Machine Learning-Based Model for Prediction of Hemorrhage Transformation in Acute Ischemic Stroke After Alteplase.

Hemorrhage transformation (HT) is the most dreaded complication of intravenous thrombolysis (IVT) in acute ischemic stroke (AIS). The prediction of HT after IVT is important in the treatment decision-making for AIS. We designed and compared different machine learning methods, capable of predicting HT in AIS after IVT. A total of 345 AIS patients who received intravenous alteplase between January 2016 and June 2021 were enrolled in this retrospective study. The demographic characteristics, clinical condition, biochemical data, and neuroimaging variables were included for analysis. HT was confirmed by head computed tomography (CT) or magnetic resonance imaging (MRI) within 48 h after IVT. Based on the neuroimaging results, all of the patients were divided into the non-HT group and the HT group. Then, the variables were applied in logistic regression (LR) and random forest (RF) algorithms to establish HT prediction models. To evaluate the accuracy of the machine learning models, the models were compared to several of the common scales used in clinics, including the multicenter stroke survey (MSS) score, safe implementation of treatments in stroke (SITS) score, and SEDAN score. The performance of these prediction models was evaluated using the receiver operating characteristic (ROC) curve (AUC). Forty-five patients had HT (13.0%) within 48 h after IVT. The ROC curve results showed that the AUCs of HT that were predicted by the RF model, LR model, MSS, SITS, and SEDAN scales after IVT were 0.795 (95% CI, 0.647-0.944), 0.703 (95% CI, 0.515-0.892), 0.657 (95% CI, 0.574-0.741), 0.660 (95% CI, 0.580-0.740) and 0.655 (95% CI, 0.571-0.739), respectively. The RF model performed better than the other models and scales. The top four most influential factors in the RF importance matrix plot were triglyceride, Lpa, the baseline NIHSS, and hemoglobin. The SHapley Additive exPlanation values made the RF prediction model clinically interpretable. In this study, an RF machine learning method was successfully established to predict HT in AIS patients after intravenous alteplase, which the sensitivity was 66.7%, and the specificity was 80.7%.

Introduction

Stroke is a major global public health issue with high morbidity and mortality. According to China's Ministry of Health, 77.8% of strokes are ischemic (1). Currently, poststroke outcomes are significantly improved by reperfusion therapy, such as intravenous thrombolysis and endovascular thrombectomy (2–6). Despite its efficacy, IVT also increases the risk of the development of hemorrage transformation (HT). The occurrence of HT significantly affects functional recovery and is independently associated with a higher mortality (7). It was shown in previous studies that a total of 10–48% of the AIS patients who received IVT developed HT (8) and the incidence of symptomatic intracranial hemorrhage (SICH) ranged from 2 to 7% (9). There is currently a lack of understanding of the benefits of IVT relative to its risks, which include HT, in both clinicians and patients (or their relatives). This results in patients not always receiving a timely intervention.

Previous reports have indicated that there are a number of clinical, laboratory, and radiographic factors associated with the risk of HT after alteplase. A systematic review and meta-analysis of 55 studies have reported on factors that are associated with HT, and these included age; the severity of the stroke; the baseline glucose; the use of antiplatelet drugs or statins; leukoaraiosis; early signs of infarction on head CT; the presence of atrial fibrillation, diabetes, previous ischemic heart disease or cerebral vascular diseases; and congestive heart failure (10). Lipoprotein (11), the neutrophil-to-lymphocyte ratio (NLR) (12) and the serum homocysteine level (13) were also reported as factors associated with HT after IVT. Numerous scoring systems have been devised to assess the risk for HT in AIS patients after IVT, and these include the Hemorrhage After Thrombolysis (HAT) score (14), multicenter stroke survey (MSS) score (15), safe implementation of treatments in stroke (SITS) score (16), SEDAN score (17) and GRASPS score (18). However, existing HT models cannot address the full complexity of all the factors that are involved. Each of the methods has limitations and disadvantages (14–18), which implies a certain degree of inaccuracy. The risk of HT after IVT in AIS patients cannot be avoided to the greatest extent when the stroke patients are only assessed based on the existing assessment system. Therefore, it is an urgent priority to find more reliable and effective methods to develop an early and timely prediction for HT after IVT in AIS patients.

Machine learning (ML) is a branch of artificial intelligence that can establish ideal models for classification, prediction, and estimation by allowing computers to “learn” from large, noisy or complex input and output datasets (19). In recent years, machine learning has been applied in medical research. For example, machine learning has been applied for the prediction of the outcomes in acute stroke patients (20) and endovascular treatment outcome in AIS (21). In the present study, we developed machine learning-based models that included a logistic regression (LR) model and a random forest (RF) model to predict HT. We then compared the predictability to several scales that are currently used in the clinic, including the MSS score, SITS score and SEDAN score.

Materials and Methods

Participants

AIS patients treated with rt-PA therapy after admission to Zhongda Hospital Affiliated to Southeast University between 1 January 2016 and 30 June 2021 were enrolled in this retrospective study. Inclusion criteria were as follows: (1) age ≥ 18 years, (2) AIS confirmed by magnetic resonance imaging (MRI), and (3) onset of stroke symptoms within 4.5 h and treated with rt-PA. Exclusion criteria were as follows: (1) additional endovascular therapy after intravenous thrombolysis, (2) no head computed tomography (CT) or MRI patients within 48 h after intravenous thrombolysis, (3) patients treated with urokinase thrombolytic therapy. This retrospective study was approved by the Research Ethics Committee of Affiliated ZhongDa Hospital and the Southeast University (approval number: 2021ZDSYLL300-P02).

Data Collection

The clinical data were collected by a certified stroke neurologist. Overall, 61 clinical variables were initially selected for the model construction.

The study collected demographic data including patient sex, age, body weight, past history (including hypertension, diabetes, coronary artery disease, atrial fibrillation, stroke, antiplatelet therapy, smoking, and alcohol consumption), clinical characteristics, including baseline indicators (the baseline NIHSS, the baseline blood pressure, early infarction signs seen on a head CT on admission and the baseline laboratory tests), the time from onset to thrombolytic therapy (OTT), the rt-PA dose (0.9 or 0.6 mg/kg), the NIHSS score (1 h after IVT), the NIHSS score (2 h after IVT), and the laboratory test results on the second day after IVT.

HT Scores

The SEDAN, MSS, and SITS scores were used for the measuring the HT scores, all of the patients were evaluated based on the scale, and they were evaluated after IVT and after having medications for 48 h. When the score was higher, there was a greater risk that the AIS patient would develop HT after IVT.

Machine-Learning Algorithms

To compare the machine-learning risk algorithms, the study population was split into a “training” cohort from which the HT risk algorithms were derived and a “validation” cohort in which the algorithms were applied and tested. The “training” cohort was derived from a random sampling of 80% of the AIS patients, and the “validation” cohort was comprised of the remaining 20% of the AIS patients. The machine learning models were trained with all of the variables as inputs to classify the patients who were likely to have HT after IVT for AIS. Two of the commonly used classes of machine-learning algorithms were utilized: logistic regression, and random forest. To evaluate the accuracy of the machine learning models, we calculated three of the common scales used in the clinic as references: the MSS score, SITS score and SEDAN score. An ROC (receiver operating characteristic) curve was used to evaluate the performance of the models.

The effects of the features on the RF prediction model were measured using the functions of the SHapley Additive exPlanations (SHAP) Python package (version 0.40.0), which assessed the importance of each feature using a game-theoretic approach (22).

Statistical Analysis

SPSS 22.0 was used for the descriptive analysis and the comparison of clinically defined groups. Continuous variables are expressed as means ± SD or as medians (interquartile range). Categorical variables are expressed as percentages. Continuous data were compared between the groups using t-test or a non-parametric test. Categorical data were compared using the χ2 test. P-values were corrected for multiple testing with the Benjamini-Hochberg false discovery rate (FDR) correction using the Statsmodels package in Python. In all cases, FDR-adjusted P < 0.05 was considered to indicate statistical significance.

Results

A total of 498 patients who received IVT with rt-PA within 4.5 h of stroke onset were included in the cohort during the study period. Ultimately, 345 patients were included (Figure 1), after excluding 20 patients with missing laboratory tests or clinical data, 81 patients who did not have acute infarct lesions that were found on DWI, 51 patients who underwent endovascular therapy, and 1 patient without a second scan. The study sample included 121 (35.1%) females and 224 (64.9%) males. Their ages ranged from 29 to 95 years with a median (IQR) of 70 years (63–81). Forty-five (13.0%) patients developed HT, and ten (2.9%) patients developed SICH (defined as a clinical deterioration expressed by an increase of at least 4 points on the NIHSS scale). The patients who developed HT were older (FDR p-value: 0.026), had higher baseline NIHSS scores (FDR p-value: 0.009), NIHSS scores 1 h after IVT (FDR p-value: 0.009), NIHSS scores 2 h after IVT (FDR p-value: 0.014), PT values (FDR p-value: 0.014), and INR values (FDR p-value: 0.043), and they had lower RBC counts (FDR p-value: 0.009), hemoglobin (FDR p-value: 0.009), lymphocytes (FDR p-value: 0.029), and triglycerides after IVT (FDR p-value: 0.009). Also, more of patients who developed HT had cardioembolic events (FDR p-value: 0.044) compared to the non-HT group. However, there were less HT patients that had small vessel occlusion (FDR p-value: 0.009) compared to the non-HT group. The demographic and clinical characteristics of this study population are shown in Table 1, Figure 2.

Figure 1

Flow chart illustrating patient selection. IVT indicates introvenous thrombolysis. HT indicates hemorrhage transformation. IVT, intravenous thrombolysis; HT, hemorrhage transformation.

Table 1

Demographic and clinical characteristics of the study sample.

Parameter	HT (n = 45)	No-HT (n = 300)	t/χ2/z value	P-value	FDR-adjusted P-value
Age, years	78 (68, 82)	70 (63, 80)	−2.847	0.004	0.026
Body weight, kg	60 (54, 70)	66 (55, 72)	−1.424	0.154	0.282
Sex
Male	24 (53.3%)	200 (66.7%)	3.055	0.080	0.190
Past history
Smoking	10 (22.2%)	115 (38.3%)	4.396	0.036	0.121
Alcohol consumption	5 (11.1%)	62 (20.7%)	2.283	0.131	0.273
Comorbidities
Hypertension, n (%)	29 (64.4%)	210 (70.0%)	0.567	0.451	0.656
Diabetes, n (%)	10 (22.2%)	81 (27.0%)	0.460	0.498	0.694
Coronary artery disease, n (%)	7 (15.6%)	39 (13.0%)	0.221	0.638	0.772
Atrial fibrillation, n (%)	15 (33.3%)	58 (19.3%)	4.597	0.032	0.114
Previous stroke, n (%)	13 (28.9%)	75 (25.0%)	0.311	0.577	0.754
PAT	6 (13.3%)	34 (11.3%)	0.153	0.696	0.810
Statin use	3 (6.7%)	13 (4.3%)	0.099	0.754	0.815
OTT	180 (120, 238)	165 (120, 200)	−1.147	0.252	0.428
SBP on admission	159 (148, 175)	152 (139, 164)	−2.302	0.021	0.084
DBP on admission	84.9 ± 14.1	85.5 ± 14.4	0.267	0.789	0.828
Rt-pa dose			0.457	0.499	0.694
0.9 mg/kg	33 (73.3%)	205 (68.3%)
0.6 mg/kg	22 (26.7%)	95 (31.7%)
Early infarction signs on admission with head CT	5 (11.1%)	25 (8.3%)	0.111	0.739	0.815
Circulation, n (%)			4.441	0.109	0.241
Anterior circulation infarct	38 (84.4%)	236 (78.7%)
Posterior circulation infarct	3 (6.7%)	50 (16.7%)
Both	4 (8.9%)	14 (4.7%)
Leukoaraiosis	35 (77.8%)	201 (67.0%)	2.103	0.147	0.277
Stroke subtype, n (%)			14.185	0.001	0.009
Cardioembolic	13 (28.9%)	41 (13.7%)	6.868	0.009	0.044
Large artery atherosclerosis	20 (44.4%)	93 (31.0%)	3.211	0.073	0.180
Small vessel occlusion	12 (26.7%)	166 (55.3%)	12.876	<0.001	0.009
Baseline NIHSS score	11 (6, 17)	7 (4, 12)	−3.211	0.001	0.009
NIHSS score (1 h after IVT)	8 (4, 17)	4 (2, 10)	−3.257	0.001	0.009
NIHSS score (2 h after IVT)	8 (4, 15)	4 (2, 10)	−3.150	0.002	0.014
Baseline laboratory test
WBC, 109/l	6.55 (4.95, 9.11)	7.36 (6.11, 8.84)	−1.852	0.064	0.169
RBC, 109/l	4.36 (3.98, 4.77)	4.69 (4.37, 5.02)	−3.221	0.001	0.009
Hemoglobin, g/l	133 (124, 145)	143 (132, 153)	−3.466	0.001	0.009
Platelet, 109/l	174 (144, 221)	193 (159, 241)	−2.012	0.044	0.141
Neutrophil, 109/l	4.54 (3.05, 6.63)	4.51 (3.53, 6.16)	−0.370	0.711	0.813
Lymphocyte, 109/l	1.45 (1.11, 2.24)	2.05 (1.37, 2.57)	−2.799	0.005	0.029
NLR	2.76 (1.73, 5.02)	2.29 (1.53, 3.82)	−1.466	0.143	0.277
PT, s	11.85 (11.28, 12.65)	11.4 (10.8, 12.0)	−3.137	0.002	0.014
APTT, s	28.91 ± 4.54	28.23 ± 3.86	−1.029	0.304	0.463
Fibrinogen, g/l	3.46 ± 0.80	3.32 ± 0.86	−1.039	0.299	0.463
INR	1.07 (1.03, 1.14)	1.05 (0.99, 1.10)	−2.667	0.008	0.043
Blood glucose, mmol/l	7.91 (6.48, 10.93)	7.13 (6.08, 8.87)	−1.875	0.061	0.169
AKP, u/l	85.0 (69.5, 96.3)	80.5 (67.0, 97.0)	−0.506	0.613	0.772
Albumin, g/l	41.9 (39.0, 45.5)	42.1 (39.2, 44.9)	−0.072	0.942	0.942
Blood urea nitrogen, mmol/l	6.48 (5.63, 7.80)	6.20 (5.00, 7.40)	−1.714	0.086	0.197
Creatinine, μmol/l	75 (67, 88)	73 (62, 87)	−1.040	0.298	0.463
Uric acid, μmol/l	326 (264, 402)	342.0 (280.0, 411.0)	−0.588	0.557	0.754
Potassium, mmol/l	3.77 ± 0.52	3.81 ± 0.45	0.470	0.639	0.772
Sodium, mmol/l	140.6 (138.2, 143.7)	140 (138, 142)	−1.505	0.132	0.273
Chloride, mmol/l	103.6 (101.0, 105.9)	103.8 (101.4, 105.7)	−0.090	0.929	0.942
Calcium, mmol/l	2.27 (2.18, 2.42)	2.29 (2.22, 2.39)	−0.300	0.764	0.815
Phosphorus, mmol/l	1.08 (0.96, 1.20)	1.08 (0.94, 1.19)	−0.482	0.630	0.772
C–reactive protein, mg/l	7.40 (3.47, 19.10)	3.94 (0.82, 8.84)	−2.438	0.015	0.069
Laboratory test on the second day after IVT
Neutrophil after IVT, 109/l	5.52 (4.73, 9.48)	5.79 (4.25, 7.16)	−1.141	0.254	0.428
Lymphocyte after IVT, 109/l	1.19 (0.86, 1.69)	1.36 (0.94, 1.88)	−1.452	0.146	0.277
NLR after IVT	4.61 (2.67, 8.78)	3.83 (2.39, 7.02)	−0.802	0.423	0.630
Potassium after IVT, mmol/l	3.65 (3.33, 3.89)	3.74 (3.58, 3.97)	−2.207	0.027	0.102
Sodium after IVT, mmol/l	139.5 (138.2,141.1)	139.6 (138.0,141.1)	−0.395	0.693	0.810
Chloride after IVT, mmol/l	104.1 (101.6, 105.9)	105.2 (102.7, 107.0)	−1.836	0.066	0.169
Calcium after IVT, mmol/l	2.21 ± 0.13	2.17 ± 0.10	−1.921	0.061	0.169
Phosphorus after IVT, mmol/l	1.01 (0.81, 1.12)	1.01 (0.87, 1.12)	−0.563	0.573	0.754
Tg, mmol/l	0.88 (0.69, 1.25)	1.26 (0.87, 1.83)	−3.666	<0.001	0.009
CHOL, mmol/l	4.32 ± 1.23	4.54 ± 1.04	1.260	0.208	0.370
HDL, mmol/l	1.24 (1.00, 1.48)	1.21 (1.02, 1.39)	−0.313	0.754	0.815
LDL, mmol/l	2.58 (1.97, 3.10)	2.71 (2.20,3.26)	−1.095	0.273	0.448
ApoA−1, g/l	1.15 (0.98, 1.30)	1.10 (0.97, 1.31)	−0.074	0.941	0.942
ApoB, g/l	0.80 (0.60, 0.94)	0.84 (0.70, 1.00)	−1.842	0.065	0.169
Lpa, mg/l	237.5 (130.5, 357.5)	148.0 (69.8, 285.0)	−2.305	0.021	0.084

PAT, previous antiplatelet therapy; SBP, systolic blood pressure; DBP, diastolic blood pressure; NLR, neutrophil-to-lymphocyte ratio; OTT, onset-to-treatment time; WBC, white blood cell; RBC, red blood cell; AKP, alkaline phosphatase; BUN, blood urea nitrogen; HDL, high-density lipoprotein cholesterol; LDL, low-density lipoprotein cholesterol, Tg, triglycerides; ApoA-1, Apolipoprotein A 1; ApoB, Apolipoprotein B; Lpa, lipoprotein (a); CHOL, cholesterol; Significant difference between the 2 groups (FDR-adjusted P-value < 0.05).

Figure 2

Comparison of Tg, Lpa, baseline NIHSS score, hemoglobin between patients in the HT and no-HT group. Tg, triglycerides; Lpa, lipoprotein(a); HT, hemorrhage transformation.

Comparison of the Models for the Prediction of Hemorrhage Transformation in Acute Ischemic Stroke After Intravenous Alteplase

The ROC curve results showed that the AUCs of HT, as predicted by the RF model, LR model, MSS, SITS, and SEDAN scores after IVT in AIS patients were 0.795 (95% CI, 0.647–0.944), 0.703 (95% CI, 0.515–0.892), 0.657 (95% CI, 0.574–0.741), 0.660 (95% CI, 0.580–0.740) and 0.655 (95% CI, 0.571–0.739), respectively. When the optimal cutoff value of the RF model was 0.169, the sensitivity was 66.7%, and the specificity was 80.7%. When the optimal cutoff value of the LR model was 0.140, the sensitivity was 60.0%, and the specificity was 78.0%. The RF model performed significantly better than the other models. The AUCs are presented in Figure 3. The importance matrix plot for the RF method is shown in Figure 4 and reveals that the top 4 parameters with the greatest importance for predicting HT of AIS after IVT in this model with all available parameters were triglyceride, Lpa, the baseline NIHSS and hemoglobin.

Figure 3

Area under the receiver operating characteristics curves (AUC) for predicting HT in AIS after Intravenous Alteplase. HT, hemorrhage transformation; AIS, acute ischemic stroke.

Figure 4

Importance matrix plot of the RF model. Top 15 parameters for predicting HT of AIS after IVT. RF, random forest; HT, hemorrhage transformation; AIS, acute ischemic stroke; IVT, intravenous thrombolysis.

SHAP Value-Based Interpretation of the RF Model

A SHAP summary plot was used to illustrate the positive or negative effects of the features attributed to the RF model (Figure 5). The SHAP dependence plot showed how a single feature of the top 4 factors affected the output of the RF prediction model (Figure 6).

Figure 5

SHAP summary plot of the top 15 features of the RF model. The higher the SHAP value of a feature (x-axis), the higher the probability of HT after IVT. Taking the feature Lpa as an example, red points are on the right whereas blue points are on the left. This means prediction scores will be smaller when patients have a low level of Lpa. SHAP, SHapley Additive exPlanation; RF, random forest; HT, hemorrhage transformation; IVT, intravenous thrombolysis; Tg, triglycerides; WBC, white blood cell; RBC, red blood cell; NLR, neutrophil-to-lymphocyte ratio.

Figure 6

SHAP dependence plot of the RF model. The SHAP dependence plot shows how a single feature affects the output of the RF prediction model. SHAP values for specific features exceed zero, representing an increased risk of HT after IVT. SHAP, SHapley Additive exPlanation; RF, random forest; HT, hemorrhage transformation; IVT, intravenous thrombolysis; Tg, triglycerides; Lpa, lipoprotein(a).

Discussion

In this retrospective cohort study, we developed and validated machine learning algorithms using 61 features to predict HT after IVT for AIS. The results demonstrated that the use of machine learning models can accurately predict HT. The RF model performed better than the LR model in predicting HT.

Numerous factors affect, with varying significance and mechanisms, the development of HT after IVT in AIS patients. To avoid HT as much as possible, it is important to consider as many influential factors as possible before proceeding with IVT. Unfortunately, the scales that are commonly used in the clinic, such as the MSS score (15), SITS score (16), and SEDAN score (17), are only calculated based on a few important variables. However, machine learning could provide a multi-characteristic analysis in deciphering the factors that are associated with the risk of HT after IVT.

Many risk factors have been confirmed to be associated with HT after IVT. A systematic review and meta-analysis of 55 studies reported that age, severity of the stroke, baseline glucose, use of antiplatelet drugs or statins, leukoaraiosis, early signs of infarction on head CT, and the presence of atrial fibrillation, diabetes, previous ischemic heart disease or cerebral vascular diseases, and congestive heart failure were associated with HT after IVT (10). In addiction, hypertension (23–25) and the NLR (12, 26–28) were also shown to be associated with HT after IVT. The results of the present study confirmed the results of previous reports. Meanwhile, the baseline NIHSS, systolic blood pressure on admission, and the NLR were ranked the top six features in the present study, and the higher their values were, the greater the risk of HT.

The NIHSS score is widely used to assess the clinical stroke severity. It has been suggested that the baseline NIHSS score is an independent risk factor for HT after IVT and that a high NIHSS score is a predictor of HT (25, 29). Moreover, it was reported that 1.6% of IVT patients had fatal HT when their NIHSS score was 5 to 10, while it increased to 6.8% when their NIHSS score was ≥ 22 (9).

The patient's blood pressure (24), especially systolic blood pressure before IVT (23, 25) was also previously shown to be a risk factor for HT. This result was consistent with the present study. This may be because if patients have higher blood pressure, there is more damage to the vascular endothelial cells.

A high NLR was reported to predict HT in AIS patients after IVT in previous studies (12, 26–28). The NLR on admission rather than the NLR post-IVT was an independent risk factor for an increased risk of HT after IVT (28). The results in this study were the same as the results of the above studies. The underlying mechanism of HT remains uncertain. It has been suggested that the NLR is associated with the inflammatory destruction of neutrophils and the protective effect of lymphocytes. Neutrophils plays a role in the disruption of the blood brain barrier (BBB) through inflammation (30, 31) and enhance the expression of matrix metalloproteinase-9 (MMP-9) (32). Rt-PA can not only promote neutrophils to release MMPs but also promote the migration of neutrophils to ischemic tissue (33). Moreover, the stress response leads to an increase in the production of catecholamines in an overactivated sympathetic nervous system, resulting in a decrease in the number and activity of lymphocytes (34).

Our study also obtained some new findings. The factors that were ranked in the top four factors in the present study were triglyceride, Lpa, the baseline NIHSS and hemoglobin, among which, triglyceride, Lpa, and hemoglobin were negelected by commonly used scales in clinic currently.

Several studies have shown that a high level of triglycerides contributes to atherosclerosis (35–37), while a low level is a risk factor for cerebral hemorrhage (38). It was also reported that a low level of triglycerides was a modest risk factor for HT after IVT (11). In the RF model-based machine learning, triglycerides were negatively associated with HT after IVT. This is probably because a high level of triglycerides is correlated with elevations in coagulation factor VII and plasminogen activator inhibitor, and the viscosity of blood and plasma (39).

The results from previous reports on the association between Lpa and stroke are controversial. Several studies have shown that there is no evidence of an association between the Lpa levels and stroke (35, 40). However, a meta-analysis suggested that Lpa was a significant risk factor for ischemic stroke (41). In the present study, we reveal that Lpa as a positive factor that increases HT after IVT, which is a novel finding. Therefore, more detailed studies on the relationship between lipids and the risk of AIS and HT after IVT may be important and informative.

It has been demonstrated that lower hemoglobin levels on admission are associated with hematoma expansion (HE) after intracranial hemorrhage and worse clinical outcomes (42). As indicated by the present study, it was also implicated to be associated with the development of HT after IVT. This may be the result of coagulopathies and prolonged bleeding in anemic patients (43).

There are several limitations in the present study. First, this was a single-center retrospective study, and this study requires validation with larger datasets from other sources. Additionally, the data was too small to further analysis the prediction of SICH. Second, the algorithm was built from the input features, and some hidden relationships may have been ignored because unknown or neglected features were not evaluated by physicians. Third, the patients did not have susceptibility-weighted imaging (SWI), which might have led to an underestimation of HT.

Conclusion

This study demonstrated that machine learning algorithms, especially the RF model, can improve HT prediction for ischemic stroke patients after intravenous alteplase. It may aid clinicians, as well as patients and families, in the process of decision-making when determining the AIS patient's eligibility for thrombolytic treatment.

Data Availability Statement

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

Ethics Statement

The studies involving human participants were reviewed and approved by the Research Ethics Committee of Affiliated ZhongDa Hospital and the Southeast University. Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

Author Contributions

YX and XL conceived and designed the research. ZZ and AJ provided technical assistance. YX and DW performed the experiments and plotted the figures. YX wrote the manuscript. XL checked the manuscript and made final modifications. All authors have read and approved the submitted version.

Funding

This study was supported by the Nanjing Medical Science and Technique Development Foundation (No. YKK21217).

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher's Note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.