A quantitative symmetry‐based analysis of hyperacute ischemic stroke lesions in noncontrast computed tomography

Abstract

Purpose Early identification of ischemic stroke plays a significant role in treatment and potential recovery of damaged brain tissue. In non-contrast CT (ncCT), the differences between ischemic changes and healthy tissue are usually very subtle during the hyper-acute phase (<8 hours from the stroke onset). Therefore, visual comparison of both hemispheres is an important step in clinical assessment. A quantitative symmetry-based analysis of texture features of ischemic lesions in non-contrast CT images may provide an important information for differentiation of ischemic and healthy brain tissue in this phase. Methods One hundred thirty-nine (139) ncCT scans of hyperacute ischemic stroke with follow-up magnetic resonance diffusion-weighted (MR-DW) images were collected. The regions of stroke were identified in the MR-DW images, which were spatially aligned to corresponding ncCT images. A state-of-the-art symmetric diffeomorphic image registration was utilized for the alignment of CT and MR-DW, for identification of individual brain hemispheres, and for localization of the region representing healthy tissue contralateral to the stroke cores. Texture analysis included extraction and classification of co-occurrence and run length texture-based image features in the regions of ischemic stroke and their contralateral regions. Results The classification schemes achieved area under the receiver operating characteristic [Az] ≈ 0.82 for the whole dataset. There was no statistically significant difference in the performance of classifiers for the data sets with time between 2 and 8 hours from symptom onset. The performance of the classifiers did not depend on the size of the stroke regions. Conclusions The results provide a set of optimal texture features which are suitable for distinguishing between hyperacute ischemic lesions and their corresponding contralateral brain tissue in non-contrast CT. This work is an initial step towards development of an automated decision support system for detection of hyperacute ischemic stroke lesions on non-contrast CT of the brain.

3D convolutional neural networks applied to CT angiography in the detection of acute ischemic stroke

“…Gillebert et al [28] found a slightly higher DSC (0.67) than the present study (0.61 at optimal threshold) for simulated data but reported high variance for performance using actual patient images. Peter et al [29] reported a ROC AUC of 0.82 which was less than that of the present study. At least one infarct detection software has been made commercially available [30].…”

“…Sensitivity calculated from Maldjian et al reported values for lentiform nucleus and insula (0.85) lower than the sensitivity calculated over the ASPECTS regions in the present study (0.93). Gillebert et al [28] as well as Peter et al [29] applied a texture and atlas-based approach to lesion segmentation. Gillebert et al [28] found a slightly higher DSC (0.67) than the present study (0.61 at optimal threshold) for simulated data but reported high variance for performance using actual patient images.…”

“…Moreover, NCCT alone or CTA + NCCT without hemispheric comparison were not investigated. Data processing could be further improved, by including atlas-based anatomical regions as described by Peter et al [29] or by introducing additional features such as the requirement of hypoattenuation to extend onto the cortical surface. In addition, finding an optimum in the CNN parameter space, exploring alternative algorithms and implementations aimed at improving specificity and reducing processing time is essential when targeting the clinical deployment.…”

Radiomics in Stroke Neuroimaging: Techniques, Applications, and Challenges

“…They found that the gray-level co-occurrence matrix (GLCM)-based tissue texture parameters were significantly different between patients and controls, and the most discriminative feature was angular second moment. In another study of 139 patients with hyperacute ischemic stroke (< 8 h), the authors identified 6 texture features from NCCT images that could differentiate ischemic lesions from their contralateral normal tissue [ 57 ]. The classification model, constructed by 3 supervised machine learning algorithms (i.e., support vector machine, decision trees, and adaboost) achieved an average AUC of 0.82.…”

“…While NCCT-based radiomic features have the ability to identify acute infarcted tissue, the information they capture is confined to a part of the lesion in a single slice (2D-radiomics) [ 56 , 57 ]. It is challenging to perform a volumetric assessment for stroke lesions that cannot be visually perceived.…”