Imaging of Acute Ischemic Stroke

Magnetic resonance (MR) imaging (MRI) is commonly used to diagnose, assess and monitor stroke. Accurate and timely segmentation of stroke lesions provides the anatomico-structural information that can aid physicians in predicting prognosis, as well as in decision making and triaging for various rehabilitation strategies. To segment stroke lesions, MR protocols, including diffusion-weighted imaging (DWI) and T2-weighted fluid attenuated inversion recovery (FLAIR) are often utilized. These imaging sequences are usually acquired with different spatial resolutions due to time constraints. Within the same image, voxels may be anisotropic, with reduced resolution along slice direction for diffusion scans in particular. In this study, we evaluate the ability of 2D and 3D U-Net Convolutional Neural Network (CNN) architectures to segment ischemic stroke lesions using single contrast (DWI) and dual contrast images (T2w FLAIR and DWI). The predicted segmentations correlate with post-stroke motor outcome measured by the National Institutes of Health Stroke Scale (NIHSS) and Fugl-Meyer Upper Extremity (FM-UE) index based on the lesion loads overlapping the corticospinal tracts (CST), which is a neural substrate for motor movement and function. Although the four methods performed similarly, the 2D multimodal U-Net achieved the best results with a mean Dice of 0.737 (95% CI: 0.705, 0.769) and a relatively high correlation between the weighted lesion load and the NIHSS scores (both at baseline and at 90 days). A monotonically constrained quintic polynomial regression yielded R² = 0.784 and 0.875 for weighted lesion load versus baseline and 90-Days NIHSS respectively, and better corrected Akaike information criterion (AIC_c) scores than those of the linear regression. In addition, using the quintic polynomial regression model to regress the weighted lesion load to the 90-Days FM-UE score results in an R² of 0.570 with a better AIC_c score than that of the linear regression. Our results suggest that the multi-contrast information enhanced the accuracy of the segmentation and the prediction accuracy for upper extremity motor outcomes. Expanding the training dataset to include different types of stroke lesions and more data points will help add a temporal longitudinal aspect and increase the accuracy. Furthermore, adding patient-specific data may improve the inference about the relationship between imaging metrics and functional outcomes.

Imaging in stroke, allows its classification into ischaemic stroke (IS) or intracranial haemorrhagic stroke (ICH), ensuring time-sensitive treatment to be administered. Imaging can also allow detection of cerebral microbleeds (CMBs), which may further determine pharmacological intervention in acute stroke. True gradient echo (T2∗GRE) or susceptibility weighted imaging (SWI) have high sensitivity for the detection of CMBs. These two sequences are included in the national guidelines; however, the implementation of these guidelines can vary depending on local interpretation and scanner capabilities.

To explore the use and application of blood sensitive MRI sequences in a specialist UK stroke unit for the detection of CMBs, to improve local practice.

A retrospective data analysis of the native database, spanning a 6-month period, was used. The data of 281 acute stroke patients with an MRI were reviewed and analysed. The MRI sequences applied, and the final diagnosis were noted for each case.

Of the 281 acute stroke patients with MRI, 259 (92.1%) had an IS, 16 (5.68%) an ICH and 6 (2.14%) had both. Overall, 13 (4.63%) had a CMB diagnosis. All of these 13 patients had a true T2∗GRE sequence. CMBs were not detected in the absence of a T2∗GRE sequence.

T2∗GRE imaging is essential for detecting CMBs. When omitted, CMB incidence can be considerably lower than that suggested in the literature. Missing CMB diagnoses in stroke patients may result in suboptimal treatment pathways, compromising the patients' standard of care.

When SWI is not available, it is imperative to always include a true T2∗GRE sequence to detect microbleeds in suspected acute stroke cases.

The aim of this study was to investigate current brain MRI practice, pattern of brain MRI requests, and their appropriateness using the American College of Radiology (ACR) Appropriateness Criteria.

We used direct observation and questionnaires to obtain data concerning routine brain MRI practice. We then retrospectively analyzed (i) demographic characteristics, (ii) clinical history, and (iii) appropriateness of brain MRI requests against published criteria.

All patients were administered the screening questionnaire; however, no reviews were undertaken directly with patients, and no signature of the radiographer was recorded. Apart from routine brain protocol, there were dedicated protocols for epilepsy and stroke. Brain MRI images from 161 patients (85 Males; 76 Females) were analyzed. The age group with most brain MRI requests were from 26 to 45 year olds. The commonest four clinical indications for imaging were brain tumour, headache, seizure, and stroke. Using the ACR Appropriateness Criteria, almost 43% of the brain MRI scans analyzed were found to be “usually appropriate”, 38% were “maybe appropriate” and 19% were categorized as “usually not appropriate”.

There was knowledge gap with regards to MRI safety in local practice, thus there is the utmost need for MRI safety training. Data on the commonest indications for performing brain MRI in this study should be used to inform local neuroradiological practice. Dedicated stroke and epilepsy MRI protocols require additional sequences i.e. MRA and 3D T1 volume acquisition, respectively. The ACR Appropriateness Criteria is recommended for use by the referring practitioners to improve appropriateness of brain MRI requests.

The diagnosis of ischemic stroke in Posterior Fossa (PF) using conventional Computed Tomography (CT) is limited. Meanwhile, the identification of PF slices in CT is a very complicated task due to their varying structures for each slices. Traditionally, the identification of PF slices by the radiologist is based on the PF structure that present in the CT images. This procedure can introduce misidentification issue and time consuming. In this paper, a method for automated classification of PF slices in CT brain images is investigated to sort the slice that contains PF for further study of ischemic stroke detection. This framework employs 11 features of Gray Level Co-Occurrence Matrix (GLCM). All the obtained features have been passed through Artificial Neural Network (ANN) classifier using Multilayer Perceptron (MLP) architecture. The experimental results show that, the combination of GLCM with ANN is promising to be applied in the classification of PF slices as they can provide 93%, 92.5% and 94.2% of precision, recall, and accuracy average rate respectively with average processing time of 0.36 seconds per slices.

Brain and vascular imaging are required components of the emergency assessment of patients with suspected stroke. Either CT or MRI may be used as the initial imaging test. MRI is more sensitive to the presence of acute and chronic ischemic lesions, and chronic microbleeds, but CT remains the most practical and used initial brain imaging test. Although, a non-enhanced CT or T2* MRI sequence showing no haemorrhage is sufficient for deciding intravenous treatment eligibility within the first 4.5 h after stroke onset, a non-invasive intracranial vascular study is strongly recommended during the initial imaging evaluation of the acute stroke patient, particularly if mechanical thrombectomy is contemplated. Advanced imaging with multimodal MRI may facilitate accurate ischemic stroke diagnosis and characterization, and should be considered as an alternative to CT, especially for the selection of patients for acute reperfusion therapy in extended time windows, and in patients in which time of stroke onset is unknown. However, MRI should only be considered in the acute stroke workflow if centres are able to achieve speed and triaging efficiency similar to that which is currently available with CT-based imaging.