Occlusion Detection Research Articles

Diagnostic accuracy of direct ophthalmoscopy in the rapid assessment of avoidable blindness protocol: A cross-sectional study

Background: Rapid Assessment of Avoidable Blindness (RAAB) relies mainly on direct ophthalmoscopy (DO) for the diagnosis of posterior segment eye diseases (PSEDs). There is, however, a growing concern that DO may not be sufficiently sensitive to detect PSED owing to inherent diagnostic limitations. Aim of the Study: The aim of the study was to determine the sensitivity and specificity of DO for the detection of PSEDs in RAAB, using indirect ophthalmoscopy (IO) as a reference standard. Materials and Methods: Participants were patients 50 years and older, presenting to the eye clinic of a tertiary hospital in Jos, between April and September 2016 who gave consent. Their visual acuity was assessed as is done in RAAB6. Those found to be unilaterally or bilaterally visually impaired underwent anterior and posterior segment eye examinations to identify the cause. Dilated fundoscopy was first performed by a senior ophthalmology resident with DO (index test), followed by IO (reference standard test) by a consultant ophthalmologist. Results: A total of 250 patients were recruited into the study, of which 188 took the index and reference standard tests. PSEDs were detected in 65 (34.6%) persons and glaucoma accounted for 87.7% of these. The sensitivity of DO for the detection of glaucoma was 95.3%, while the specificity was 91.0%. Sensitivity and specificity for the detection of diabetic retinopathy and central retinal vein occlusion were both 100%. Specificity for the detection of other PSEDs was good even though sensitivity was below acceptable limits. Conclusion: The performance of DO against IO in this study, suggests that it is a satisfactory modality for the detection of the most common PSEDs in RAAB.

Gaze Weakness Neglect and Speech: An Acute Stroke Scale for Large Vessel Occlusion in the Emergency Department for Faster Treatment

Background: Despite the advancement in acute ischemic stroke with large vessel occlusion (LVO), golden time is lost in assessment lengthy neurological examination and redundantly in the Emergency department, often after emergency medical service prehospital stroke scale evaluation indicating possible LVO. A simple acute ischemic stroke scale (AISS) of the cortical representations of the anterior circulation can rapidly predict LVO, saving precious time to initiate early intravenous tissue plasminogen activator and endovascular mechanical thrombectomy. We proposed an ASIS in the emergency department called Gaze Weakness Neglect Speech (GWNS) to evaluate its feasibility and predictability for the detection of LVO in anterior circulation in the emergency department. Additionally, to evaluate if time can be gained that has been lost in obtaining National Institute of Health stroke Scale (NIHSS) and computed tomographic angiography (CTA), avoiding unnecessary radiation. Methods: This is a prospective observational study. An institutional review board permission was obtained, and patient enrollment started in January 2020 and ended in January 2021. Consecutive patients from January 2020 to September 2021 were selected from the database. The GWNS stroke scale was used by stroke and vascular neurologist during the emergency triage. The GWNS stroke scale scores range from 0 to 4 (1 for positive 0 for negative). The GWNS stroke scale assesses gaze deviation or gaze preference (G), presence of any weakness (W), neglect/disregard (N), and any speech impairment (S). Demographic data, CTA/cerebral angiographic data, and scores from NIHSS were also collected. The collected data was analyzed by a biostatistician to determine the association between the GWNS scale score and LVO. Results: In our study,109 qualifying patients were selected. Fifty-eight patients had GWNS stroke scale score of 3 or 4, with 57 having confirmed LVO and 1 presenting after a seizure. The GWNS stroke score ≥3 (0.86) correlated with LVO better than NIHSS (0.67), regardless of hemisphere side involvement. The GWNS stroke scale score of ≥3 also was effective in detection of proximal and distal blood vessels occlusion in the anterior circulation (Internal carotid artery, middle cerebral artery and its branches). A GWNS stroke scale score of ≥3 with presence of gaze was the most predictive for LVO (0.9) followed by neglect/disregards (0.8). The time to obtain GWNS stroke scale was 1.5 minutes (range 1-3) and time to obtain/interpretation CTA was 41.3 +/- 7.4 minutes after emergency department arrival (range: 29-51 minutes). Conclusions: Our Gaze Weakness Neglect Speech stroke scale can be performed rapidly in the emergency department and is highly predictive of LVO in the internal carotid artery, middle cerebral artery and middle cerebral branches. A GWNS stroke scale score of ≥3 is highly predictive of LVO, especially when gaze or neglect is present. Patients can potentially bypass CTA or advanced imaging in future studies, saving precious time and millions of brain cells for better outcome.

MFF-Net: Multimodal Feature Fusion Network for 3D Object Detection

In complex traffic environment scenarios, it is very important for autonomous vehicles to accurately perceive the dynamic information of other vehicles around the vehicle in advance. The accuracy of 3D object detection will be affected by problems such as illumination changes, object occlusion, and object detection distance. To this purpose, we face these challenges by proposing a multimodal feature fusion network for 3D object detection (MFF-Net). In this research, this paper first uses the spatial transformation projection algorithm to map the image features into the feature space, so that the image features are in the same spatial dimension when fused with the point cloud features. Then, feature channel weighting is performed using an adaptive expression augmentation fusion network to enhance important network features, suppress useless features, and increase the directionality of the network to features. Finally, this paper increases the probability of false detection and missed detection in the non-maximum suppression algorithm by increasing the one-dimensional threshold. So far, this paper has constructed a complete 3D target detection network based on multimodal feature fusion. The experimental results show that the proposed achieves an average accuracy of 82.60% on the Karlsruhe Institute of Technology and Toyota Technological Institute (KITTI) dataset, outperforming previous state-of-the-art multimodal fusion networks. In Easy, Moderate, and hard evaluation indicators, the accuracy rate of this paper reaches 90.96%, 81.46%, and 75.39%. This shows that the MFF-Net network has good performance in 3D object detection.

PFYOLOv4: An Improved Small Object Pedestrian Detection Algorithm

With the development of deep convolutional neural networks, the effect of pedestrian detection has been rapidly improved. However, there are still many problems in small target pedestrian detection, for example noise (such as light) interference, target occlusion, and low detection accuracy. In order to solve the above problems, based on YOLOv4 algorithm, this paper proposes an improved small target pedestrian detection algorithm named PF_YOLOv4. The algorithm is improved in three aspects on the basis of the YOLOv4 algorithm: firstly, a soft thresholding module is added to the residual structure of the backbone network to perform noise reduction process on interference factors, such as light to enhance the robustness of the algorithm; secondly, the depthwise separable convolution replaces the traditional convolution in the YOLOv4 residual structure, to reduce the number of network model parameters; finally, the Convolutional Block Attention Module (CBAM) is added after the output feature map of the backbone network to enhance of the network feature expression. Experimental results show that the PF_YOLOv4 algorithm outperforms most of the state-of-the-art algorithms in detecting small target pedestrians. The mean Average Precision (mAP) of the PF_YOLOv4 algorithm is 2.35% higher than that of the YOLOv4 algorithm and 9.67% higher than that of the YOLOv3 algorithm, while the detection speed is slightly higher than that of YOLOv4 algorithm.

Automatic Anomaly Monitoring in Public Surveillance Areas

With the dramatic increase in video surveillance applications and public safety measures, the need for an accurate and effective system for abnormal/suspicious activity classification also increases. Although it has multiple applications, the problem is very challenging. In this paper, a novel approach for detecting normal/abnormal activity has been proposed. We used the Gaussian Mixture Model (GMM) and Kalman filter to detect and track the objects, respectively. After that, we performed shadow removal to segment an object and its shadow. After object segmentation we performed occlusion detection method to detect occlusion between multiple human silhouettes and we implemented a novel method for region shrinking to isolate occluded humans. Fuzzy c-mean is utilized to verify human silhouettes and motion based features including velocity and optical flow are extracted for each identified silhouettes. Gray Wolf Optimizer (GWO) is used to optimize feature set followed by abnormal event classification that is performed using the XG-Boost classifier. This system is applicable in any surveillance application used for event detection or anomaly detection. Performance of proposed system is evaluated using University of Minnesota (UMN) dataset and UBI (University of Beira Interior)-Fight dataset, each having different type of anomaly. The mean accuracy for the UMN and UBI-Fight datasets is 90.14% and 76.9% respectively. These results are more accurate as compared to other existing methods.

Tracking and Analysis of Pedestrian’s Behavior in Public Places

Crowd management becomes a global concern due to increased population in urban areas. Better management of pedestrians leads to improved use of public places. Behavior of pedestrian’s is a major factor of crowd management in public places. There are multiple applications available in this area but the challenge is open due to complexity of crowd and depends on the environment. In this paper, we have proposed a new method for pedestrian’s behavior detection. Kalman filter has been used to detect pedestrian’s using movement based approach. Next, we have performed occlusion detection and removal using region shrinking method to isolate occluded humans. Human verification is performed on each human silhouette and wavelet analysis and particle gradient motion are extracted for each silhouettes. Gray Wolf Optimizer (GWO) has been utilized to optimize feature set and then behavior classification has been performed using the Extreme Gradient (XG) Boost classifier. Performance has been evaluated using pedestrian’s data from avenue and UBI-Fight datasets, where both have different environment. The mean achieved accuracies are 91.3% and 85.14% over the Avenue and UBI-Fight datasets, respectively. These results are more accurate as compared to other existing methods.

A Novel Method for Digital Orthophoto Generation from Top View Constrained Dense Matching

The digital orthophoto is an image with both map geometric accuracy and image characteristics, which is commonly used in geographic information systems (GIS) as a background image. Existing methods for digital orthophoto generation are generally based on a 3D reconstruction. However, the digital orthophoto is only the top view of the 3D reconstruction result with a certain spatial resolution. The computation about the surfaces vertical to the ground and details less than the spatial resolution is redundant for digital orthophoto generation. This study presents a novel method for digital orthophoto generation based on top view constrained dense matching (TDM). We first reconstruct some sparse points using the features in the image sequence based on the structure-from-motion (SfM) method. Second, we use a raster to locate the sparse 3D points. Each cell indicates a pixel of the output digital orthophoto. The size of the cell is related to the required spatial resolution. Only some cells with initial values from the sparse 3D points are considered seed cells. The values of other cells around the seed points are computed from a top-down propagation based on color constraints and occlusion detection from multiview-related images. The propagation process continued until the entire raster was occupied. Since the process of TDM is on a raster and only one point is saved in each cell, TDM effectively eliminate the redundant computation. We tested TDM on various scenes and compared it with some commercial software. The experiments showed that our method’s accuracy is the same as the result of commercial software, together with a time consumption decrease as the spatial resolution decreases.

About Segmath, a new Cerebral Vascular Segmentation Software after CTA

Objectives: The new segmentation software Segmath delivers a 3D view of the cerebral vascular structures without superposition of bony or other structures. This will, according to the literature, improve the workflow of stroke patients and increase the occlusion detection rate on the original CTA. Materials and methods: The software written in MATLAB is based on the analysis of the local Hessian matrix with new original functions of the resulting local eigenvalues. No user intervention in the segmentation process is needed. Results: The validation of the new software yields good results both with synthetic data and real CTA’s. Conclusion: This segmentation software is a powerful additional diagnostic tool available to radiologists and neurologists examining and treating stroke patients. This will improve the workflow of suspected stroke patients.

Automated detection of intracranial large vessel occlusions using Viz.ai software: Experience in a large, integrated stroke network.

Endovascular thrombectomy is an evidence-based treatment for large vessel occlusion (LVO) stroke. Commercially available artificial intelligence has been designed to detect the presence of an LVO on computed tomography angiogram (CTA). We compared Viz.ai-LVO (San Francisco, CA, USA) to CTA interpretation by board-certified neuroradiologists (NRs) in a large, integrated stroke network. From January 2021 to December 2021, we compared Viz.ai detection of an internal carotid artery (ICA) or middle cerebral artery first segment (MCA-M1) occlusion to the gold standard of CTA interpretation by board-certified NRs for all code stroke CTAs. On a monthly basis, sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV) were calculated. Trend analyses were conducted to evaluate for any improvement of LVO detection by the software over time. 3851 patients met study inclusion criteria, of whom 220 (5.7%) had an ICA or MCA-M1 occlusion per NR. Sensitivity and specificity were 78.2% (95% CI 72%-83%) and 97% (95% CI 96%-98%), respectively. PPV was 61% (95% CI 55%-67%), NPV 99% (95% CI 98%-99%), and accuracy was 95.9% (95% CI 95.3%-96.5%). Neither specificity or sensitivity improved over time in the trend analysis. Viz.ai-LVO has high specificity and moderately high sensitivity to detect an ICA or proximal MCA occlusion. The software has the potential to streamline code stroke workflows and may be particularly impactful when emergency access to NRs or vascular neurologists is limited.

Occlusion detection of traffic signs by voxel-based ray tracing using highly detailed models and MLS point clouds of vegetation

Visibility analysis plays a vital role in the design and placing of traffic signs in the urban street environment. This work investigates the occlusion detection of traffic lights and traffic signs caused by vegetation. The presented analysis method is built upon the inputs from the expected situation reflected by a highly detailed 3D city model and the as-is situation captured by 3D Mobile Laser Scanning (MLS). The model contains the location and orientation of streets, traffic lights, and traffic signs; the measurements add detail on irregular-shaped and morphing objects such as vegetation, respectively. The analysis covers the visibility of traffic lights and traffic signs by ray-tracing in an occupancy grid that is generated by the voxelization of the space. The voxels facilitate the distinction between occupied and empty space. The identification of unknown volumes is added and considered in the decision process, to cope with the regions invisible to the sensor. As output, we provide a visibility metric and detailed 3D space descriptions on different levels of granularity, including the knowledge of the semantic classes of traversed voxels. During the whole process, the awareness of unknown volumes is added to an otherwise binary decision between visible and invisible targets. Experiments are conducted on the TUM-MLS-2016 dataset. Results demonstrate that the proposed method is feasible for the detection of occlusions by vegetation in the street scenario, and reveal that the identification of unknown volumes proves necessary for a reliable interpretation of the measurements. • Agent-based visibility analysis of traffic installations. • Combination of a street environment model with actual MLS measurements of vegetation. • Hence solving that realistic volumetric modeling of trees is often not present in city models. • Inclusion of information on volumes not covered by measurements. • The analysis gains an additional level of confidence depending on lines of sight traveling through space of unknown occupation.

Modified GAN for Natural Occlusion Detection and Inpainting of Raw Footage From Video Surveillance

Low resolution and occlusion are mainly prominent in images taken from certain unconstrained environments such as raw footage from video surveillance. In this work, a deep generative adversarial network for joint face completion and face super-resolution is proposed. It will be really useful in the current COVID-19 scenario as people wearing masks are a common sight. Given an input of a low-resolution face image with occlusion, the generator aims to recover a high-resolution face image without occlusion. The discriminator uses a set of carefully designed losses to assure the high quality of the recovered high-resolution face images without occlusion. Experimental results on CelebA database show that the proposed approach outperforms the state-of-the-art methods in jointly performing face super-resolution and face completion, and shows good generalization ability in cross-database testing. MSSIM showed an accuracy of around 80% for test cases. The recorded values of generator adversarial loss, generator pixel loss, and discriminator loss are 0.93, 0.10, and 0.003, respectively.

Safety monitoring method of moving target in underground coal mine based on computer vision processing

Coal is one of the main energy sources in China. The country attaches great importance to the development of coal mining industry, and coal production is on the rise. At the same time, mine safety accidents are becoming more and more frequent, and the country is paying more and more attention to mine safety accidents. The underground environment of coal mine is complex, noisy and uneven, and there will be problems such as occlusion and high false detection rate during video monitoring. In order to ensure the safety of underground personnel, moving target detection and tracking based on video monitoring information is of great significance for coal mine safety production. The purpose of this paper is to study how to analyze and study the monitoring of moving targets in coal mines based on computer vision processing, and describe the image processing methods. This paper puts forward the problem of target monitoring, which is based on image processing, and then elaborates on the concept of image enhancement and related algorithms. From the average gradient, the algorithm in this paper is 56.60% higher than the histogram equalization algorithm, and 68.26% higher than the dark primary color prior dehazing algorithm. and designs and analyzes cases of image enhancement in coal mines. The experimental results show that the information entropy of the algorithm in this paper is 31.10% higher than that of the dark primary color prior dehazing algorithm, and 18.72% higher than that of the histogram equalization algorithm. It can be seen that the algorithm in this paper can achieve better enhancement effect.

Robust self-supervised monocular visual odometry based on prediction-update pose estimation network

Visual odometry aims at estimating the camera pose from video sequence, which is an important part of visual Simultaneous Localization and Mapping (SLAM). In this paper, we propose a novel prediction-update pose estimation network, PU-PoseNet, for self-supervised monocular visual odometry. It allows the network to use the effective information of the previous frame in estimating the current pose. The long-time pose consistency constraint-based motion weighted photometric loss is designed to make the network to pay more attention to the pixels of stationary objects and enhance the time consistency of estimation results. The depth image-based occlusion detection, the depth smoothness loss and auto-Masking are used to construct the depth consistency constraint loss term to reduce the influences of interferences such as occlusion. To further improve the robustness and the accuracy of the proposed method, both the depth consistency constraint and the variational auto-encoder are used for network training. For frame missing cases, a novel frame missing training strategy is used to make our method adapt frame missing cases. Extensive experiments on the KITTI dataset have validated the effectiveness of our proposed method.

Head-to-head comparison of commercial artificial intelligence solutions for detection of large vessel occlusion at a comprehensive stroke center.

PurposeDespite the availability of commercial artificial intelligence (AI) tools for large vessel occlusion (LVO) detection, there is paucity of data comparing traditional machine learning and deep learning solutions in a real-world setting. The purpose of this study is to compare and validate the performance of two AI-based tools (RAPID LVO and CINA LVO) for LVO detection.Materials and methodsThis was a retrospective, single center study performed at a comprehensive stroke center from December 2020 to June 2021. CT angiography (n = 263) for suspected stroke were evaluated for LVO. RAPID LVO is a traditional machine learning model which primarily relies on vessel density threshold assessment, while CINA LVO is an end-to-end deep learning tool implemented with multiple neural networks for detection and localization tasks. Reasons for errors were also recorded.ResultsThere were 29 positive and 224 negative LVO cases by ground truth assessment. RAPID LVO demonstrated an accuracy of 0.86, sensitivity of 0.90, specificity of 0.86, positive predictive value of 0.45, and negative predictive value of 0.98, while CINA demonstrated an accuracy of 0.96, sensitivity of 0.76, specificity of 0.98, positive predictive value of 0.85, and negative predictive value of 0.97.ConclusionBoth tools successfully detected most anterior circulation occlusions. RAPID LVO had higher sensitivity while CINA LVO had higher accuracy and specificity. Interestingly, both tools were able to detect some, but not all M2 MCA occlusions. This is the first study to compare traditional and deep learning LVO tools in the clinical setting.

Noise-optimized virtual monoenergetic reconstructions of dual-energy CT angiographies improve assessability of the lower leg arterial segments in peripheral arterial occlusive disease

IntroductionThe aim of this study was to evaluate the influence of a noise optimized virtual monoenergetic reconstruction algorithm (VMI+) on the image quality and assessability of dual energy (DE) computed tomography angiography (CTA) of the lower extremity runoff. MethodsA total of 118 lower extremity runoff CTA performed on a 3rd generation DE-CT scanner in 109 patients (54 females; 75.6 ± 9.5 years) were included in this retrospective study. Axial image stacks were reconstructed with a standard 120 kV setting and VMI+ of different keV levels. Objective image quality criteria (contrast attenuation, signal-to-noise [SNR] and contrast-to-noise ratio [CNR]) were measured. Two radiologists evaluated subjective image quality regarding intraluminal attenuation and image noise using a 5-point Likert scale. Diagnostic accuracy for significant stenosis (>75%) and vessel occlusion was assessed for 120 kV and 50 keV VMI+ images rated by two radiologists. In all patients, a digital subtraction angiography (DSA) rated by on board-certified radiologist served as the standard of reference. ResultsIntraluminal attenuation was highest in 40/50 keV VMI+ while SNR were similar to 120 kV images. In subjective assessment, intraluminal contrast of 50 keV images was deemed superior compared to 120 kV despite higher image noise. Sensitivity, specificity, and accuracy for detection of a vessel occlusion were similar in 50 keV VMI+ compared to 120 kV (70%/92%/84%; 70%/91%/83%; p < 0.001) but 13 of 118 (11%) lower leg runoffs were only assessable with 50 keV VMI+. ConclusionVMI+ reconstructions improve assessability of DE-CTA by increased luminal attenuation with consistent image noise, also allowing the evaluation of lower leg arterial segments inassessable with standard reconstructions. Implications for practiceProviding higher intraluminal attenuation and similar image noise compared with conventional reconstructions, 50 keV VMI+ may be appropriate for routine evaluation of DE-CTA.

An algorithm for the labeling and interactive visualization of the cerebrovascular system of ischemic strokes

During the diagnosis of ischemic strokes, the Circle of Willis and its surrounding vessels are the arteries of interest. Their visualization in case of an acute stroke is often enabled by Computed Tomography Angiography (CTA). Still, the identification and analysis of the cerebral arteries remain time consuming in such scans due to a large number of peripheral vessels which may disturb the visual impression. We propose VirtualDSA++, an algorithm designed to segment and label the cerebrovascular tree on CTA scans. Especially with stroke patients, labeling is a delicate procedure, as in the worst case whole hemispheres may not be present due to impeded perfusion. Hence, we extended the labeling mechanism for the cerebral arteries to identify occluded vessels. In the work at hand, we place the algorithm in a clinical context by evaluating the labeling and occlusion detection on stroke patients, where we have achieved labeling sensitivities comparable to other works between 92% and 95%. To the best of our knowledge, ours is the first work to address labeling and occlusion detection at once, whereby a sensitivity of 67% and a specificity of 81% were obtained for the latter. VirtualDSA++ also automatically segments and models the intracranial system leading to further processing possibilities. We present the generic concept of iterative systematic search for pathways on all nodes of said model, which enables new interactive features. Exemplary, we derive in detail, firstly, the interactive planning of vascular interventions like the mechanical thrombectomy and secondly, the interactive suppression of vessel structures that are not of interest in diagnosing strokes (like veins). We discuss both features as well as further possibilities emerging from the proposed concept.

Revascularization in the treatment of acute ischemic stroke

Objective ‒ to analyze the experience of providing care to patients with acute stroke in the conditions of the multidisciplinary neurosurgical department of Poltava Regional Clinical Hospital named after M.V. Sklifosovsky.Materials and methods. In the neurosurgical department in 2020 treated 1,148 patients aged 18 to 83, among whom 49 % were patients with acute stroke. Ischemic stroke was observed in 54.7 % of people, and hemorrhagic stroke in 45.3 %. A total of 173 patients with a diagnosis of acute ischemic stroke were hospitalized in 2020, of which 54 patients were delivered within 4.5 hours from the onset of the disease and received thrombolytic therapy and 20 patients within 6 hours who underwent mechanical thrombectomy, with of them, bridging was used in 5 patients. To determine the presence of signs of a stroke and pre-notification, the following scales were used: FAST and «МОЗОК-ЧАС». Neuroimaging was performed as an emergency using native SCT and SCT-angiography or MRT and MRT-angiography in 100 % of cases. To evaluate the results of neuroimaging, the ASPECTS scale was used for stroke in the anterior hemisphere and pc-ASPECTS – in the vertebrobasilar basin. The NIHSS scale was used to determine the severity of the patient; the following scales were used to determine the presence of occlusion of a large vessel: RACE, BRAIN2. A modified Rankin scale was used to assess the patient’s functional status. In the treatment of ischemic stroke, we divided all patients into 2 groups: the first – 41 % patients, who are subject to revascularization treatment, the second – 59 % patients, who are shown only optimal drug therapy. Revascularization had 2 treatment options. The first option provided for thrombolytic therapy in case of detection of small vessel occlusion, the second – mechanical thrombectomy in case of detection of large vessel occlusion. In cases where the patient was admitted within the therapeutic window, bridging therapy was performed, i.e. intravenous administration of plasminogen activator and simultaneous endovascular thrombectomy. In case of simultaneous detection of a stroke and a heart attack in a patient, the Canadian Stroke Best Practice Recommendations for Acute Stroke Management (2018) were used as a basis, according to which the specifics of the management of patients undergoing revascularization treatment are defined. The following EVT techniques were used in our study: ADAPT – in 5 % of patients, Solumbra – in 10 % of patients and SAVE – in 85 % of patients during endovascular thrombectomy.Results. Thrombolytic therapy was carried out – 31 % of the total number of treated patients with ischemic stroke, mechanical thrombectomy – 10 % of the total number of treated patients. Decompressive craniectomies were performed in 6 % of patients with ischemic stroke.Conclusions. Analysis of the results of treatment of patients with ischemic stroke using the above methods indicates a good result at discharge from the medical institution and later, namely on the 90th day after the treatment.

Automated detection of intracranial artery stenosis and occlusion in magnetic resonance angiography: A preliminary study based on deep learning

Background and objectivesIntracranial atherosclerotic stenosis of a major intracranial artery is the common cause of ischemic stroke. We evaluate the feasibility of using deep learning to automatically detect intracranial arterial steno-occlusive lesions from time-of-flight magnetic resonance angiography. MethodsIn a retrospective study, magnetic resonance images with radiological reports of intracranial arterial stenosis and occlusion were extracted. The images were randomly divided into a training set and a test set. The manual annotation of lesions with a bounding box labeled “moderate stenosis,” “severe stenosis,” “occlusion,” and “absence of signal” was considered as ground truth. A deep learning algorithm based on you only look once version 5 (YOLOv5) detection model was developed with the training set, and its sensitivity and positive predictive values to detect lesions were evaluated in the test set. ResultsA dataset of 200 examinations consisted of a total of 411 lesions—242 moderate stenoses, 84 severe stenoses, 70 occlusions, and 15 absence of signal. The magnetic resonance images contained 291 lesions in the training set and 120 lesions in the test set. The sensitivity and positive predictive values were 64.2 and 83.7%, respectively. The detection sensitivity in relation to the location was greatest in the internal carotid artery (86.2%). ConclusionsApplying deep learning algorithms in the automated detection of intracranial arterial steno-occlusive lesions from time-of-flight magnetic resonance angiography is feasible and has great potential.

Image level detection of large vessel occlusion on 4D-CTA perfusion data using deep learning in acute stroke

ObjectivesAutomated image-level detection of large vessel occlusions (LVO) could expedite patient triage for mechanical thrombectomy. A few studies have previously attempted LVO detection using artificial intelligence (AI) on CT angiography (CTA) images. To our knowledge this is the first study to detect LVO existence and location on raw 4D-CTA/ CT perfusion (CTP) images using neural network (NN) models. Materials and MethodsRetrospective study using data from a level-I stroke center was performed. A total of 306 (187 with LVO, and 119 without) patients were evaluated. Image pre-processing included co-registration, normalization and skull stripping. Five consecutive time-points for each patient were selected to provide variable contrast density in data. Additional data augmentation included rotation and horizonal image flipping. Our model architecture consisted of two neural networks, first for classification (based on hemispheric asymmetry), followed by second model for exact site of LVO detection. Only cases deemed positive by the classification model were routed to the detection model, thereby reducing false positives and improving specificity. The results were compared with expert annotated LVO detection. ResultsUsing a 80:20 split for training and validation, the combination of both classification and detection model achieved a sensitivity of 86.5%, a specificity of 89.5%, and an accuracy of 87.5%. A 5-fold cross-validation using the entire data achieved a mean sensitivity of 82.7%, a specificity of 89.8%, and an accuracy of 85.5% and a mean AUC of 0.89 (95% CI: 0.85-0.93). ConclusionOur findings suggest that accurate image-level LVO detection is feasible on CTP raw images.

The implementation of artificial intelligence significantly reduces door-in-door-out times in a primary care center prior to transfer.

Viz LVO artificial intelligence (AI) software utilizes AI-powered large vessel occlusion (LVO) detection technology which automatically identifies suspected LVO through CT angiogram (CTA) imaging and alerts on-call stroke teams. This analysis was performed to determine whether AI software can reduce the door-in-door-out (DIDO) time interval within the primary care center (PSC) prior to transfer to the comprehensive care center (CSC). We compared the DIDO time interval for all LVO transfer patients from a single-spoke PSC to our CSC prior to (February 2017 to November 2018) and after (November 2018 to June 2020) incorporating AI. Using a stroke database at a CSC, demographics, DIDO time at PSC, modified Rankin Scale (mRS) at 90-days, mortality rate at discharge, length of stay (LOS), and intracranial hemorrhage rates were examined. There were a total of 63 patients during the study period (average age 69.99 ± 13.72, 55.56% female). We analyzed 28 patients pre-AI (average age 71.64 ± 12.28, 46.4% female), and 35 patients post-AI (average age 68.67 ± 14.88, 62.9% female). After implementing the AI software, the mean DIDO time interval within the PSC significantly improved by 102.3 min (226.7 versus 124.4 min; p = 0.0374). The incorporation of the AI software was associated with a significant improvement in DIDO times within the PSC as well as CTA to door-out time in the PSC. More extensive studies are warranted to expand on the ability of AI technology to improve transfer times and outcomes for LVO patients.