DGA3-Net: A parameter-efficient deep learning model for ASPECTS assessment for acute ischemic stroke using non-contrast computed tomography

Detecting the early signs of stroke using non-contrast computerized tomography (NCCT) is essential for the diagnosis of acute ischemic stroke (AIS). However, the hypoattenuation in NCCT is difficult to precisely identify, and accurate assessments of the Alberta Stroke Program Early CT Score (ASPECTS) are usually time-consuming and require experienced neuroradiologists. To this end, this study proposes DGA3-Net, a convolutional neural network (CNN)-based model for ASPECTS assessment via detecting early ischemic changes in ASPECTS regions. DGA3-Net is based on a novel parameter-efficient dihedral group CNN encoder to exploit the rotation and reflection symmetry of convolution kernels. The bounding volume of each ASPECTS region is extracted from the encoded feature, and an attention-guided slice aggregation module is used to aggregate features from all slices. An asymmetry-aware classifier is then used to predict stroke presence via comparison between ASPECTS regions from the left and right hemispheres. Pre-treatment NCCTs of suspected AIS patients were collected retrospectively, which consists of a primary dataset (n = 170) and an external validation dataset (n = 90), with expert consensus ASPECTS readings as ground truth. DGA3-Net outperformed two expert neuroradiologists in regional stroke identification (F1 = 0.69) and ASPECTS evaluation (Cohen's weighted Kappa = 0.70). Our ablation study also validated the efficacy of the proposed model design. In addition, class-relevant areas highlighted by visualization techniques corresponded highly with various well-established qualitative imaging signs, further validating the learned representation. This study demonstrates the potential of deep learning techniques for timely and accurate AIS diagnosis from NCCT, which could substantially improve the quality of treatment for AIS patients.

Comparing the benefit of ASPECTS on maximum intensity projection images of computed tomography angiography to source images and noncontract computed tomography in predicting infarct volume and collaterals extent

INTRODUCTION

In acute ischemic strokes (AIS), the Alberta Stroke Program Early Computed Tomography Score (ASPECTS) and CT perfusion (CTP) are commonly used to determine mechanical thrombectomy eligibility. Prior work suggests that CTA source image (CTA-SI) ASPECTS (CTA_asp) and a newly described CTA maximum intensity projection (CTA-MIP) ASPECTS (MIP_asp) better predict the final infarct core. Our goal was to compare MIP_asp to CTA_asp and non-contrast CT ASPECTS (NCCT_asp) for predicting ischemic core and collaterals established by CTP.

METHODS AND MATERIALS

A single institution, retrospective database for AIS due to internal carotid artery (ICA) or proximal middle cerebral artery (MCA) occlusions between January 2016 and February 2021 was reviewed. We rated ASPECTS on NCCT, CTA-SI, and CTA-MIP at baseline, then used the automated RAPID software to measure CTP ischemic core volume. The accuracy of each ASPECTS in predicting ischemic core volume (ICV) >70 cc and Hypoperfusion intensity ratio (HIR) >0.4 was compared using the receiver operating characteristic (ROC) curve.

RESULTS

122/319 patients fulfilled the inclusion criteria. Area under the curve (AUC) for MIP_asp was significantly higher than NCCT_asp and CTA_asp for predicting ICV >70 cc (0.95 vs. 0.89 and 0.95 vs. 0.92, P =0.03 and P = 0.04). For predicting HIR >0.4, AUC for MIP_asp was significantly higher than NCCT_asp and CTA_asp (0.85 vs. 0.72 and 0.85 vs. 0.81, P < 0.001 and P < 0.01).

CONCLUSION

The predictive accuracy of detecting ischemic stroke with ICV >70cc and HIR >0.4 can be significantly improved using the MIP_asp instead of CTA_asp or NCCT_asp.