Predicting Hematoma Expansion after Spontaneous Intracranial Hemorrhage Through a Radiomics Based Model

Abstract

Purpose: Intracranial hemorrhage (ICH) is characterized as bleeding into the brain tissue, intracranial space, and ventricles and is the second most disabling form of stroke. Hematoma expansion (HE) following ICH has been correlated with significant neurological decline and death. For early detection of patients at risk, deep learning prediction models were developed to predict whether hematoma due to ICH will expand. This study aimed to explore the feasibility of HE prediction using a radiomic approach to help clinicians better stratify HE patients and tailor intensive therapies timely and effectively.

Materials and methods: Two hundred ICH patients with known hematoma evolution, were enrolled in this study. An open-source python package was utilized for the extraction of radiomic features from both non-contrast computed tomography (NCCT) and magnetic resonance imaging (MRI) scans through characterization algorithms. A total of 99 radiomic features were extracted and different features were selected for network inputs for the NCCT and MR models. Seven supervised classifiers: Support Vector Machine, Naïve Bayes, Decision Tree, Random Forest, Logistic Regression, K-Nearest Neighbor and Multilayer Perceptron were used to build the models. A training:testing split of 80:20 and 20 iterations of Monte Carlo cross validation were performed to prevent overfitting and assess the variability of the networks, respectively. The models were fed training datasets from which they learned to classify the data based on pre-determined radiomic categories.

Results: The highest sensitivity among the NCCT classifier models was seen with the support vector machine (SVM) and logistic regression (LR) of 72 ± 0.3% and 73 ± 0.5%, respectively. The MRI classifier models had the highest sensitivity of 68 ± 0.5% and 72 ± 0.5% for the SVM and LR models, respectively.

Conclusions: This study indicates that the NCCT radiomics model is a better predictor of HE and that SVM and LR classifiers are better predictors of HE due to their more cautious approach indicated by a higher sensitivity metric.

Keywords: Artificial Intelligence; Brain; Hematoma Expansion; Hemorrhagic Stroke; Non-contrast Computed Tomography.

Figure 1:

The 200 Intracranial hemorrhage patients used for this retrospective study all had one of the following hemorrhage types: intraparenchymal (n=181), intraventricular (n=45), subdural (n=13), or subarachnoid (n=19) (Seymour, 2021).

Figure 2:

NCCT slice with bounding box around hemorrhagic region on the left image and automatic segmentation of hematoma in the right image (Seymour, 2021).

Figure 3:

Represents an MRI central slice on the left and its corresponding binary hemorrhage segmentation label (Seymour, 2021).

Figure 4:

Demonstrates a diagram of the HE study that has an ICH cohort of 200, with 70 HE patients and 130 non-HE cases. Image segmentation of the hemorrhagic area was conducted along with feature extraction using the Pyradiomics package to extract radiomic features. A training:testing split of 80:20, class weighting of 0.77:1.43 for non-HE:HE, and 20 iterations of Monte Carlo cross validation was performed for each classification algorithm (Seymour, 2021).

Figure 5:

A comparison of all six ROC curves for each of the following classifiers for the NCCT model: SVM, NB, RF, DT, LR, and MLP (Seymour, 2021).

Figure 6:

A comparison of all six ROC curves for each of the following classifiers for the MRI radiomics model: SVM, NB, RF, DT, LR, and KNN (Seymour, 2021).