Prognosis modelling of adverse events for post-PCI treated AMI patients based on inflammation and nutrition indexes

Abstract

Objective

This study aimed to evaluate the predictive performance of inflammatory and nutritional indices for adverse cardiovascular events (ACE) in patients with acute myocardial infarction (AMI) after percutaneous coronary intervention (PCI) using a machine learning (ML) algorithm.

Methods

AMI patients who underwent PCI were recruited and randomly divided into non/ACE groups. Inflammatory and nutritional indices were graded according to the laboratory examination reports. Logistic Regression was used to screen for factors that were significant for ML model establishment. The performances of the algorithms were evaluated in terms of accuracy, kappa, F1, receiver operating characteristic, precision recall curve, etc.

Results

Age, LVEF%, Killip Grade, heart rate, creatinine, albumin, neutrophil/lymphocyte ratio (NLR), platelet/lymphocyte ratio (PLR), and prognostic nutritional index (PNI) were significantly correlated with ACE by Logistic regression analysis (P < 0.05). These nine factors were employed to establish stepwise regression (SR), random forest (RF), naïve Bayes (NB), decision trees (DT), and artificial neutron network (ANN), whose performances were evaluated in terms of accuracy, kappa, F1, receiver operating characteristic, precision recall curve, etc. The accuracy of the decision tree was greater than that of other trees. The area under the curves was the highest in the ANN model compared with the other models.

Conclusion

ANN predictive performance had an advantage over other ML algorithms based on age, LVEF%, Killip Grade, heart rate, creatinine, albumin, NLR, PLR, and PNI.

Introduction

Cardiovascular disease is the leading cause of mortality and morbidity worldwide [1]. Approximate fourth five of cardiovascular death results from acute myocardial infarction (AMI) [2, 3]. The clinical application of percutaneous coronary intervention (PCI) has revolutionized the treatment and management of AMI, improving the quality of life and duration of heart attack [4]. Early blood flow restoration methods for patients mainly include pharmacological thrombolysis and PCI. With the establishment of chest pain centers and widespread development of PCI and other revascularization techniques, the survival rate of patients with AMI has significantly improved. However, prognosis management is crucial for patient quality of life, as numerous studies have indicated that AMI patients often experience adverse cardiovascular events (ACE) after percutaneous coronary artery treatment [5]. These events include acute heart failure, malignant arrhythmias, cardiogenic shock, sudden cardiac death, non-fatal stroke, cardiac arrest, and myocardial reinfarction. Therefore, adverse cardiovascular events following revascularization in AMI patients should not be overlooked.

Inflammation plays a pivotal role in the pathogenesis of atherosclerosis and its complications and serves as both a marker and mediator of disease progression. The systemic inflammatory response following AMI and subsequent PCI can influence the healing process, affecting myocardial repair and propensity for adverse outcomes. Although the prognostic nutritional index (PNI) in patients with AMI remains controversial, Huang et al. suggested that PNI might be a useful factor to predict patients at high risk of 1-year all-cause mortality among AMI patients [6]. Inflammatory processes play an important role in the development of AMI at different phases. In particular, low-grade inflammation continues to exist in the late stages of AMI [7]. Higher neutrophil counts and NLR could be associated with an increase in the odds ratio of ACE after AMI in patients treated with PCI [7, 8]. Nutritional status is another critical yet often overlooked determinant of outcomes in patients [9]. Malnutrition, characterized by deficiencies in essential nutrients, can impair immune function, exacerbate inflammation, and delay or complicate the healing process [10, 11]. Zengin et al. found that lower controlling nutritional status (CONUT) was associated with adverse events in AMI patients with ST segment elevation [12]. Conversely, optimal nutrition can bolster immune response, mitigate inflammatory damage, and promote myocardial repair. The integration of nutritional assessment into the prognostic evaluation of patients with AMI post-PCI represents a holistic approach to patient care, acknowledging the interdependence of nutrition, inflammation, and cardiovascular health. Kanda et al. suggested that the Geriatric Nutritional Risk Index might be a potential indicator for classifying mortality among AMI patients after PCI [13].

As a subset of artificial intelligence, machine learning (ML) can demonstrate underlying patterns or relationships based on medical information [14,15,16]. It could assist clinical doctors and medical scientists to explore disclosed invaluable risk factors, which promote predictive accuracy in the early detection, diagnosis, and prognosis of diseases [17, 18]. Far away from traditional statistical approaches, ML algorithms can manage big data with numerous variables and identify subtle patterns [17]. Its competence is particularly beneficial for disease prognosis and adventure prediction in the real world. ML integrates diverse data types to provide more accurate and personalized treatment strategies. ML has been used to predict ACE in AMI patients, but it has not been used to predict the prognosis of ACE after PCI treatment for AMI Patients [19,20,21]. To build a reliable model to predict these events is crucial for optimizing post-PCI management strategies, decreasing risks, and improving patient outcomes.

Materials and methods

Study population

This study recruited AMI patients who were admitted to the Emergency Department of the Chinese People’s Liberation Army (PLA) Joint Logistics Support Force 920 Hospital and underwent percutaneous coronary intervention (PCI) treatment from January 1, 2019, to December 31, 2023. This study was approved by the Medical Ethics Committee (NO 920IEC/AF/61/2021 − 01.1). To ensure our study cohort was well-defined and representative of the target population, participants inclusion and exclusion criteria were stated below. All participants fulfilled the following inclusion criteria: (1) patients were diagnosed with AMI, including ST-segment elevation myocardial infarction (STEMI) or non-ST-segment elevation myocardial infarction (NSTEMI); (2) AMI patients who received PCI treatment. Before surgery, all patients received thrombolytic therapy with aspirin 300 mg and clopidogrel 300 mg or aspirin 300 mg and ticagrelor 180 mg per os. (3) The age of the subjects was ≥ 18 years. 4) Patients with AMI post-PCI experienced ACE within 24 h. 5) All subjects voluntarily participated in this study and provided written informed consent. Exclusion criteria were as follows: (1) subjects with severe infection (depending on thoracic imaging, microbiological testing, or white blood cell count). (2) Subjects with severe liver, kidney, lung, etc. dysfunction; (3) Subjects diagnosed with severe hypoalbuminemia combined with chronic inflammatory, hematologic, rheumatic, or immunological diseases and various cancers. (4) Patients with genetic diseases in their families. (5) Patients were treated with corticosteroids or immunosuppressants. (6) Subjects were diagnosed with other cardiac diseases, including valvular dysfunction, constrictive pericarditis, myocarditis, non-ischemic cardiomyopathy, congenital heart disease, rheumatic heart disease, or others. (7) Patients with acute AMI were treated with thrombolytic therapy alone. (8) Patients did not complete the clinical data. (9) Patients who rejected signs of informed consent. All procedures were conducted in compliance with the guidelines of the Declaration of Helsinki (7th version).

Clinical data collection

Clinical information of patients with AMI was obtained from the Electronic Medical Records System, including age, sex, type of myocardial infarction, Killip categories, history of previous myocardial infarction, smoking history, hypertension, hyperlipidemia, hyperuricemia, diabetes, history of cerebrovascular accidents, systolic blood pressure (SBP), diastolic blood pressure (DBP), and heart rate (HR). Laboratory examinations included serum potassium (K+), white blood cell count (WBC), lymphocyte count (LY), neutrophil count (NE), monocyte count (MONO), platelet count (PLT), serum albumin (ALB), triglycerides (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), homocysteine (Hcy), creatinine (Cr), uric acid (UA), cardiac troponin I (CTN-I), and creatine kinase-MB (CK-MB). The left ventricular ejection fraction (LVEF) was measured and recorded using an ultrasonic cardiogram.

Inflammatory indices were calculated based on laboratory examination, such as neutrophil/lymphocyte ratio (NLR), monocyte/lymphocyte ratio (MLR), platelet/lymphocyte ratio (PLR), systemic immune-inflammatory index (SII), and systemic inflammation response index (SIRI). The SII was calculated using the formula: platelet count × neutrophil count/lymphocyte count. The SIRI equation was calculated as neutrophil count × monocyte count/lymphocyte count.

Prognostic nutritional index (PNI) and controlling nutritional status (CONUT) scores were used to indicate the patients’ nutritional status. The PNI equation was as follows: serum albumin (g/L) + 5 × peripheral blood lymphocyte count (10⁹/L). The CONUT Score was evaluated based on serum albumin, lymphocyte, and cholesterol levels. Serum albumin (g/L): ≥35 points, 0; 30 ~ 34.9 points, 2; 25 ~ 29.9 points, 4; <25 points, 6. Total lymphocyte count (per mm³): ≥1600, 0; 1200–1599, 1; 800–1199, 2; <800, 3. Total cholesterol (TC) (mg/dL): >180, 0; 140–180, 1; 100–139, 2; <100, 3. A total score greater than or equal to 3 indicates a risk of malnutrition.

Adverse cardiovascular events and grouping

The primary outcome of this study was nosocomial adverse cardiovascular events (ACE), which included cardiogenic death, cardiogenic shock, malignant arrhythmias (ventricular tachycardia, ventricular fibrillation, and atrial fibrillation), acute left heart failure, nonfatal cerebral stroke, myocardial re-infarction, and cardiac arrest [22]. These seven types of events emerged after PCI in hospital inpatients. Depending on whether ACE occurred in AMI patients post-PCI, participants were separated into two groups: non-ACE and ACE groups.

Statistical analysis

All data were analyzed using R software (version 4.4.0) and its package. Quantitative data are presented as means and 95% confidence intervals. Normality was assessed using the Kolmogorov-Smirnov test and quantile-quantile plot. Quantitative data with normal distribution were compared using Student’s t-test, while non-normally distributed data were analyzed using the Wilcoxon test. As two categories of non-ACE and ACE, logistic regression was employed to evaluate the relationship between ACE and the different factors. Significant factors were used to construct ML algorithm including stepwise regression (SR), random forest (RF), naïve bayes (NB), decision trees (DT), and artificial neutron network (ANN) in R package of “caret.” Two-thirds of the samples were randomly selected as the training data to build the models. The remaining one-third of the samples were used to test model efficacy. The performance of the predictive model was evaluated using accuracy, kappa, F1, specificity, sensitivity, precision, recall, balanced accuracy, McNemar’s test, ROC, and PRC. A two-tailed P value less than 0.05 (two-tailed) was considered statistically significant.

Results

Clinical characteristics

A total of 1078 patients were included in our study, comprising 861 males (79.86%) and 217 females (20.13%). Figure 1 shows the different disease compositions of ACE. The percentage of patients with acute left heart failure in the ACE group was 22.66%, which was the highest disease type compared to other ACE types. Cardiogenic shock and malignant arrhythmias were the second and third most common types of ACE, with patient percentages of 14.84% and 13.28%, respectively. The fourth patient percentage (6.25%) with ACE was a two-type ACE (cardiogenic shock and acute left heart failure) and three types of ACE (cardiogenic death, cardiogenic shock, and cardiac arrest). Comorbidity of six or seven diseases was not found in ACE.

Fig. 1

Different disease composition of ACE based on cardiogenic death, cardiogenic shock, malignant arrhythmias, acute left heart failure, nonfatal cerebral stroke, myocardial re-infarction, and cardiac arrest

The age of the ACE group was 66.31 (56.00, 78.00) years, which was greater than that of the non-ACE group was 59.97 (56.00–78.00) years (P < 0.001). The BMI was higher in ACE group rather than non-ACE group, whose values were 24.86 (23.59, 26.14) and 24.05 (23.85, 24.24), respectively (P = 0.209). Heart rate (HR) was also higher in the ACE group than in the non-ACE group (P < 0.001). In contrast, SBP, DBP, and LVEF% were elevated in the non-ACE group compared to those in the ACE group (P < 0.001). Table 1 depicts the basic clinical characteristics between the non-ACE and ACE groups; significant differences were found in sex, Killip grade, smoking history, hyperuricemia, and cerebrovascular accident (P < 0.05, respectively). Statistically significant differences were observed between the non-ACE and ACE groups in laboratory testing, including the levels of WBC, LY, NE, MONO, Cr, UA, ALB, Hcy, TC, and CTN-I (Table 2; P < 0.05) (Table 3).

Table 1 Basic clinical characteristics between non-ACE and ACE group

Table 2 Laboratory examination of non-ACE and ACE group

Table 3 Comparison of inflammatory and nutritional index between non-ACE and ACE group

Grading of inflammatory and nutritional conditions

Five inflammatory indices were evaluated and compared between non-ACE and non-ACE groups. The NLR, MLR, PLR, SII, and SIRI scores were lower in the non-ACE group than in the ACE group (P < 0.001). The PNI score was 44.74 (44.35, 45.13) in the non-ACE group, greater than 41.31 (40.28, 42.34) in the ACE group (P < 0.001). However, the CONUT score of the non-ACE group was significantly lower than that of the ACE group (2.00 vs. 3.00, P < 0.001).

Logistic regression

To further explore the association between clinical data or indices and ACE, significant variables were used to investigate the relationship with ACE. Table 4 shows the logistic regression analysis between clinical data and ACE. Six variables were significantly associated with ACE: LVEF%, age, Killip grade, HR, creatinine, and albumin (P < 0.05). The B values were positive for Killip grade, HR, and creatinine (0.397, 0.028, 0.011, respectively), but the LVEF% and albumin were negatively correlated with ACE, as their B values were negative (-0.104, -0.059, respectively). For inflammatory indices, NLR and PLR were significantly related to ACE (P = 0.0470 and 0.0194, respectively). NLR was positively associated with ACE (B = 0.066), whereas PLR was negatively associated with ACE (B = -0.008). Among nutritional factors, only PNI was significantly negatively correlated with ACE (P = 0.0362, B = -0.065) (Table 5).

Table 4 Logistic regression of clinical data with ACE

Table 5 Logistic regression of inflammatory index and nutritional grade with ACE

ML performance

According to the logistic regression analysis, significant factors were selected to construct the ML models, as shown in Table 6. In the ML model composed of three inflammatory or nutritional factors (NLR, PLR, and PNI), the SR, RF, and NB algorithms were successfully built. Specificities were approximately or greater than 0.9. However, their sensitivities were less than 0.5. The accuracy of the RF model was higher than that of the other two models, but the balanced accuracy of SR was greater than that of the other two models, achieving a value of 0.668. The precision and recall values were less than or equal to 0.5. The highest Kappa and F1 values were found for the NB model. The DT and ANN models were not successfully constructed because all predictive values were 0, which represented non-ACE according to their algorithm.

Table 6 Performance comparison of machine learning algorithm

By combining three inflammatory/nutritional factors and six clinical variables, SR, RF, NB, DT, and ANN models were successfully constructed, as shown in Table 6. The accuracies of the SR, DT, and ANN were greater than 0.9. The SR model exhibited the highest balanced accuracy (0.792). The Kappa and F1 values of the nine referent factor model were higher than those of the three referent factors. The specificities of the RF, DT, and ANN models were greater than 0.95, and the RF model had the highest specificity. The SR sensitivity was 0.667, which was greater than 0.5.

Each single index of inflammation or nutrition possessed the lowest AUC for PRC and ROC (Fig. 2A). In particular, the area under the ROC and PRC of PNI was 0.322 less than 0.5 and 0.084 less than 0.1. The areas under the curve of ROC and PRC in different ML models are described in Table 7; Fig. 2B&C. ANN models including nine referent factors showed the maximum in ROC and PRC, which had arrived at 0.783 and 0.479, respectively, in Table 7; Fig. 2C. Among the ANN factors, NLR was the most important factor in constructing the ANN model, meanwhile the least important factor (Fig. 2E). The neuron network illustrated input layers with nine factors, a hidden layer with five factors, and an output layer with ACE, as shown in Fig. 2F. The second largest PRC AUC of ML was for the SR model with three referent factors, but its PRC AUC was the third largest (Table 7; Fig. 2B). The DT model was built with nine factors; however, only five factors were used to stratify this model (Fig. 2D). The root node had a Killip grade of less than 4. 87% patients were divided into the non-ACE group, whose Killip grade was less than 4. The first internal node was the LVEF%, which was ≥ 43. Other internal nodes were age, HR, and PLR, with thresholds of 73, 79, and 163, respectively.

Fig. 2

Curves of ROC/PRC and visualization of machine learning model parameters. A. ROC and PRC of inflammatory and nutritional indexes. The area under curve of ROC and PRC in PNI were less than 0.5 and 0.1, respectively. B. ROC and PRC of SR, RF and NB algorithm with three significant factors. C. ROC and PRC of SR, RF, NB, DT and ANN algorithm with nine significant factors. D. Decision tree classification. E. Importance of nine factors in ANN algorithm. F. visualization of ANN

Table 7 Areas under curves (AUC) of ROC and PRC in different machine learning algorithm

Discussion

The prognosis of ACEs following PCI in patients with AMI is a pivotal concern for cardiologists in nosocomial patients. Huang et al. investigated nosocomial MACE after PCI for acute STEMI and found that eGFR, LVEF, cTnI, SBP, and Killp classification might be used to predict in-hospital MACE [23]. ACEs are valuable indicators for postoperative treatment success and safety. Inflammatory and nutritional conditions are dynamic parameters that may have the strongest predictive capacity following PCI. Several algorithms have been used to successfully and unsuccessfully predict ACE or MACE in AMI patients after PCI [24, 25]. Our study compared a variety of ML algorithms to predict ACE occurrence based on clinical data, laboratory examinations, inflammatory indices, and nutritional grading. This finding underscores the potential of ANN in predicting ACE in post-PCI AMI patients.

Higher age and lower LVEF% were risk factors

Consistent with previous studies, our analysis identified that patients who experienced ACE post-PCI were older and exhibited higher resting heart rates than their non-ACE counterparts. Several studies have examined the link between advancing age and heightened risk, revealing that older adults experience poorer outcomes and a higher incidence of complications following PCI than younger adults [26, 27]. In contrast to younger AMI patients, older adults with AMI face specific medical challenges, including frailty, multiple coexisting conditions, and heightened risks during procedures, complicating management after PCI. Additionally, a higher LVEF was protective, aligning with the understanding that better-preserved cardiac function post-AMI mitigates the risk of subsequent adverse events. Low LVEF is recognized as an important indicator of ACE in patients with AMI, which has been reported by researchers [28, 29]. Chyrchel et al. found that lower LVEF was significantly associated with a higher risk of ACE, which is consistent with our results. These findings reinforce the importance of age and cardiac function as simple yet powerful predictors of adverse event risk, which were integrated into comprehensive risk-predictive models.

Thrombolytic medication usage before PCI

All participants received thrombolytic therapy with aspirin, clopidogrel, or ticagrelor. Nils et al. compared prasugrel and ticagrelor treatment outcomes for acute coronary syndrome (ACS) in a cohort study, which suggested that prasugrel could reduce the risk of all-cause mortality, MI, and stroke [30]. But a meta-analysis conducted by Pravesh et al. implied that prasugrel land ticagrelor presented comparable adverse events in patients [31]. In overweight ACS patients, single prasugrel or dual prasugrel combined with clopidogrel treatment was associated with clinical benefit, but dual therapy was associated with non-overweight ACS patients only [32].

Creatinine and albumin roles in ACE

The relationship between creatinine level and ACE is complex. Elevated creatinine levels were associated with an increased odds ratio of ACE, reflecting the well-documented link between renal dysfunction and cardiovascular risk. Whether renal dysfunction is caused by contracted media or AMI itself is controversial. Immediate or mediate studies have shown that contract media usage accelerates ROS production and oxidative stress, which induces apoptosis and necrosis of cell membranes [33]. However, the negative association in certain analyses suggests a nonlinear relationship or the influence of confounding factors, warranting further investigation of the role of renal function in post-PCI prognosis. Our findings revealed a negative association between albumin levels and the risk of ACE, underscoring the potential protective role of albumin. Albumin, a serum protein, has emerged as a critical biomarker for the prognosis of ACE in patients with post-PCI AMI. This inverse relationship suggests that higher albumin levels may confer lower risk of adverse cardiovascular events [34]. The protective effects of albumin can be attributed to its anti-inflammatory and nutritional properties. A decreased albumin level represents a state of chronic inflammation or malnutrition that is unfavorable for cardiovascular disease [35]. Otherwise, the inflammation can induce malnutrition. ACE patients might experience more severe inflammation in local or systemic regions. Higher inflammation would prohibit albumin synthesis and promote its catabolism, so that serum albumin decreases in elevated inflammatory diseases [35, 36].

Suitable performance of ANN based on inflammatory and nutritional indicators

AMI is one of the leading causes of mortality worldwide, and the progression of atherosclerosis exacerbates chronic inflammation and malnutrition [37]. Intermountain Risk Score, Naples Score, Serum Uric Acid/Albumin Ratio were applied as indicator to evaluate the AMI mortality or post-PCI [38,39,40]. The intertwining of malnutrition and inflammation forms a vicious cycle, culminating in the advancement of atherosclerosis and ultimately increasing the mortality rate of cardiovascular diseases, thus highlighting the interconnection between nutritional status, inflammation, and cardiovascular diseases [41]. Published studies have demonstrated that monocytes have a closer relationship with the occurrence and development of atherosclerosis than platelets do. Following acute myocardial infarction, monocytes are recruited and subsequently differentiate into macrophages to mediate the development and resolution of inflammation, thereby producing inflammatory cytokines and nitric oxide and participating in the degradation of necrotic myocardium [42]. Meanwhile, neutrophils recruited to the infarcted myocardial site release large amounts of reactive oxygen species, promoting oxidative damage to the vascular wall and simultaneously generating neutrophil extracellular traps (NETs) to eliminate pathogens [43]. Neutrophils can be recruited before platelet activation and promote platelet activation and deposition, causing further myocardial damage in the infarcted area. Subsequently, neutrophils can maintain a chronic inflammatory environment [44]. Lymphocyte depletion can exacerbate oxidative and inflammatory damage and accelerate plaque rupture during acute myocardial infarction [45,46,47]. Li et al.. evaluate the effect of the SII in AMI patients with ACE, which could be a reliable biomarker to predict a higher risk of ACE [48]. This study demonstrated that SIRI serves as the optimal inflammatory indicator for predicting in-hospital ACE and is an independent predictor of malignant arrhythmias following PCI in patients with AMI. The PNI, composed of serum albumin and lymphocytes, was initially used to evaluate tumor-related prediction and gradually provides information on nutrition, inflammation, and immunology [49, 50]. It has been found that the prognosis of acute myocardial infarction is closely related to nutritional status [51, 52]. Initially, lymphocytes and their subgroups may participate in the microcirculatory obstruction during the ischemia-reperfusion process of myocardial infarction and release inflammatory factors, further aggravating myocardial injury [53, 54]. However, lymphocyte subgroups represented by T cells play a major role in inhibiting inflammatory responses and activating fibrosis to protect continuously damaged myocardial tissues during acute myocardial infarction [55]. As for the other component of the PNI, serum albumin levels temporarily decrease under the inflammatory effect of myocardial infarction, with studies indicating that serum albumin is an independent predictor of heart failure and cardiovascular death after AMI [34]. Thus, Cheng et al.. suggested that PNI and modified PNI presented lower prognostic value for AMI [56] Moreover, research has also found that PNI is independently associated with the long-term survival of acute heart failure patients with decreased or preserved left ventricular ejection fraction [57, 58]. In this study, PNI served as an independent predictor of in-hospital ACE in patients with AMI, with lower PNI levels being independently associated with acute heart failure, cardiogenic shock, sudden cardiac death, and cardiac arrest.

Generally, the performance of the nine factor-referenced models is more optimal than that of the three reference models because more variables tune down mean squared error [59, 60]. Deng et al. summarized previous prediction models for MACE after PCI, which implicated that Logistics and COX approaches were applied popularly [24]. However, SR, RF, NB, DT and ANN had been used to build models for neurosurgery, chronic kidney disease, heart disease, etc [61,62,63,64]. The superior performance of ANN over other ML algorithms in our study is a testament to the potential of advanced computational techniques for transforming cardiovascular prognostication. By integrating a wide array of variables, including age, LVEF%, Killip Grade, heart rate, creatinine, albumin, NLR, PLR, and PNI, ANN models can uncover complex nonlinear relationships that traditional statistical methods might miss. This approach not only enhances predictive accuracy, but also offers a more nuanced understanding of the multifactorial nature of post-PCI outcomes. Our study highlights the prognostic significance of inflammation and nutrition in a post-PCI setting. Elevated PLR, NLR, and PNI were associated with an increased risk of ACE, underscoring the intertwined roles of systemic inflammation and malnutrition status in AMI outcomes. These findings are in line with emerging evidence suggesting that targeted anti-inflammatory therapies and nutritional interventions may offer novel avenues for reducing post-PCI cardiovascular risks. Zhang et al. built a machine learning model to predict ACE in patients with AMI who underwent PCI. They found that artificial neural networks performed better than random forests, k-nearest neighbors, support vector machines, and logistic regression models [65]. However, they and other researchers have not analyzed the clinical outcomes of patients with AMI through a combination analysis of nutrition and inflammation. Meanwhile, ANN are advantageous for handling problems with complex internal, nonlinear, and dynamic mechanisms [66]. Their strengths include self-adaptability, self-organization, fault tolerance, and the ability to perform nonlinear mapping. Because of these qualities, ANNs have found wide application in many disciplines [60, 65].

This study is limited by its single-center, retrospective design and the relatively small sample size, especially the sample sizes of the non-ACE and ACE groups, which are not in disequilibrium. Additionally, there may be some information bias, and the research group was not subjected to long-term follow-up, further contributing to its limitations. Future clinical trials are needed to extend these findings to long-term outcomes.

Conclusions

Our findings highlight the intricate interplay between demographic, clinical, biochemical, inflammatory, and nutritional factors in determining the risk of ACE post-PCI in patients with AMI. The application of ANN and other ML techniques represents a promising frontier in cardiovascular medicine, offering the potential to refine risk stratification, personalize patient management, and ultimately improve clinical outcomes. Future research should focus on validating these models in diverse populations and exploring the integration of ML-based prognostication into clinical workflow.