An automatic deep-learning approach for the prediction of post-stroke epilepsy after an initial intracerebral hemorrhage based on non-contrast computed tomography imaging

Abstract

Background: Post-stroke epilepsy (PSE) is a common and significant complication that often occurs after stroke, and affects patients' prognosis and overall quality of life. In recent years, non-contrast computed tomography (NCCT) has become the preferred method for the clinical diagnosis of intracerebral hemorrhage (ICH). This study aimed to develop and validate a triple deep-learning model, simply named, the post-stroke epilepsy network (PSENet), to predict PSE in ICH patients based on NCCT.

Methods: A total of 1,130 patients (62 with PSE and 1,068 without PSE) who experienced an initial ICH at our hospital were enrolled in this study. Using five-fold cross-validation, all patients were randomly divided into training and validation sets at a ratio of 4:1. Next, the no-new-Net (nnU-Net) was used to automatically segment the ICH for the subsequent quantitative analysis. A triple deep-learning model was developed to extract the PSE-related features and incorporate the deep-learning features related to cortical involvement (FCI) and ICH volume to predict PSE. This model was compared with three clinical models constructed using random forest. Model performance was mainly evaluated using the area under the curve (AUC).

Results: The nnU-Net had a high Dice score of 0.923. The proposed PSENet, which incorporated multiple features, showed excellent diagnostic performance, and had an accuracy of 0.876, a F1-score of 0.621, a recall of 0.716, a specificity of 0.897, and an AUC of 0.840, which significantly surpassed the AUC of the baseline clinical model (AUC =0.787).

Conclusions: Based on our findings, the developed PSENet could be used to predict PSE quickly after the first ICH, especially in scenarios in which reliable clinical information is lacking on admission.

Keywords: Intracerebral hemorrhage (ICH); deep learning; post-stroke epilepsy (PSE); prediction.

Figure 1

Patient recruitment flowchart. ICH, intracerebral hemorrhage; IVH, intraventricular hemorrhage; ROI, region of interest; CT, computed tomography; PSE, post-stroke epilepsy.

Figure 2

The flow chart of the experiment. (A) In the image preprocessing and quantitative analysis stage, the segmentation mask was first obtained, and the quantitative analysis was then conducted. (B) In our proposed end-to-end PSENet for PSE prediction, two weak image enhancements and one strong image enhancement were performed. A consistency loss was implemented to ensure the similarity of the FPSE, and cross-entropy loss was applied separately on both augmentations. Further, the fusion of multiple extracted features created a comprehensive feature space, improving the performance of the PSENet. NCCT, non-contrast computed tomography; CT_BET, computed tomography_brain extraction tool; nnU-Net, no-new-Net; SVR-LSM, support vector regression lesion-symptom mapping; ICH, intracerebral hemorrhage; ROI, region of interest; FPSE, features related to post-stroke epilepsy; FCI, features related to cortical involvement; FPSE + CI, features related to post-stroke epilepsy and cortical involvement; FPSE + CI + IV, features related to post-stroke epilepsy, cortical involvement and intracerebral hemorrhage volume; FIV, features related to intracerebral hemorrhage volume; PSE, post-stroke epilepsy; PSENet, post-stroke epilepsy network.

Figure 3

Visualization of five patient examples. From top to bottom, the NCCT image, ground truth, and segmentation results. NCCT, non-contrast computed tomography.

Figure 4

ROC curves for the different models. (A) PSENet compared with different deep-learning models. (B) AUCs and DeLong test of PSENet with different preprocessed data. (C) PSENet compared with different clinical models. (D) DeLong test results of the PSENet and clinical models. C3D, convolutional 3D; AUC, area under the curve; ViT, vision transformer; PSE, post-stroke epilepsy; PSENet, post-stroke epilepsy network; LANE, L = lobar hemorrhage, A = age <65 years, N = National Institute of Health Stroke Scale score ≥15, E = early seizures; CAVE, C = cortical involvement, A = age <65 years, V = ICH volume >10 mL, E = early seizures; ROC, receiver operating characteristic.

Figure 5

Confusion matrices for the (A) PSENet, (B) baseline clinical model, (C) CAVE score model, and (D) LANE score model. PSE, post-stroke epilepsy; PSENet, post-stroke epilepsy network; CAVE, C = cortical involvement, A = age <65 years, V = ICH volume >10 mL, E = early seizures; LANE, L = lobar hemorrhage, A = age <65 years, N = National Institute of Health Stroke Scale score ≥15, E = early seizures.

Figure 6

Mean prediction probability histogram of different models. (A) PSENet. (B) Baseline clinical model. (C) CAVE score model. (D) LANE score model. The green dashed line represents the classification threshold (50%). True positive means PSE, and true negative means non-PSE. PSE, post-stroke epilepsy; PSENet, post-stroke epilepsy network; CAVE, C = cortical involvement, A = age <65 years, V = ICH volume >10 mL, E = early seizures; LANE, L = lobar hemorrhage, A = age <65 years, N = National Institute of Health Stroke Scale score ≥15, E = early seizures.

Figure 7

Decision curve analysis for the PSENet and the baseline clinical model. The net benefit of the PSENet is higher than the baseline clinical model, and the all/no intervention strategy. The threshold probability is displayed on the x-axis, while the net benefit is shown on the y-axis. The “None” curve represents no intervention, with all individuals being negative and a net benefit rate of 0, presented as a horizontal line. The “All” curve represents the all intervention, with all individuals being positive, and the net benefit rate decreasing with increasing intervention, showing a negative slope. PSE, post-stroke epilepsy; PSENet, post-stroke epilepsy network.

Figure 8

Grad-Cam heatmap of the PSENet and Resnet34. From top to bottom, the cropped ROI, the visualization results of the ResNet34, and the visualization results of the PSENet. The red region represents a larger weight, which can be decoded by the color bar on the right. ROI, region of interest; PSE, post-stroke epilepsy; PSENet, post-stroke epilepsy network; Grad-Cam, gradient-weighted class activation mapping.

Figure 9

Feature importance of the baseline clinical model for the prediction of PSE. ICH, intracerebral hemorrhage, NlHSS, National Institute of Health Stroke Scale; PSE, post-stroke epilepsy.