, et al. Clinical Outcomes of Percutaneous Coronary Intervention for Bifurcation Lesions According to Medina Classification. J Am Heart Assoc. 2022 Sep 6;11(17):e025459. https://doi.org/10.1161/JAHA.122.025459
7. Katona A, von Koch S, Andell P, Völz S, Omerovic E, Fröbert O, et al. Long-term prognosis after coronary bifurcation PCI-A nationwide observational study. PLoS One. 2025 Mar 26;20(3):e0317628. https://doi.org/10.1371/journal.pone.0317628
8. Panayotov P, Mileva N, Vassilev D. Current Challenges in Coronary Bifurcation Interventions. Medicina (Kaunas). 2024 Sep 3;60(9):1439. https://doi.org/10.3390/medicina60091439
9. Zhao S, Wu W, Samant S, Khan B, Kassab GS, Watanabe Y, et al. Patient-specific computational simulation of coronary artery bifurcation stenting. Sci Rep. 2021 Aug 13;11(1):16486. https://doi.org/10.1038/s41598-021-95026-2
10. Arunothayaraj S, Behan MW, Lefevre T, Lassen JF, Chieffo A, Stankovic G, et al. Stepwise provisional versus systematic culotte for stenting of true coronary bifurcation lesions: five-year follow-up of the multicentre randomised EBC TWO Trial. EuroIntervention. 2023;19(4):e297-307. https://doi.org/10.4244/EIJ-D-23-00211
11. Sadeghi AM, Behjati M, Najafabadian B, Shirneshan K, Sadeghi MM. Long-Term Survival Following Heart Transplantation in a Patient With Dilated Cardiomyopathy and Papillary Thyroid Carcinoma. Acta Med Iran. 2026;64(8). https://doi.org/10.18502/acta.v64i5.22217
12. Najafabadian B, Nasrabadi AM, Rashidi S. GSC-ABTA: A group-level brain sources connectivity framework based on adaptive block tensor analysis. Biomed Signal Process Control. 2025;110:108336. https://doi.org/10.1016/j.bspc.2025.108336
13. Bujak K, Verardi FM, Arevalos V, Gabani R, Spione F, Rajwa P, et al. Clinical outcomes following different stenting techniques for coronary bifurcation lesions: a systematic review and network meta-analysis of randomised controlled trials. EuroIntervention. 2023 Oct 23;19(8):664-75. https://doi.org/10.4244/EIJ-D-23-00013
14. Avesta M, Avesta M, Yousefian Baboukani A, Najafabadian B, Behjati M. CardioMD-FT2RL: type-2 fuzzy reinforcement learning for uncertainty-aware multimodal cardiac diagnosis. In: 10th International Congress on Fuzzy and Intelligent Systems of Iran (CFIS2025). 2025.
15. Avesta M, Avesta M, Yousefian Baboukani A, Najafabadian B, Behjati M. STA-EEGNet: spatial-temporal attention enhanced EEGNet architecture with adaptive model compression for efficient brain-computer interfaces. In: 34th International Conference on Electrical Engineering. 2026.
16. Avesta M, Avesta M, Yousefian Baboukani A, Najafabadian B, Behjati M. EchoFusion: multi-view transformer architecture with anatomical constraints for comprehensive echocardiographic assessment. In: 11th International Conference on Signal Processing & Intelligent Systems. 2025. https://doi.org/10.1109/ICSPIS68676.2025.11551724
17. Avesta M, Avesta M, Yousefian Baboukani A, Najafabadian B, Behjati M. Multi-agent large language model framework for comprehensive ECG signal analysis: delineation, classification, and arrhythmia subtyping with temporal hierarchical reasoning. In: 34th International Conference on Electrical Engineering. 2026.
18. Erabi G, Karimi S, Behkar A, Allafi D, Shahveghar R, Pooryai A, et al. Coconut Left Atrium: Causes and Management. Echocardiography. 2025 Sep;42(9):e70282. https://doi.org/10.1111/echo.70 282
19. Burzotta F, Louvard Y, Lassen JF, Lefèvre T, Finet G, Collet C, et al. Percutaneous coronary intervention for bifurcation coronary lesions using optimised angiographic guidance: the 18th consensus document from the European Bifurcation Club. EuroIntervention. 2024 Aug 5;20(15):e915-26. https://doi.org/10.4244/EIJ-D-24-00160
20. Najafabadian B, Baghoolizadeh H, Asadi E, Najafabadian A. ECG amplifier design with high input impedance for long-term monitoring. In: International Congress of Recent Advances in Engineering. 2013.
21. Najafabadian B, Baghoolizadeh H, Asadi E, Najafabadian A. A novel method for Automated Real-Time ECG signal processing and long term monitoring. In: International Congress of Recent Advances in Engineering. 2013.
22. Louvard Y, Medina A. Definitions and classifications of bifurcation lesions and treatment. EuroIntervention. 2015;11 Suppl V:V23-6. https://doi.org/10.4244/EIJV11SVA5
23. Suleiman S, Coughlan JJ, Touma G, Szirt R. Contemporary Management of Isolated Ostial Side Branch Disease: An Evidence-based Approach to Medina 001 Bifurcations. Interv Cardiol. 2021 Apr 5;16:e06. https://doi.org/10.15420/icr.2020.30
24. Gil RJ, Bil J, Kern A, Iñigo Garcia LA, Formuszewicz R, Dobrzycki S, et al. Comparison of dedicated BIOSS bifurcation stents with regular drug-eluting stents for coronary artery bifurcated lesions: Pooled analysis from two randomized studies. Cardiol J. 2018;25(3):308-16. https://doi.org/10.5603/CJ.a2017.0098
25. Feldmann K, Cami E, Safian RD. Planning percutaneous coronary interventions using computed tomography angiography and fractional flow reserve-derived from computed tomography: A state-of-the-art review. Catheter Cardiovasc Interv. 2019 Feb 1;93(2):298-304. https://doi.org/10.1002/ccd.27817
26. Katona A, von Koch S, Andell P, Völz S, Omerovic E, Fröbert O, et al. Long-term prognosis after coronary bifurcation PCI-A nationwide observational study. PLoS One. 2025 Mar 26;20(3):e0317628. https://doi.org/10.1371/journal.pone.0317628
27. Raphael CE, O’Kane PD. Contemporary approaches to bifurcation stenting. JRSM Cardiovasc Dis. 2021 Feb 24;10:2048004021992190. https://doi.org/10.1177/2048004021992190
28. Najafabadian B, Najafabadian A, Motaharinia M, Tahmasebi GM, Abdollahzadeh T. Transcutaneous Electrical Stimulation with high frequency Signals: volume conductor and circuit models. J Med Signals Sens. 2024.
29. Kan J, Zhang JJ, Sheiban I, Santoso T, Munawar M, Tresukosol D, et al. 3-Year outcomes after 2-stent with provisional stenting for complex bifurcation lesions defined by DEFINITION criteria. JACC Cardiovasc Interv. 2022;15(13):1310-20. https://doi.org/10.1016/j.jcin.2022.05.026
30. Sousa JP, Puga L, Ribeiro J, Lopes J, Saleiro C, Gomes R, et al. Provisional versus 2-stent strategies for coronary bifurcations: is a bird in the hand worth two in the bush? Eur Heart J. 2020;41(Suppl 2):ehaa946-2541. https://doi.org/10.1093/ehjci/ehaa946.2541
31. Kerdabadi SA, Hoseini M, Najafabadian B, Hamidipour H. A high CMRR low noise CMOS amplifier for electrocardiogram signal recording applications. In: Proc 4th Nat Conf New Ideas Elect Eng. 2015:1-6.
32. Latib A, Colombo A. Bifurcation disease: what do we know, what should we do? JACC Cardiovasc Interv. 2008 Jun;1(3):218-26. https://doi.org/10.1016/j.jcin.2007.12.008
33. Farooq V, Brugaletta S, Serruys PW. Contemporary and evolving risk scoring algorithms for percutaneous coronary intervention. Heart. 2011 Dec;97(23):1902-13. https://doi.org/10.1136/heartjnl-2011-300718
34. Nakahara T, Dweck MR, Narula N, Pisapia D, Narula J, Strauss HW. Coronary Artery Calcification: From Mechanism to Molecular Imaging. JACC Cardiovasc Imaging. 2017 May;10(5):582-93. https://doi.org/10.1016/j.jcmg.2017.03.005
35. Sajjadieh Khajouei A, Goli F, Aria A, Babak A, Shahabi J, et al. Prognostic Clinical Characteristics Associated with High Mortality in COVID-19 Patients with Bradycardia: A Case Series Study. Int Cardiovasc Res J. 2024;18(1):e151449. https://doi.org/10.5812/icrj-151449
36. Khosravi A, Zand I, Shirvani E, Najafabadian B, Behjati M. Machine Learning with SHAP-Driven Interpretability Enhances Decision-Making in Coronary Bifurcation Percutaneous Coronary Intervention: A Prospective Study. Mod Care J. 2025;22(4):e160108. https://doi.org/10.5812/mcj-160108
37. Petrosian O, Zou J. Explainable AI: Efficiency Sequential Shapley Updating Approach. IEEE Access. 2024;12:166414-23. https://doi.org/10.1109/ACCESS.2024.3495543
38. Wang Z, Samsten I, Papapetrou P. Counterfactual explanations for survival prediction of cardiovascular ICU patients. In: International Conference on Artificial Intelligence in Medicine. Cham: Springer; 2021:338-48. https://doi.org/10.1007/978-3-030-77211-6_38
ARYA Atherosclerosis Journal
Volume 22, Issue 3
May and June 2026
Pages 13-26

Home

Guide for Authors

Submit Manuscript

Reviewers

Contact Us