Antiplatelet therapy for AMI continues to face significant challenges in balancing the reduction of thrombotic events with the control of bleeding risks. While traditional DAPT effectively reduces ischemic risk, its “one-size-fits-all” approach has notable limitations. In recent years, individualized treatment strategies based on platelet function testing (PFT), genotyping, and dynamic risk stratification have emerged as promising approaches to optimize clinical decision-making.

Current guidelines recommend adjusting the duration of DAPT dynamically according to ischemic and bleeding risks. Although DAPT reduces ischemic events following stent implantation, it also increases bleeding risks, particularly in patients with ACS or high bleeding risk (HBR). A trial randomly assigned 6002 patients to prasugrel monotherapy following PCI using everolimus-eluting stents. Patient outcomes after 1 month of prasugrel monotherapy were compared to those after 1 month of DAPT. The results showed that prasugrel monotherapy did not significantly reduce bleeding risk compared to DAPT, despite demonstrating non-inferiority in ischemic outcomes. However, it was associated with an increased risk of stent thrombosis and revascularization, especially among ACS patients. Consequently, DAPT remains the standard therapy for early ischemic protection [23]. The PRECISE-DAPT score integrates 7 indicators, including hemoglobin levels, renal function, and bleeding history. It is used to classify patients into non-high, high, and extremely high bleeding risk categories, providing a standardized basis for de-escalation therapy [24]. For instance, HBR patients receiving 1 month of prasugrel monotherapy exhibited significantly lower major bleeding risks (particularly in the ACS subgroup) compared to prolonged DAPT, with non-inferior ischemic endpoints [25]. Nevertheless, the applicability of this score to East Asian populations requires further validation.

The timing, time window, and efficacy of antiplatelet therapy in AMI patients remain controversial topics. Specifically, the optimal timing of pre-treatment with the P2Y₁₂ inhibitor in STEMI patients is still being debated. A registry study involving 1624 patients undergoing direct PCI demonstrated that pre-PCI administration, as compared to intra-procedural administration, reduced the 30-day MACE risk, particularly when administered $\ge$80 minutes before PCI [26]. Peak effects were observed at 180–240 minutes. This finding underscores the potential benefits of optimizing preoperative medication protocols to enhance ischemic protection, while balancing operational feasibility and bleeding risks.

Precision interventions guided by PFT and genotyping have shown improved outcomes. CYP2C19 gene polymorphisms significantly influence clopidogrel metabolism, with carriers of loss-of-function (LOF) alleles exhibiting a 30% higher ischemic risk [27]. A prospective study confirmed that a genotyping-guided de-escalation strategy (switching LOF carriers to ticagrelor or prasugrel) reduced ischemic events by 22% and bleeding risks by 53% at 12 months [28]. Despite its cost-effectiveness, the application of CYP2C19-guided strategies is limited by poor access to genetic testing, especially in developing countries, and insufficient standardization of result interpretation. AI-assisted decision-making systems may accelerate the analysis of genotype-phenotype associations, although the generalizability of these algorithms requires further verification. Additionally, rapid genotyping platforms, such as point-of-care testing, can expedite precise decision-making but must address challenges such as the potential omission of rare variants and delays in clinical decision-making (median genotyping time: 48 h).

Current antiplatelet strategies for AMI are designed to balance the reduction of thrombotic risk with the potential for bleeding complications. A standardized DAPT regimen is not universally recommended due to interpatient variability. Dynamic risk stratification tools, such as the PRECISE-DAPT score, may help to determine optimal treatment duration and guide de-escalation strategies, particularly in HBR patients. Precision medicine approaches, including the use of CYP2C19 genotyping to inform the selection of P2Y₁₂ inhibitors, can improve clinical outcomes among poor metabolizers of clopidogrel. Personalized therapeutic strategies improve the overall risk-benefit profile, and genotyping has demonstrated cost-effectiveness in this context. However, challenges remain regarding the accessibility and turnaround time of genotyping technologies. Future directions include the broad implementation of rapid point-of-care genetic testing, as well as the integration of AI-based clinical decision support systems to facilitate real-time, individualized treatment (Table 2).

The management of AMI centers primarily on repairing damaged myocardium and inhibiting adverse remodeling processes. Conventional cell-based therapies encounter challenges such as immunogenicity, arrhythmia risks, and difficulties in large-scale production. In contrast, extracellular vesicles (EVs), also known as exosomes, have emerged as an exciting prospect in regenerative medicine due to their immunomodulatory, pro-angiogenic, and non-toxic properties. EVs can precisely regulate inflammatory responses and cellular fate within the infarcted microenvironment by delivering functional nucleic acids (e.g., miRNA) and proteins. Mesenchymal stem cell-derived EVs (MSCATV-EVs), pre-treated with atorvastatin, carry high levels of miR-139-3p, which drives macrophage polarization from the pro-inflammatory M1 type to the reparative M2 type by inhibiting the Stat1 signaling pathway [29]. In rat models, intramyocardial injection of MSCATV-EVs reduces the infarct area by 28%, increases left ventricular ejection fraction (LVEF) by 15%, and significantly decreases the expression of pro-inflammatory cytokines such as IL-1$\beta$ and TNF-$\alpha$. Mechanistic studies confirm that knockdown of miR-139-3p in EVs abolishes their therapeutic effects, highlighting this pathway as a critical target for enhancing myocardial repair.

Small EVs derived from human cardiac progenitor cells (CPC-sEVs) and produced under Good Manufacturing Practice (GMP) standards represent a breakthrough in clinical-grade, “off-the-shelf” therapies. CPC-sEVs are compatible with coronary artery (CA) delivery during PCI procedures. In pig AMI models, IC infusion of CPC-sEVs (2 $\times$ 10¹¹ particles) reduced the infarct area by 25% compared to intramyocardial injection (18.0% vs. 23.9%), promoted angiogenesis at the infarct border zone, reduced myocardial fibrosis, and prevented operation-related hemodynamic damage [30]. The standardized preparation and immediate availability (“off-the-shelf” product) of CPC-sEVs facilitates their rapid clinical translation. These findings support the clinical application of IC-infused CPC-sEVs as a novel extracellular therapy for AMI patients, overcoming the limitations of traditional cell-based approaches.

EVs derived from mesenchymal/stromal cells or cardiac progenitor cells exhibit immunomodulatory, pro-angiogenic, and cardioprotective effects, offering advantages such as stability, scalability, and reduced immunogenicity compared to conventional cell therapies. Preclinical studies in large animal models (pigs, primates) demonstrate that EVs delivered via IC, intramyocardial (IM), or through engineered stents reduce infarct size, improve left ventricular function, and mitigate adverse remodeling. The key mechanisms include regulation of macrophage polarization, inhibition of fibrosis, and promotion of angiogenesis. While IC delivery aligns with PCI workflows and provides sustained benefits of up to 3 months, IM injection carries risks of procedural damage [31].

Despite the promising preclinical outcomes with EVs, significant translational challenges remain. A primary hurdle is the scalable manufacturing of clinical-grade EVs under strict GMP conditions. This includes standardizing the isolation methods (e.g., ultrafiltration, size-exclusion chromatography), defining critical quality attributes (e.g., particle concentration, surface markers, miRNA cargo), and ensuring batch-to-batch consistency [32]. Furthermore, optimal dosing regimens (single vs. multiple doses), delivery routes (intracoronary vs. systemic), and long-term biodistribution and safety profiles in humans are yet to be fully established [33].

Several early-phase clinical trials are directly addressing these challenges. The EV-AMI trial is evaluating the safety and efficacy of allogeneic MSC-derived exosomes administered intravenously in patients with AMI. Similarly, the SECRET-HF trial is investigating the impact of cardiosphere-derived cell (CDC)-exosomes in patients with non-ischemic cardiomyopathy. These studies will provide crucial initial data on safety, tolerability, and preliminary efficacy in humans.

A recent study developed a composite hydrogel-EV system, DHPM(4APPC)-EVs, which addresses the limitation of short retention times. This system was constructed using CXCR2-overexpressing exosomes and responds to acidic pH/H₂O₂/MMP9 signals in the infarct microenvironment, enabling the following sequential release: pH-triggered gelation captures exosomes at the infarct site, followed by H₂O₂-induced stabilization and MMP9-mediated sustained release. In rat AMI models, DHPM(4APPC)-EVs reduce infarct size, enhance LVEF, suppress pro-inflammatory cytokines (IL-1$\beta$, IL-6, TNF-$\alpha$), and promote collagen III synthesis. Compared to repeat exosome injections, the hydrogel platform extends exosome retention ($>$7 days), reduces systemic toxicity, and improves survival rates. This biomaterial strategy also resolves the issue of rapid EV clearance during long-term cardiac repair after AMI, offering a highly promising non-invasive alternative [34].

EVs purified from pooled human platelet concentrates (SCPL-EVs) exhibit potent cardioprotective effects in MI/reperfusion (I/R) injury models. In a rodent model, SCPL-EVs reduced cardiomyocyte apoptosis, enhanced angiogenesis, and improved cardiac function. Crucially, platelet concentrates are clinical-grade materials with established safety profiles for transfusion, thus streamlining the regulatory process. Preliminary human trials of platelet-derived EVs for wound healing found no safety concerns, supporting their translational potential. The scalable and cost-effective production of SCPL-EVs, coupled with their stability and low immunogenicity, makes them promising candidates for rapid advancement into clinical trials for AMI [35].

EVs derived from MSC/CPC sources mitigate myocardial injury via targeted delivery of miRNAs/proteins, thereby promoting macrophage M2 polarization, angiogenesis, and antifibrotic effects. Intracoronary infusion optimizes delivery [36], while smart hydrogels enhance retention ($>$7 days). However, clinical translation still faces challenges in terms of GMP standardization, pharmacokinetics, and regulatory approval. Ongoing trials (e.g., EV-AMI, NCT05191381) are currently evaluating the safety and efficacy of allogeneic EVs. Future directions include engineered EVs with tailored cargo and hybrid hydrogel systems for sustained release, enabling progress toward clinical application in cardiac regeneration. The convergence of EV biology and biomaterial engineering holds immense potential for developing effective and clinically translatable therapies for cardiac regeneration.

The integration of AI into precision medicine has significantly transformed the diagnosis and risk assessment of CVDs. In 2021, Chang et al. [37] introduced a multi-label AI system designed to detect arrhythmias and STEMI. By leveraging a long short-term memory (LSTM) model, this system simultaneously diagnosed 12 types of arrhythmias and STEMI using 12-lead ECGs. It demonstrated very high accuracy (AUC of 0.987), outperforming both cardiologists and commercial algorithms. This highlights its potential for real-time triage of patients with chest pain [37].

Such systems can be integrated into emergency department workflows to provide immediate decision support, helping clinicians prioritize high-risk patients for urgent catheterization based on AI-interpreted ECG findings. This practice is increasingly being reflected in recent chest pain guidelines that emphasize the use of advanced diagnostic technologies.

Expanding on this work, Han et al. [38] in 2021 tackled the challenge of asynchrony in smartwatch ECG signals by developing a residual neural network enhanced with a self-attention mechanism. Their model demonstrated diagnostic accuracy comparable to traditional 12-lead ECGs (AUC ranging from 0.845 to 0.880) across various configurations, enabling outpatient settings to conduct real-time risk stratification. This technology allows continuous ambulatory monitoring of individuals with suspected arrhythmias, aligning with the 2023 ACC/AHA Guidelines on the Management of Patients with Chronic Coronary Disease, which endorse the use of wearable devices for enhancing patient surveillance and early intervention [39].

In 2023, De Michieli et al. [40] validated the diagnostic efficacy of an artificial intelligence Electrocardiogram (AI-ECG) for identifying left ventricular systolic dysfunction (LVSD) in a study of 1977 emergency department patients. This tool identified high-risk individuals with a MACE rate of 48% within two years, showcasing its complementary prognostic value alongside high-sensitivity troponin T (hs-cTnT) biomarkers and highlighting the role of AI in addressing diagnostic gaps in acute care scenarios. The above findings support the recommendations of the 2021 European Society of Cardiology (ESC) Guidelines for the Diagnosis and Treatment of Acute and Chronic Heart Failure, which encourage the use of multidimensional risk stratification tools to identify high-risk patients who may benefit from early advanced imaging or intervention [41].

Large-scale observational studies further confirm the scalability of AI. For instance, the 2022 EURObservational program by the ESC analyzed data from 3620 NSTEMI patients across 59 countries, revealing variations in guideline adherence and patient outcomes [42]. Although not directly AI-driven, such registry studies emphasize the need for AI tools in standardizing care for high-risk populations and minimizing variability. AI-driven clinical decision support systems (CDSS) can now be embedded within electronic health records (EHRs) to alert physicians when guideline-recommended treatments such as DAPT or statin use are overlooked, thereby improving adherence and reducing practice variation.

Awasthi et al. [43] reported in 2023 a series of deep learning models that utilized 7.1 million ECGs to detect coronary artery calcification (CAC $\ge$300), obstructive coronary artery disease (CAD), and left ventricular regional wall motion abnormalities, which are indicators of MI. These models achieved AUROC values of 0.88, 0.85, and 0.94, respectively, surpassing traditional 10-year atherosclerotic cardiovascular disease (ASCVD) scores by identifying subclinical disease and improving the prediction accuracy of short-term events [44]. These tools are particularly useful in primary care settings for refining cardiovascular risk stratification beyond conventional scores, enabling earlier initiation of preventive therapies as recommended by the 2023 ACC/AHA Guideline on the Primary Prevention of Cardiovascular Disease [39].

Bock et al. [44] showcased advancements in multimodal integration by combining clinical variables and exercise stress test ECGs to predict functional coronary artery disease (fCAD) [43]. Herman et al. [45] validated an AI-based, 12-lead ECG model in an international cohort for detecting acute coronary occlusion myocardial infarction (OMI). This model outperformed STEMI criteria (sensitivity 80.6% vs. 32.5%) and matched expert interpretations (accuracy 90.9%), demonstrating its utility in prioritizing urgent revascularization for NSTEMI patients [45]. This approach supports the 2023 ESC Guidelines for the Management of Acute Coronary Syndromes, which stress the importance of early detection of OMI regardless of ST-segment elevation, particularly in patients with ongoing symptoms and ambiguous ECG changes.

Approximately 24% to 35% of patients with non-ST-segment elevation acute coronary syndrome (NSTE-ACS) experience complete coronary artery OMI requiring immediate reperfusion. Current reliance on ST-segment elevation (STE) for diagnosing OMI often results in delayed treatment and higher mortality rates. Researchers recently developed a machine learning (ML) model using data from three U.S. cohorts comprising 7313 patients. This ML model detected subtle ECG patterns that were overlooked by clinicians, such as ST-segment depression and T-wave inversion. When combined with clinical judgment, the OMI score reclassified one-third of chest pain patients, reducing uncertain evaluations. This framework improved OMI detection, thereby facilitating earlier reperfusion and optimizing resource allocation, although further prospective trials are necessary to validate clinical outcomes [46]. Incorporating such ML-based scores into existing ACS algorithms could help emergency physicians identify candidates for urgent angiography more accurately, thereby implementing the Class IIa recommendation in current guidelines regarding early invasive strategy in high-risk NSTE-ACS patients. Zepeda-Echavarria et al. [47] developed a portable four-electrode mini-ECG device capable of detecting AMI caused by different culprit vessels (left anterior descending artery, right coronary artery, and left circumflex artery) with a sensitivity of 65% and specificity of 92%. This innovation enables rapid and targeted ischemia assessment in emergency settings, reducing the intervention time [47]. Their integrated model, employing Shapley Additive Explanations (SHAP) analysis to pinpoint key predictors like age and coronary heart disease history, outperformed cardiologists (AUC 0.71 vs. 0.64), highlighting the importance of integrating AI into clinical decision-making. Such portable devices are especially valuable in resource-limited settings or pre-hospital environments, allowing paramedics to perform ECG acquisition and AI-supported analysis en route to the hospital. This shortens the door-to-balloon time, a key performance indicator in STEMI care systems.

The role of AI in clinical workflows is expected to expand significantly by 2025. The ROMIAE study applied an AI ECG algorithm to 8493 emergency patients, achieving diagnostic accuracy comparable to the HEART score (AUROC 0.878 vs. 0.877) while combining high-sensitivity troponin data to enhance risk stratification, giving a net reclassification improvement of 19.6% [48]. This suggests that AI can improve clinical decision-making by prioritizing high-risk patients and minimizing unnecessary catheterizations. The 2022 AHA/ACC Key Data Elements and Definitions for Chest Pain and Acute Myocardial Infarction also advocates the incorporation of AI-derived metrics into existing risk scores to improve triage accuracy and reduce low-yield cardiac testing [49].

Collectively, these studies highlight the transformative impact of AI on precision medicine—from enhancing risk prediction through multimodal data integration, to enabling immediate diagnosis via wearable/mobile devices. Despite challenges related to demographic biases and ensuring universal applicability, the trend toward AI-driven personalized cardiovascular care continues to gain momentum (Fig. 1).

Fig. 1.

Scenario map of AI applications in AMI management. AI, artificial intelligence; AMI, acute myocardial infarction; FFR, fractional flow reserve; ECG, Electrocardiogram; ST, ST Segment; AI-ECG, Artificial intelligence-Electrocardiogram; CTA, Computed Tomography Angiography. This figure was created using Edraw Max (www.edrawsoft.com).

Recent advancements in AI have begun to revolutionize precision medicine, particularly in cardiovascular care, offering tools that enhance clinical decision-making and personalize patient management. For example, Samant et al. [50] introduced the Artificial Intelligence-Simulations-Extended Reality (AISER) paradigm, which integrates AI, computational simulations, and extended reality. AISER shows promise in transforming preprocedural planning, virtual trials, and medical education in interventional cardiology. These authors emphasized how AI-driven analytics, in combination with patient-specific simulations and augmented reality (AR), can optimize procedures such as stent deployment and structural heart interventions, providing intuitive visual guidance that supports surgical execution.

Another study employed integrative ML to identify diagnostic genes (AQP9 and SOCS3) from leukocyte transcriptomic data in AMI, which were further validated through correlations with immune cell infiltration [51]. This approach highlights the growing role of AI in biomarker discovery and individualized risk assessment, offering a pathway toward more targeted diagnostic strategies.

Koo et al. [52] developed an AI-enabled, quantitative coronary plaque analysis model (AI-QCPHA) that integrates FFR, plaque burden, and myocardial blood flow. This model outperformed conventional CAD-RADS/HRP scoring in identifying culprit lesions in 351 patients with acute coronary syndrome, demonstrating its potential as a decision-support tool in catheterization laboratories. Similarly, Yang et al. [53] created an explainable XGBoost model to predict in-hospital mortality among Chinese STEMI patients, enhancing interpretability and clinical adoption.

Further expanding the clinical applicability of AI, Lu et al. [54] conducted a multi-center echocardiographic substudy within the PARAGON-HF trial, linking left ventricular mass, filling pressures, and right ventricular dysfunction to renal decline in heart failure with preserved ejection fraction (HFpEF) patients. ML identified LV mass index and E/e’ ratio as independent predictors of renal events, highlighting the value of AI in elucidating complex cardiorenal interactions. In parallel, Aghezzaf et al. [55] deployed a model using transthoracic echocardiography (TTE) parameters—including Left Ventricular Outflow Tract Velocity Time Integral (LVOT VTI), Tricuspid Annular Plane Systolic Excursion (TAPSE), and Systolic Pulmonary Artery Pressures (PAP)—to predict in-hospital MACE in intensive cardiac care unit (ICCU) patients. This model outperformed traditional Thrombolysis In Myocardial Infarction (TIMI) and Global Registry of Acute Coronary Events (GRACE) scores.

Collectively, these studies underscore the expanding role of AI in cardiology—from enhancing risk prediction and biomarker discovery to providing real-time diagnostic support. By improving accuracy and operationalizing data-rich workflows, AI helps to bridge the gap between data and clinical practice. However, challenges remain regarding the generalizability of models across diverse populations, as well as the seamless integration of these tools into routine care. Future efforts should focus on developing standardized reporting guidelines and validating AI systems within real-world clinical pathways to ensure equitable and effective adoption.

AI-enhanced ECG analysis has shown potential in improving the identification of CVDs, especially in emergency department settings, thereby enabling more rapid and accurate treatment decisions. However, despite the substantial promise of AI in cardiology, several limitations and challenges remain. The effective implementation of AI-based ECG interpretation is constrained by issues such as systemic bias. Dataset imbalances may introduce biases related to age, gender, and race. Insufficient representation of demographic diversity can adversely affect model performance. For example, age-related biases in training datasets might fail to account for ECG changes associated with aging. Physiological differences between sexes can also influence ECG patterns, meaning that datasets skewed toward one gender may compromise accuracy for the other gender. Furthermore, given that genetic variations can impact ECG readings, it is essential to ensure racial diversity within datasets. A recent study assessed an AI-enhanced ECG algorithm for detecting cardiac amyloidosis post-development [56]. Subtle variations were noted in its performance. The key findings were that overall AUC decreased from 0.91 (original version) to 0.84 (current validation). Although robustness was demonstrated across various demographics, including age, sex, and amyloid subtypes (AL/ATTR), notable discrepancies were identified among Hispanic patients (AUC 0.66), as well as specific ECG features such as left bundle branch block (AUC 0.76) and hypertrophy (AUC 0.75). Patterns associated with low voltage and infarcts maintained high accuracy ($>$0.90). The algorithm showed temporal stability, with no significant reduction in performance over a three-year span. This research underscores the importance of subgroup-specific validation to mitigate bias risks and ensure generalizability, particularly for underrepresented ethnic populations and complex ECG phenotypes. These limitations highlight critical considerations for clinical deployment and the future refinement of algorithms [56].

The ideal length of DAPT for patients at high risk of bleeding continues to be a subject of discussion. Current guidelines (2025) recommend ticagrelor or prasugrel over clopidogrel for 12 months in ACS patients without high bleeding risk. Shortened DAPT (1–3 months) followed by ticagrelor monotherapy is favored in ACS patients with high bleeding risk. In STEMI, prasugrel/ticagrelor is preferred over clopidogrel during PCI. For patients requiring long-term oral anticoagulation, aspirin is discontinued 1–4 weeks post-PCI, with continued P2Y₁₂ inhibition (preferably clopidogrel). Upstream ticagrelor/clopidogrel may be considered in NSTE-ACS planned for invasive evaluation $>$24 h. Radial access and intracoronary imaging are recommended to reduce ischemic or rebleeding risks. Updated recommendations prioritize tailored DAPT duration and P2Y₁₂ agent selection based on ischemic or bleeding risk, comorbidities, and procedural factors [1]. With regard to the optimal DAPT duration post-PCI in ACS, proponents have advocated 12-month DAPT based on landmark trials (CURE, TRITON, PLATO), emphasizing the reduced ischemic events despite increased bleeding risks. Critics argue these trials inadequately evaluated duration, citing newer evidence which shows that shorter DAPT (1–3 months) lowers bleeding risks without compromising ischemic outcomes in HBR patients. Meta-analyses have shown heterogeneity in trial populations (selection bias) and inconsistent ischemic event rates. Key challenges include balancing ischemic protection with bleeding risks, optimizing post-DAPT monotherapy, and addressing gaps in evidence for diverse ACS subgroups. Further large-scale RCTs are needed to refine personalized DAPT strategies [57].

While temporary Mechanical Circulatory Support (tMCS) adoption in AMI-CS is rising, its optimal timing and device-specific efficacy are still being debated. A large observational study (n = 294,839) using U.S. hospital data (2016–2020) found that early tMCS initiation (24 h) reduced in-hospital mortality (OR 0.90, p 0.001), shortened stays (7 vs. 15 days), and lowered costs compared to delayed use. However, no survival benefit was observed in unselected populations. Subgroup analyses revealed that VA-ECMO (OR 0.85, p = 0.041) and Impella (OR 0.92, p = 0.040) significantly improved outcomes with early use, whereas IABP showed minimal impact. Early adopters were younger and more likely to undergo revascularization. Limitations included coding bias and incomplete hemodynamic data, emphasizing the need for randomized trials to refine patient selection [58]. A separate multicenter study (n = 421) found that early Impella-initiated LV unloading ($\le$2 h before/after VA-ECMO) in CS patients reduced 30-day mortality (HR 0.64) and improved ventilator weaning (OR 2.17) compared to delayed ($>$2 h) unloading. Progressive delays correlated with increased mortality risk and hemodynamic deterioration (e.g., prolonged aortic retrograde flow exacerbating LV overload). While early unloading mitigated ischemia-reperfusion injury, randomized trials are needed to confirm the benefits and guide clinical protocols [59]. Although early tMCS improves outcomes in select AMI-CS patients, such as VA-ECMO and Impella users, this is not universal. Device selection and timing are therefore critical, with younger, revascularized patients gaining the most benefit. Hemodynamic monitoring and randomized trials are essential to personalize tMCS strategies in CS.