Fast Automatic Detection of Large Vessel Occlusions on CT Angiography

Automated detection of large vessel occlusion using deep learning: a pivotal multicenter study and reader performance study

BACKGROUND

To evaluate the stand-alone efficacy and improvements in diagnostic accuracy of early-career physicians of the artificial intelligence (AI) software to detect large vessel occlusion (LVO) in CT angiography (CTA).

METHODS

This multicenter study included 595 ischemic stroke patients from January 2021 to September 2023. Standard references and LVO locations were determined by consensus among three experts. The efficacy of the AI software was benchmarked against standard references, and its impact on the diagnostic accuracy of four residents involved in stroke care was assessed. The area under the receiver operating characteristic curve (AUROC), sensitivity, and specificity of the software and readers with versus without AI assistance were calculated.

RESULTS

Among the 595 patients (mean age 68.5±13.4 years, 56% male), 275 (46.2%) had LVO. The median time interval from the last known well time to the CTA was 46.0 hours (IQR 11.8-64.4). For LVO detection, the software demonstrated a sensitivity of 0.858 (95% CI 0.811 to 0.897) and a specificity of 0.969 (95% CI 0.943 to 0.985). In subjects whose symptom onset to imaging was within 24 hours (n=195), the software exhibited an AUROC of 0.973 (95% CI 0.939 to 0.991), a sensitivity of 0.890 (95% CI 0.817 to 0.936), and a specificity of 0.965 (95% CI 0.902 to 0.991). Reading with AI assistance improved sensitivity by 4.0% (2.17 to 5.84%) and AUROC by 0.024 (0.015 to 0.033) (all P<0.001) compared with readings without AI assistance.

CONCLUSIONS

The AI software demonstrated a high detection rate for LVO. In addition, the software improved diagnostic accuracy of early-career physicians in detecting LVO, streamlining stroke workflow in the emergency room.

Viz LVO versus Rapid LVO in detection of large vessel occlusion on CT angiography for acute stroke

BACKGROUND

Endovascular thrombectomy improves outcomes and reduces mortality for large vessel occlusion (LVO) and is time-sensitive. Computer automation may aid in the early detection of LVOs, but false values may lead to alarm desensitization. We compared Viz LVO and Rapid LVO for automated LVO detection.

METHODS

Data were retrospectively extracted from Rapid LVO and Viz LVO running concurrently from January 2022 to January 2023 on CT angiography (CTA) images compared with a radiologist interpretation. We calculated diagnostic accuracy measures and performed a McNemar test to look for a difference between the algorithms' errors. We collected demographic data, comorbidities, ejection fraction (EF), and imaging features and performed a multiple logistic regression to determine if any of these variables predicted the incorrect classification of LVO on CTA.

RESULTS

360 participants were included, with 47 large vessel occlusions. Viz LVO and Rapid LVO had a specificity of 0.96 and 0.85, a sensitivity of 0.87 and 0.87, a positive predictive value of 0.75 and 0.46, and a negative predictive value of 0.98 and 0.97, respectively. A McNemar test on correct and incorrect classifications showed a statistically significant difference between the two algorithms' errors (P=0.00000031). A multiple logistic regression showed that low EF (Viz P=0.00125, Rapid P=0.0286) and Modified Woodcock Score >1 (Viz P=0.000198, Rapid P=0.000000975) were significant predictors of incorrect classification.

CONCLUSION

Rapid LVO produced a significantly larger number of false positive values that may contribute to alarm desensitization, leading to missed alarms or delayed responses. EF and intracranial atherosclerosis were significant predictors of incorrect predictions.

StrokeClassifier: ischemic stroke etiology classification by ensemble consensus modeling using electronic health records

Determining acute ischemic stroke (AIS) etiology is fundamental to secondary stroke prevention efforts but can be diagnostically challenging. We trained and validated an automated classification tool, StrokeClassifier, using electronic health record (EHR) text from 2039 non-cryptogenic AIS patients at 2 academic hospitals to predict the 4-level outcome of stroke etiology adjudicated by agreement of at least 2 board-certified vascular neurologists' review of the EHR. StrokeClassifier is an ensemble consensus meta-model of 9 machine learning classifiers applied to features extracted from discharge summary texts by natural language processing. StrokeClassifier was externally validated in 406 discharge summaries from the MIMIC-III dataset reviewed by a vascular neurologist to ascertain stroke etiology. Compared with vascular neurologists' diagnoses, StrokeClassifier achieved the mean cross-validated accuracy of 0.74 and weighted F1 of 0.74 for multi-class classification. In MIMIC-III, its accuracy and weighted F1 were 0.70 and 0.71, respectively. In binary classification, the two metrics ranged from 0.77 to 0.96. The top 5 features contributing to stroke etiology prediction were atrial fibrillation, age, middle cerebral artery occlusion, internal carotid artery occlusion, and frontal stroke location. We designed a certainty heuristic to grade the confidence of StrokeClassifier's diagnosis as non-cryptogenic by the degree of consensus among the 9 classifiers and applied it to 788 cryptogenic patients, reducing cryptogenic diagnoses from 25.2% to 7.2%. StrokeClassifier is a validated artificial intelligence tool that rivals the performance of vascular neurologists in classifying ischemic stroke etiology. With further training, StrokeClassifier may have downstream applications including its use as a clinical decision support system.

A self-supervised learning approach for registration agnostic imaging models with 3D brain CTA

Deep learning-based neuroimaging pipelines for acute stroke typically rely on image registration, which not only increases computation but also introduces a point of failure. In this paper, we propose a general-purpose contrastive self-supervised learning method that converts a convolutional deep neural network designed for registered images to work on a different input domain, i.e., with unregistered images. This is accomplished by using a self-supervised strategy that does not rely on labels, where the original model acts as a teacher and a new network as a student. Large vessel occlusion (LVO) detection experiments using computed tomographic angiography (CTA) data from 402 CTA patients show the student model achieving competitive LVO detection performance (area under the receiver operating characteristic curve [AUC] = 0.88 vs. AUC = 0.81) compared to the teacher model, even with unregistered images. The student model trained directly on unregistered images using standard supervised learning achieves an AUC = 0.63, highlighting the proposed method's efficacy in adapting models to different pipelines and domains.

Software with artificial intelligence-derived algorithms for analysing CT brain scans in people with a suspected acute stroke: a systematic review and cost-effectiveness analysis

Background

Artificial intelligence-derived software technologies have been developed that are intended to facilitate the review of computed tomography brain scans in patients with suspected stroke.

Objectives

To evaluate the clinical and cost-effectiveness of using artificial intelligence-derived software to support review of computed tomography brain scans in acute stroke in the National Health Service setting.

Methods

Twenty-five databases were searched to July 2021. The review process included measures to minimise error and bias. Results were summarised by research question, artificial intelligence-derived software technology and study type. The health economic analysis focused on the addition of artificial intelligence-derived software-assisted review of computed tomography angiography brain scans for guiding mechanical thrombectomy treatment decisions for people with an ischaemic stroke. The de novo model (developed in R Shiny, R Foundation for Statistical Computing, Vienna, Austria) consisted of a decision tree (short-term) and a state transition model (long-term) to calculate the mean expected costs and quality-adjusted life-years for people with ischaemic stroke and suspected large-vessel occlusion comparing artificial intelligence-derived software-assisted review to usual care.

Results

A total of 22 studies (30 publications) were included in the review; 18/22 studies concerned artificial intelligence-derived software for the interpretation of computed tomography angiography to detect large-vessel occlusion. No study evaluated an artificial intelligence-derived software technology used as specified in the inclusion criteria for this assessment. For artificial intelligence-derived software technology alone, sensitivity and specificity estimates for proximal anterior circulation large-vessel occlusion were 95.4% (95% confidence interval 92.7% to 97.1%) and 79.4% (95% confidence interval 75.8% to 82.6%) for Rapid (iSchemaView, Menlo Park, CA, USA) computed tomography angiography, 91.2% (95% confidence interval 77.0% to 97.0%) and 85.0 (95% confidence interval 64.0% to 94.8%) for Viz LVO (Viz.ai, Inc., San Fransisco, VA, USA) large-vessel occlusion, 83.8% (95% confidence interval 77.3% to 88.7%) and 95.7% (95% confidence interval 91.0% to 98.0%) for Brainomix (Brainomix Ltd, Oxford, UK) e-computed tomography angiography and 98.1% (95% confidence interval 94.5% to 99.3%) and 98.2% (95% confidence interval 95.5% to 99.3%) for Avicenna CINA (Avicenna AI, La Ciotat, France) large-vessel occlusion, based on one study each. These studies were not considered appropriate to inform cost-effectiveness modelling but formed the basis by which the accuracy of artificial intelligence plus human reader could be elicited by expert opinion. Probabilistic analyses based on the expert elicitation to inform the sensitivity of the diagnostic pathway indicated that the addition of artificial intelligence to detect large-vessel occlusion is potentially more effective (quality-adjusted life-year gain of 0.003), more costly (increased costs of £8.61) and cost-effective for willingness-to-pay thresholds of £3380 per quality-adjusted life-year and higher.

Limitations and conclusions

The available evidence is not suitable to determine the clinical effectiveness of using artificial intelligence-derived software to support the review of computed tomography brain scans in acute stroke. The economic analyses did not provide evidence to prefer the artificial intelligence-derived software strategy over current clinical practice. However, results indicated that if the addition of artificial intelligence-derived software-assisted review for guiding mechanical thrombectomy treatment decisions increased the sensitivity of the diagnostic pathway (i.e. reduced the proportion of undetected large-vessel occlusions), this may be considered cost-effective.

Future work

Large, preferably multicentre, studies are needed (for all artificial intelligence-derived software technologies) that evaluate these technologies as they would be implemented in clinical practice.

Study registration

This study is registered as PROSPERO CRD42021269609.

Funding

This award was funded by the National Institute for Health and Care Research (NIHR) Evidence Synthesis programme (NIHR award ref: NIHR133836) and is published in full in Health Technology Assessment; Vol. 28, No. 11. See the NIHR Funding and Awards website for further award information.

StrokeClassifier: Ischemic Stroke Etiology Classification by Ensemble Consensus Modeling Using Electronic Health Records

Determining the etiology of an acute ischemic stroke (AIS) is fundamental to secondary stroke prevention efforts but can be diagnostically challenging. We trained and validated an automated classification machine intelligence tool, StrokeClassifier, using electronic health record (EHR) text data from 2,039 non-cryptogenic AIS patients at 2 academic hospitals to predict the 4-level outcome of stroke etiology determined by agreement of at least 2 board-certified vascular neurologists' review of the stroke hospitalization EHR. StrokeClassifier is an ensemble consensus meta-model of 9 machine learning classifiers applied to features extracted from discharge summary texts by natural language processing. StrokeClassifier was externally validated in 406 discharge summaries from the MIMIC-III dataset reviewed by a vascular neurologist to ascertain stroke etiology. Compared with stroke etiologies adjudicated by vascular neurologists, StrokeClassifier achieved the mean cross-validated accuracy of 0.74 (±0.01) and weighted F1 of 0.74 (±0.01). In the MIMIC-III cohort, the accuracy and weighted F1 of StrokeClassifier were 0.70 and 0.71, respectively. SHapley Additive exPlanation analysis elucidated that the top 5 features contributing to stroke etiology prediction were atrial fibrillation, age, middle cerebral artery occlusion, internal carotid artery occlusion, and frontal stroke location. We then designed a certainty heuristic to deem a StrokeClassifier diagnosis as confidently non-cryptogenic by the degree of consensus among the 9 classifiers, and applied it to 788 cryptogenic patients. This reduced the percentage of the cryptogenic strokes from 25.2% to 7.2% of all ischemic strokes. StrokeClassifier is a validated artificial intelligence tool that rivals the performance of vascular neurologists in classifying ischemic stroke etiology for individual patients. With further training, StrokeClassifier may have downstream applications including its use as a clinical decision support system.

Automated Intracranial Clot Detection: A Promising Tool for Vascular Occlusion Detection in Non-Enhanced CT

(1) Background: to test the diagnostic performance of a fully convolutional neural network-based software prototype for clot detection in intracranial arteries using non-enhanced computed tomography (NECT) imaging data. (2) Methods: we retrospectively identified 85 patients with stroke imaging and one intracranial vessel occlusion. An automated clot detection prototype computed clot location, clot length, and clot volume in NECT scans. Clot detection rates were compared to the visual assessment of the hyperdense artery sign by two neuroradiologists. CT angiography (CTA) was used as the ground truth. Additionally, NIHSS, ASPECTS, type of therapy, and TOAST were registered to assess the relationship between clinical parameters, image results, and chosen therapy. (3) Results: the overall detection rate of the software was 66%, while the human readers had lower rates of 46% and 24%, respectively. Clot detection rates of the automated software were best in the proximal middle cerebral artery (MCA) and the intracranial carotid artery (ICA) with 88-92% followed by the more distal MCA and basilar artery with 67-69%. There was a high correlation between greater clot length and interventional thrombectomy and between smaller clot length and rather conservative treatment. (4) Conclusions: the automated clot detection prototype has the potential to detect intracranial arterial thromboembolism in NECT images, particularly in the ICA and MCA. Thus, it could support radiologists in emergency settings to speed up the diagnosis of acute ischemic stroke, especially in settings where CTA is not available.

Performance of deep learning-based autodetection of arterial stenosis on head and neck CT angiography: an independent external validation study

PURPOSE

To externally validate the performance of automated stenosis detection on head and neck CT angiography (CTA) and investigate the impact factors using an independent bi-center dataset with digital subtraction angiography (DSA) as the ground truth.

MATERIAL AND METHODS

Patients who underwent head and neck CTA and DSA between January 2019 and December 2021 were retrospectively included. The degree of stenosis was automatically evaluated using CerebralDoc based on CTA. The performance of CerebralDoc across levels (per-patient, per-region, per-vessel, and per-segment) and thresholds (≥ 50%, ≥ 70%, and = 100%) was evaluated. Logistic regression was performed to identify independent factors associated with false negative results.

RESULTS

296 patients were analyzed. Specificity across levels and thresholds was high, exceeding 92%. The area under the curve ranged from poor (0.615, 95% CI: 0.544, 0.686; at the region-based analysis for stenosis ≥ 70%) to excellent (0.945, 95% CI: 0.905, 0.985; at the patient-based analysis for stenosis ≥ 50%). Sensitivity ranged from 0.714 (95% CI: 0.675, 0.750) at the segment-based analysis for stenosis ≥ 70% to 0.895 (95% CI: 0.849, 0.919) at the patient-based analysis for stenosis ≥ 50%. The multiple logistic regression analysis revealed that false negative results were primarily more likely to specific stenosis locations (particularly the M2 segment and skull base segment of the internal carotid artery) and occlusion.

CONCLUSIONS

CerebralDoc has the potential to automated stenosis detection on head and neck CTA, but further efforts are needed to optimize its performance.

Deep-learning based detection of vessel occlusions on CT-angiography in patients with suspected acute ischemic stroke

Swift diagnosis and treatment play a decisive role in the clinical outcome of patients with acute ischemic stroke (AIS), and computer-aided diagnosis (CAD) systems can accelerate the underlying diagnostic processes. Here, we developed an artificial neural network (ANN) which allows automated detection of abnormal vessel findings without any a-priori restrictions and in <2 minutes. Pseudo-prospective external validation was performed in consecutive patients with suspected AIS from 4 different hospitals during a 6-month timeframe and demonstrated high sensitivity (≥87%) and negative predictive value (≥93%). Benchmarking against two CE- and FDA-approved software solutions showed significantly higher performance for our ANN with improvements of 25-45% for sensitivity and 4-11% for NPV (p ≤ 0.003 each). We provide an imaging platform ( https://stroke.neuroAI-HD.org ) for online processing of medical imaging data with the developed ANN, including provisions for data crowdsourcing, which will allow continuous refinements and serve as a blueprint to build robust and generalizable AI algorithms.

The diagnostic performance of artificial intelligence algorithms for identifying M2 segment middle cerebral artery occlusions: A systematic review and meta-analysis

BACKGROUND

Artificial intelligence (AI)-based algorithms have been developed to facilitate rapid and accurate computed tomography angiography (CTA) assessment in proximal large vessel occlusion (LVO) acute ischemic stroke, including internal carotid artery and M1 occlusions. In clinical practice, however, the detection of medium vessel occlusion (MeVO) represents an ongoing diagnostic challenge in which the added value of AI remains unclear.

PURPOSE

To assess the diagnostic performance of AI platforms for detecting M2 occlusions.

METHODS

Studies that report the diagnostic performance of AI-based detection of M2 occlusions were screened, and sensitivity and specificity data were extracted using the semi-automated AutoLit software (Nested Knowledge, MN) platform. STATA (version 16 IC; Stata Corporation, College Station, Texas, USA) was used to conduct all analyses.

RESULTS

Eight studies with a low risk of bias and significant heterogeneity were included in the quantitative and qualitative synthesis. The pooled estimates of sensitivity and specificity of AI platforms for M2 occlusion detection were 64% (95% CI, 53 to 74%) and 97% (95% CI, 84 to 100%), respectively. The area under the curve (AUC) in the SROC curve was 0.79 (95% CI, 0.74 to 0.83).

CONCLUSION

The current performance of the AI-based algorithm makes it more suitable as an adjunctive confirmatory tool rather than as an independent one for M2 occlusions. With the rapid development of such algorithms, it is anticipated that newer generations will likely perform much better.

Impact of an automated large vessel occlusion detection tool on clinical workflow and patient outcomes

Purpose

Automated large vessel occlusion (LVO) tools allow for prompt identification of positive LVO cases, but little is known about their role in acute stroke triage when implemented in a real-world setting. The purpose of this study was to evaluate the automated LVO detection tool's impact on acute stroke workflow and clinical outcomes.

Materials and methods

Consecutive patients with a computed tomography angiography (CTA) presenting with suspected acute ischemic stroke were compared before and after the implementation of an AI tool, RAPID LVO (RAPID 4.9, iSchemaView, Menlo Park, CA). Radiology CTA report turnaround times (TAT), door-to-treatment times, and the NIH stroke scale (NIHSS) after treatment were evaluated.

Results

A total of 439 cases in the pre-AI group and 321 cases in the post-AI group were included, with 62 (14.12%) and 43 (13.40%) cases, respectively, receiving acute therapies. The AI tool demonstrated a sensitivity of 0.96, a specificity of 0.85, a negative predictive value of 0.99, and a positive predictive value of 0.53. Radiology CTA report TAT significantly improved post-AI (mean 30.58 min for pre-AI vs. 22 min for post-AI, p < 0.0005), notably at the resident level (p < 0.0003) but not at higher levels of expertise. There were no differences in door-to-treatment times, but the NIHSS at discharge was improved for the pre-AI group adjusted for confounders (parameter estimate = 3.97, p < 0.01).

Conclusion

Implementation of an automated LVO detection tool improved radiology TAT but did not translate to improved stroke metrics and outcomes in a real-world setting.

Big Data in Stroke: How to Use Big Data to Make the Next Management Decision

The last decade has seen significant advances in the accumulation of medical data, the computational techniques to analyze that data, and corresponding improvements in management. Interventions such as thrombolytics and mechanical thrombectomy improve patient outcomes after stroke in selected patients; however, significant gaps remain in our ability to select patients, predict complications, and understand outcomes. Big data and the computational methods needed to analyze it can address these gaps. For example, automated analysis of neuroimaging to estimate the volume of brain tissue that is ischemic and salvageable can help triage patients for acute interventions. Data-intensive computational techniques can perform complex risk calculations that are too cumbersome to be completed by humans, resulting in more accurate and timely prediction of which patients require increased vigilance for adverse events such as treatment complications. To handle the accumulation of complex medical data, a variety of advanced computational techniques referred to as machine learning and artificial intelligence now routinely complement traditional statistical inference. In this narrative review, we explore data-intensive techniques in stroke research, how it has informed the management of stroke patients, and how current work could shape clinical practice in the future.

Benchmark dataset for clot detection in ischemic stroke vessel-based imaging: CODEC-IV

We present an annotated dataset for the purposes of creating a benchmark in Artificial Intelligence for automated clot detection. While there are commercial tools available for automated clot detection on computed tomographic (CT) angiographs, they have not been compared in a standardized manner whereby accuracy is reported on a publicly available benchmark dataset. Furthermore, there are known difficulties in automated clot detection - namely, cases where there is robust collateral flow, or residual flow and occlusions of the smaller vessels - and it is necessary to drive an initiative to overcome these challenges. Our dataset contains 159 multiphase CTA patient datasets, derived from CTP and annotated by expert stroke neurologists. In addition to images where the clot is marked, the expert neurologists have provided information about clot location, hemisphere and the degree of collateral flow. The data is available on request by researchers via an online form, and we will host a leaderboard where the results of clot detection algorithms on the dataset will be displayed. Participants are invited to submit an algorithm to us for evaluation using the evaluation tool, which is made available at together with the form at https://github.com/MBC-Neuroimaging/ClotDetectEval.

Machine-learning algorithm in acute stroke: real-world experience

AIM

To assess the clinical performance of a commercially available machine learning (ML) algorithm in acute stroke.

MATERIALS AND METHODS

CT and CT angiography (CTA) studies of 104 consecutive patients (43 females, age range 19-93, median age 62) performed for suspected acute stroke at a single tertiary institution with real-time ML software analysis (RAPID™ ASPECTS and CTA) were included. Studies were retrospectively reviewed independently by two neuroradiologists in a blinded manner.

RESULTS

The cohort included 24 acute infarcts and 16 large vessel occlusions (LVO). RAPID™ ASPECTS interpretation demonstrated high sensitivity (87.5%) and NPV (87.5%) but very poor specificity (30.9%) and PPV (30.9%) for detection of acute ischaemic parenchymal changes. There was a high percentage of false positives (51.1%). In cases of proven LVO, RAPID™ ASPECTS showed good correlation with neuroradiologists' blinded independent interpretation, Pearson correlation coefficient = 0.96 (both readers), 0.63 (RAPID™ vs reader 1), 0.69 (RAPID™ vs reader 2). RAPID™ CTA interpretation demonstrated high sensitivity (92.3%), specificity (85.3%), and negative predictive (NPV) (98.5%) with moderate positive predictive value (PPV) (52.2%) for detection of LVO (N=13). False positives accounted for 12.5% of cases, of which 27.3% were attributed to arterial stenosis.

CONCLUSION

RAPID™ CTA was robust and reliable in detection of LVO. Although demonstrating high sensitivity and NPV, RAPID™ ASPECTS interpretation was associated with a high number of false positives, which decreased clinicians' confidence in the algorithm. However, in cases of proven LVO, RAPID™ ASPECTS performed well and had good correlation with neuroradiologists' blinded interpretation.

AI software detection of large vessel occlusion stroke on CT angiography: a real-world prospective diagnostic test accuracy study

BACKGROUND

Artificial intelligence (AI) software is increasingly applied in stroke diagnostics. However, the actual performance of AI tools for identifying large vessel occlusion (LVO) stroke in real time in a real-world setting has not been fully studied.

OBJECTIVE

To determine the accuracy of AI software in a real-world, three-tiered multihospital stroke network.

METHODS

All consecutive head and neck CT angiography (CTA) scans performed during stroke codes and run through an AI software engine (Viz LVO) between May 2019 and October 2020 were prospectively collected. CTA readings by radiologists served as the clinical reference standard test and Viz LVO output served as the index test. Accuracy metrics were calculated.

RESULTS

Of a total of 1822 CTAs performed, 190 occlusions were identified; 142 of which were internal carotid artery terminus (ICA-T), middle cerebral artery M1, or M2 locations. Accuracy metrics were analyzed for two different groups: ICA-T and M1 ±M2. For the ICA-T/M1 versus the ICA-T/M1/M2 group, sensitivity was 93.8% vs 74.6%, specificity was 91.1% vs 91.1%, negative predictive value was 99.7% vs 97.6%, accuracy was 91.2% vs 89.8%, and area under the curve was 0.95 vs 0.86, respectively. Detection rates for ICA-T, M1, and M2 occlusions were 100%, 93%, and 49%, respectively. As expected, the algorithm offered better detection rates for proximal occlusions than for mid/distal M2 occlusions (58% vs 28%, p=0.03).

CONCLUSIONS

These accuracy metrics support Viz LVO as a useful adjunct tool in stroke diagnostics. Fast and accurate diagnosis with high negative predictive value mitigates missing potentially salvageable patients.

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

OBJECTIVES

Automated 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 METHODS

Retrospective 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.

RESULTS

Using 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).

CONCLUSION

Our findings suggest that accurate image-level LVO detection is feasible on CTP raw images.

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

Purpose

Despite 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 methods

This 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.

Results

There 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.

Conclusion

Both 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.

Validation of a machine learning software tool for automated large vessel occlusion detection in patients with suspected acute stroke

PURPOSE

CT angiography (CTA) is the imaging standard for large vessel occlusion (LVO) detection in patients with acute ischemic stroke. StrokeSENS LVO is an automated tool that utilizes a machine learning algorithm to identify anterior large vessel occlusions (LVO) on CTA. The aim of this study was to test the algorithm's performance in LVO detection in an independent dataset.

METHODS

A total of 400 studies (217 LVO, 183 other/no occlusion) read by expert consensus were used for retrospective analysis. The LVO was defined as intracranial internal carotid artery (ICA) occlusion and M1 middle cerebral artery (MCA) occlusion. Software performance in detecting anterior LVO was evaluated using receiver operator characteristics (ROC) analysis, reporting area under the curve (AUC), sensitivity, and specificity. Subgroup analyses were performed to evaluate if performance in detecting LVO differed by subgroups, namely M1 MCA and ICA occlusion sites, and in data stratified by patient age, sex, and CTA acquisition characteristics (slice thickness, kilovoltage tube peak, and scanner manufacturer).

RESULTS

AUC, sensitivity, and specificity overall were as follows: 0.939, 0.894, and 0.874, respectively, in the full cohort; 0.927, 0.857, and 0.874, respectively, in the ICA occlusion cohort; 0.945, 0.914, and 0.874, respectively, in the M1 MCA occlusion cohort. Performance did not differ significantly by patient age, sex, or CTA acquisition characteristics.

CONCLUSION

The StrokeSENS LVO machine learning algorithm detects anterior LVO with high accuracy from a range of scans in a large dataset.

Deep learning-based classification of DSA image sequences of patients with acute ischemic stroke

PURPOSE

Recently, a large number of patients with acute ischemic stroke benefited from the use of thrombectomy, a minimally invasive intervention technique for mechanically removing thrombi from the cerebrovasculature. During thrombectomy, 2D digital subtraction angiography (DSA) image sequences are acquired simultaneously from the posterior-anterior and the lateral view to control whether thrombus removal was successful, and to possibly detect newly occluded areas caused by thrombus fragments split from the main thrombus. However, such new occlusions, which would be treatable by thrombectomy, may be overlooked during the intervention. To prevent this, we developed a deep learning-based approach to automatic classification of DSA sequences into thrombus-free and non-thrombus-free sequences.

METHODS

We performed a retrospective study based on the single-center DSA data of thrombectomy patients. For classifying the DSA sequences, we applied Long Short-Term Memory or Gated Recurrent Unit networks and combined them with different Convolutional Neural Networks used as feature extractor. These network variants were trained on the DSA data by using five-fold cross-validation. The classification performance was determined on a test data set with respect to the Matthews correlation coefficient (MCC) and the area under the curve (AUC). Finally, we evaluated our models on patient cases, in which overlooking thrombi during thrombectomy had happened.

RESULTS

Depending on the specific model configuration used, we obtained a performance of up to 0.77[Formula: see text]0.94 for the MCC[Formula: see text]AUC, respectively. Additionally, overlooking thrombi could have been prevented in the reported patient cases, as our models would have classified the corresponding DSA sequences correctly.

CONCLUSION

Our deep learning-based approach to thrombus identification in DSA sequences yielded high accuracy on our single-center test data set. External validation is now required to investigate the generalizability of our method. As demonstrated, using this new approach may help reduce the incident risk of overlooking thrombi during thrombectomy in the future.

Hyperdense Artery Sign in Patients With Acute Ischemic Stroke-Automated Detection With Artificial Intelligence-Driven Software

Background

Hyperdense artery sign (HAS) on non-contrast CT (NCCT) can indicate a large vessel occlusion (LVO) in patients with acute ischemic stroke. HAS detection belongs to routine reporting in patients with acute stroke and can help to identify patients in whom LVO is not initially suspected. We sought to evaluate automated HAS detection by commercial software and compared its performance to that of trained physicians against a reference standard.

Methods

Non-contrast CT scans from 154 patients with and without LVO proven by CT angiography (CTA) were independently rated for HAS by two blinded neuroradiologists and an AI-driven algorithm (Brainomix®). Sensitivity and specificity were analyzed for the clinicians and the software. As a secondary analysis, the clot length was automatically calculated by the software and compared with the length manually outlined on CTA images as the reference standard.

Results

Among 154 patients, 84 (54.5%) had CTA-proven LVO. HAS on the correct side was detected with a sensitivity and specificity of 0.77 (CI:0.66–0.85) and 0.87 (0.77–0.94), 0.8 (0.69–0.88) and 0.97 (0.89–0.99), and 0.93 (0.84–0.97) and 0.71 (0.59–0.81) by the software and readers 1 and 2, respectively. The automated estimation of the thrombus length was in moderate agreement with the CTA-based reference standard [intraclass correlation coefficient (ICC) 0.73].

Conclusion

Automated detection of HAS and estimation of thrombus length on NCCT by the tested software is feasible with a sensitivity and specificity comparable to that of trained neuroradiologists.

Fully Automated Thrombus Segmentation on CT Images of Patients with Acute Ischemic Stroke

Thrombus imaging characteristics are associated with treatment success and functional outcomes in stroke patients. However, assessing these characteristics based on manual annotations is labor intensive and subject to observer bias. Therefore, we aimed to create an automated pipeline for consistent and fast full thrombus segmentation. We used multi-center, multi-scanner datasets of anterior circulation stroke patients with baseline NCCT and CTA for training (n = 228) and testing (n = 100). We first found the occlusion location using StrokeViewer LVO and created a bounding box around it. Subsequently, we trained dual modality U-Net based convolutional neural networks (CNNs) to segment the thrombus inside this bounding box. We experimented with: (1) U-Net with two input channels for NCCT and CTA, and U-Nets with two encoders where (2) concatenate, (3) add, and (4) weighted-sum operators were used for feature fusion. Furthermore, we proposed a dynamic bounding box algorithm to adjust the bounding box. The dynamic bounding box algorithm reduces the missed cases but does not improve Dice. The two-encoder U-Net with a weighted-sum feature fusion shows the best performance (surface Dice 0.78, Dice 0.62, and 4% missed cases). Final segmentation results have high spatial accuracies and can therefore be used to determine thrombus characteristics and potentially benefit radiologists in clinical practice.

Artificial intelligence software for diagnosing intracranial arterial occlusion in patients with acute ischemic stroke

PURPOSE

To evaluate the diagnostic performance of AI software in diagnosing intracranial arterial occlusions in the proximal anterior circulation at CT angiography (CTA) and to compare it to manual reading performed in clinical practice.

METHODS

Patients with acute ischemic stroke underwent CTA to detect arterial occlusion in the proximal anterior circulation. Retrospective review of CTA scans by two neuroradiologists served as reference standard. Sensitivity and specificity of AI software (StrokeViewer) were compared to those of manual reading using the McNemar test. The proportions of correctly detected occlusions in the distal internal carotid artery and/or M1 segment of the middle cerebral artery (large vessel occlusion [LVO]) and in the M2 segment of the middle cerebral artery (medium vessel occlusion [MeVO]) were calculated.

RESULTS

Of the 474 patients, 75 (15.8%) had an arterial occlusion in the proximal anterior circulation according to the reference standard. Sensitivity of StrokeViewer software was not significantly different compared to that of manual reading (77.3% vs. 78.7%, P = 1.000). Specificity of StrokeViewer software was significantly lower than that of manual reading (88.5% vs. 100%, P < 0.001). StrokeViewer software correctly identified 40 of 42 LVOs (95.2%) and 18 of 33 MeVOs (54.5%). StrokeViewer software detected 8 of 16 (50%) intracranial arterial occlusions which were missed by manual reading.

CONCLUSION

The current AI software detected intracranial arterial occlusion with moderate sensitivity and fairly high specificity. The AI software may detect additional occlusions which are missed by manual reading. As such, the use of AI software may be of value in clinical stroke care.

Emerging Detection Techniques for Large Vessel Occlusion Stroke: A Scoping Review

Background: Large vessel occlusion (LVO) is the obstruction of large, proximal cerebral arteries and can account for up to 46% of acute ischaemic stroke (AIS) when both the A2 and P2 segments are included (from the anterior and posterior cerebral arteries). It is of paramount importance that LVO is promptly recognised to provide timely and effective acute stroke management. This review aims to scope recent literature to identify new emerging detection techniques for LVO. As a good comparator throughout this review, the commonly used National Institutes of Health Stroke Scale (NIHSS), at a cut-off of ≥11, has been reported to have a sensitivity of 86% and a specificity of 60% for LVO. Methods: Four electronic databases (Medline via OVID, CINAHL, Scopus, and Web of Science), and grey literature using OpenGrey, were systematically searched for published literature investigating developments in detection methods for LVO, reported from 2015 to 2021. The protocol for the search was published with the Open Science Framework (10.17605/OSF.IO/A98KN). Two independent researchers screened the titles, abstracts, and full texts of the articles, assessing their eligibility for inclusion. Results: The search identified 5,082 articles, in which 2,265 articles were screened to assess their eligibility. Sixty-two studies remained following full-text screening. LVO detection techniques were categorised into 5 groups: stroke scales (n = 30), imaging and physiological methods (n = 15), algorithmic and machine learning approaches (n = 9), physical symptoms (n = 5), and biomarkers (n = 3). Conclusions: This scoping review has explored literature on novel and advancements in pre-existing detection methods for LVO. The results of this review highlight LVO detection techniques, such as stroke scales and biomarkers, with good sensitivity and specificity performance, whilst also showing advancements to support existing LVO confirmatory methods, such as neuroimaging.

Automated Large Artery Occlusion Detection in Stroke: A Single-Center Validation Study of an Artificial Intelligence Algorithm

INTRODUCTION

Expediting notification of lesions in acute ischemic stroke (AIS) is critical. Limited availability of experts to assess such lesions and delays in large vessel occlusion (LVO) recognition can negatively affect outcomes. Artificial intelligence (AI) may aid LVO recognition and treatment. This study aims to evaluate the performance of an AI-based algorithm for LVO detection in AIS.

METHODS

Retrospective analysis of a database of AIS patients admitted in a single center between 2014 and 2019. Vascular neurologists graded computed tomography angiographies (CTAs) for presence and site of LVO. Studies were analyzed by the Viz-LVO Algorithm® version 1.4 - neural network programmed to detect occlusions from the internal carotid artery terminus (ICA-T) to the Sylvian fissure. Comparisons between human versus AI-based readings were done by test characteristic analysis and Cohen's kappa. Primary analysis included ICA-T and/or middle cerebral artery (MCA)-M1 LVOs versus non-LVOs/more distal occlusions. Secondary analysis included MCA-M2 occlusions.

RESULTS

610 CTAs were analyzed. The AI algorithm rejected 2.5% of the CTAs due to poor quality, which were excluded from the analysis. Viz-LVO identified ICA-T and MCA-M1 LVOs with a sensitivity of 87.6%, specificity of 88.5%, and accuracy of 87.9% (AUC 0.88, 95% CI: 0.85-0.92, p < 0.001). Cohen's kappa was 0.74. In the secondary analysis, the algorithm yielded a sensitivity of 80.3%, specificity of 88.5%, and accuracy of 82.7%. The mean run time of the algorithm was 2.78 ± 0.5 min.

CONCLUSION

Automated AI reading allows for fast and accurate identification of LVO strokes with timely notification to emergency teams, enabling quick decision-making for reperfusion therapies or transfer to specialized centers if needed.

Positive predictive value and stroke workflow outcomes using automated vessel density (RAPID-CTA) in stroke patients: One year experience

PURPOSE

Several new techniques have emerged for detecting anterior circulation large vessel occlusion by quantifying relative vessel density including RAPID-CTA, potentially allowing for faster triage and decreased time to mechanical thrombectomy. We present our one-year experience on positive predictive value of RAPID-CTA for the detection of large vessel occlusion in patients presenting with stroke symptoms and its effect on treatment time and clinical outcomes.

MATERIALS AND METHODS

Three hundred and ten patients presenting with stroke symptoms with relative vessel density <60% on RAPID-CTA were included (average age 70 years, 145 male, 165 female). Examinations were considered positive if there was evidence of large vessel occlusion or high grade stenosis. Computed tomography angiography to groin puncture time was calculated during one-year time intervals before and after RAPID-CTA installation. Ninety-day Modified Rankin Scale scores were obtained for patients in each cohort.

RESULTS

Of the 310 patients, 270 had large vessel occlusion or high grade stenosis (87% positive predictive value), with 161 having large vessel occlusion. Using 45% relative vessel density threshold, 129/161 large vessel occlusion were detected (80% sensitivity) and 163/172 examinations were positive (95% positive predictive value). Computed tomography angiography to groin puncture time was significantly lower after deployment of RAPID-CTA (93 min vs 68 min, p<0.05). Average 90 day modified Rankin Scale score was lower in the RAPID-CTA group with a higher percentage of patients with functional independence, although the data was not statistically significant.

CONCLUSION

RAPID-CTA had high positive predictive value for large vessel occlusion with a 45% relative vessel density threshold, which could facilitate active worklist reprioritization. Time to treatment was significantly lower and clinical outcomes were improved after deployment of RAPID-CTA.

Comparison of convolutional neural networks for detecting large vessel occlusion on computed tomography angiography

PURPOSE

Artificial intelligence diagnosis and triage of large vessel occlusion may quicken clinical response for a subset of time-sensitive acute ischemic stroke patients, improving outcomes. Differences in architectural elements within data-driven convolutional neural network (CNN) models impact performance. Foreknowledge of effective model architectural elements for domain-specific problems can narrow the search for candidate models and inform strategic model design and adaptation to optimize performance on available data. Here, we study CNN architectures with a range of learnable parameters and which span the inclusion of architectural elements, such as parallel processing branches and residual connections with varying methods of recombining residual information.

METHODS

We compare five CNNs: ResNet-50, DenseNet-121, EfficientNet-B0, PhiNet, and an Inception module-based network, on a computed tomography angiography large vessel occlusion detection task. The models were trained and preliminarily evaluated with 10-fold cross-validation on preprocessed scans (n = 240). An ablation study was performed on PhiNet due to superior cross-validated test performance across accuracy, precision, recall, specificity, and F1 score. The final evaluation of all models was performed on a withheld external validation set (n = 60) and these predictions were subsequently calibrated with sigmoid curves.

RESULTS

Uncalibrated results on the withheld external validation set show that DenseNet-121 had the best average performance on accuracy, precision, recall, specificity, and F1 score. After calibration DenseNet-121 maintained superior performance on all metrics except recall.

CONCLUSIONS

The number of learnable parameters in our five models and best-ablated PhiNet directly related to cross-validated test performance-the smaller the model the better. However, this pattern did not hold when looking at generalization on the withheld external validation set. DenseNet-121 generalized the best; we posit this was due to its heavy use of residual connections utilizing concatenation, which causes feature maps from earlier layers to be used deeper in the network, while aiding in gradient flow and regularization.

Cost-effectiveness of artificial intelligence aided vessel occlusion detection in acute stroke: an early health technology assessment

Background

Limited evidence is available on the clinical impact of artificial intelligence (AI) in radiology. Early health technology assessment (HTA) is a methodology to assess the potential value of an innovation at an early stage. We use early HTA to evaluate the potential value of AI software in radiology. As a use-case, we evaluate the cost-effectiveness of AI software aiding the detection of intracranial large vessel occlusions (LVO) in stroke in comparison to standard care. We used a Markov based model from a societal perspective of the United Kingdom predominantly using stroke registry data complemented with pooled outcome data from large, randomized trials. Different scenarios were explored by varying missed diagnoses of LVOs, AI costs and AI performance. Other input parameters were varied to demonstrate model robustness. Results were reported in expected incremental costs (IC) and effects (IE) expressed in quality adjusted life years (QALYs).

Results

Applying the base case assumptions (6% missed diagnoses of LVOs by clinicians, $40 per AI analysis, 50% reduction of missed LVOs by AI), resulted in cost-savings and incremental QALYs over the projected lifetime (IC: − $156, − 0.23%; IE: + 0.01 QALYs, + 0.07%) per suspected ischemic stroke patient. For each yearly cohort of patients in the UK this translates to a total cost saving of $11 million.

Conclusions

AI tools for LVO detection in emergency care have the potential to improve healthcare outcomes and save costs. We demonstrate how early HTA may be applied for the evaluation of clinically applied AI software for radiology.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13244-021-01077-4.

Automated emergent large vessel occlusion detection by artificial intelligence improves stroke workflow in a hub and spoke stroke system of care

BACKGROUND

Emergent large vessel occlusion (ELVO) acute ischemic stroke is a time-sensitive disease.

OBJECTIVE

To describe our experience with artificial intelligence (AI) for automated ELVO detection and its impact on stroke workflow.

METHODS

We conducted a retrospective chart review of code stroke cases in which VizAI was used for automated ELVO detection. Patients with ELVO identified by VizAI were compared with patients with ELVO identified by usual care. Details of treatment, CT angiography (CTA) interpretation by blinded neuroradiologists, and stroke workflow metrics were collected. Univariate statistical comparisons and linear regression analysis were performed to quantify time savings for stroke metrics.

RESULTS

Six hundred and eighty consecutive code strokes were evaluated by AI; 104 patients were diagnosed with ELVO during the study period. Forty-five patients with ELVO were identified by AI and 59 by usual care. Sixty-nine mechanical thrombectomies were performed.Median time from CTA to team notification was shorter for AI ELVOs (7 vs 26 min; p<0.001). Door to arterial puncture was faster for transfer patients with ELVO detected by AI versus usual care transfer patients (141 vs 185 min; p=0.027). AI yielded a time savings of 22 min for team notification and a 23 min reduction in door to arterial puncture for transfer patients.

CONCLUSIONS

AI automated alerts can be incorporated into a comprehensive stroke center hub and spoke system of care. The use of AI to detect ELVO improves clinically meaningful stroke workflow metrics, resulting in faster treatment times for mechanical thrombectomy.

Diagnostic accuracy of automated occlusion detection in CT angiography using e-CTA

Background and aim

The aim of this study was to assess the diagnostic accuracy of e-CTA (product name) (Brainomix) in the automatic detection of large vessel occlusions in anterior circulation stroke.

Methods

Of 487 CT angiographies from patients with large vessel occlusions stroke, 327 were used to train the algorithm while the remaining cases together with 140 negative CT angiographies were used to validate its performance against ground truth. Of these 301 cases, 144 were randomly selected and used for an additional comparative analysis against 4 raters. Sensitivity, specificity, positive and negative predictive value (PPV and NPV), accuracy and level of agreement with ground truth (Cohen’s Kappa) were determined and compared to the performance of a neuroradiologist, a radiology resident, and two neurology residents.

Results

e-CTA had a sensitivity and specificity of 0.84 (0.77–0.89) and 0.96 (0.91–0.98) respectively for the detection of any large vessel occlusions on the correct side in the whole validation cohort. This performance was identical in the comparative analysis subgroup and was within the range of physicians at different levels of expertise: 0.86–0.97 and 0.91–1.00, respectively. For the detection of proximal occlusions, it was 0.92 (0.84–0.96) and 0.98 (0.94–1.00) for the whole cohort and 0.93 (0.80–0.98) and 1.00 (0.95–1.00) for the comparative analysis, respectively for e-CTA. The range was 0.8–0.97 for sensitivity and 0.97–1.00 for specificity for the four physicians.

Conclusions

The performance of e-CTA in detecting any large vessel occlusions is comparable to less experienced physicians but is similar to experienced physicians for detecting proximal large vessel occlusions.

CTA Protocols in a Telestroke Network Improve Efficiency for Both Spoke and Hub Hospitals

BACKGROUND AND PURPOSE

Telestroke networks support screening for patients with emergent large-vessel occlusions who are eligible for endovascular thrombectomy. Ideal triage processes within telestroke networks remain uncertain. We characterize the impact of implementing a routine spoke hospital CTA protocol in our integrated telestroke network on transfer and thrombectomy patterns.

MATERIALS AND METHODS

A protocol-driven CTA process was introduced at 22 spoke hospitals in November 2017. We retrospectively identified prospectively collected patients who presented to a spoke hospital with National Institutes of Health Stroke Scale scores ≥6 between March 1, 2016 and March 1, 2017 (pre-CTA), and March 1, 2018 and March 1, 2019 (post-CTA). We describe the demographics, CTA utilization, spoke hospital retention rates, emergent large-vessel occlusion identification, and rates of endovascular thrombectomy.

RESULTS

There were 167 patients pre-CTA and 207 post-CTA. The rate of CTA at spoke hospitals increased from 15% to 70% (P < .001). Despite increased endovascular thrombectomy screening in the extended window, the overall rates of transfer out of spoke hospitals remained similar (56% versus 54%; P = .83). There was a nonsignificant increase in transfers to our hub hospital for endovascular thrombectomy (26% versus 35%; P = .12), but patients transferred >4.5 hours from last known well increased nearly 5-fold (7% versus 34%; P < .001). The rate of endovascular thrombectomy performed on patients transferred for possible endovascular thrombectomy more than doubled (22% versus 47%; P = .011).

CONCLUSIONS

Implementation of CTA at spoke hospitals in our telestroke network was feasible and improved the efficiency of stroke triage. Rates of patients retained at spoke hospitals remained stable despite higher numbers of patients screened. Emergent large-vessel occlusion confirmation at the spoke hospital lead to a more than 2-fold increase in thrombectomy rates among transferred patients at the hub.

High-Performance Automated Anterior Circulation CT Angiographic Clot Detection in Acute Stroke: A Multireader Comparison

Background Identification of large vessel occlusion (LVO) is critical to the management of acute ischemic stroke and prerequisite to endovascular therapy in recent trials. Increasing volumes and data complexity compel the development of fast, reliable, and automated tools for LVO detection to facilitate acute imaging triage. Purpose To investigate the performance of an anterior circulation LVO detection platform in a large mixed sample of individuals with and without LVO at cerebrovascular CT angiography (CTA). Materials and Methods In this retrospective analysis, CTA data from recent cerebrovascular trials (CRISP [ClinicalTrials.gov NCT01622517] and DASH) were enriched with local repositories from 11 worldwide sites to balance demographic and technical variables in LVO-positive and LVO-negative examinations. CTA findings were reviewed independently by two neuroradiologists from different institutions for intracranial internal carotid artery (ICA) or middle cerebral artery (MCA) M1 LVO; these observers were blinded to all clinical variables and outcomes. An automated analysis platform was developed and tested for prediction of LVO presence and location relative to reader consensus. Discordance between readers with respect to LVO presence or location was adjudicated by a blinded tertiary reader at a third institution. Sensitivity, specificity, and receiver operating characteristics were assessed by an independent statistician, and subgroup analyses were conducted. Prespecified performance thresholds were set at a lower bound of the 95% CI of sensitivity and specificity of 0.8 or greater at mean times to notification of less than 3.5 minutes. Results A total of 217 study participants (mean age, 64 years ± 16 [standard deviation]; 116 men; 109 with positive findings of LVO) were evaluated. Prespecified performance thresholds were exceeded (sensitivity, 105 of 109 [96%; 95% CI: 91, 99]; specificity, 106 of 108 [98%; 95% CI: 94, 100]). Sensitivity and specificity estimates across age, sex, location, and vendor subgroups exceeded 90%. The area under the receiver operating characteristic curve was 99% (95% CI: 97, 100). Mean processing and notification time was 3 minutes 18 seconds. Conclusion The results confirm the feasibility of fast automated high-performance detection of intracranial internal carotid artery and middle cerebral artery M1 occlusions. © RSNA, 2021 See also the editorial by Kloska in this issue.

Diagnostic performance of single-phase CT angiography in detecting large vessel occlusion in ischemic stroke: A systematic review

PURPOSE

To systematically review the diagnostic performance of single-phase CT angiography (CTA) in detecting intracranial large vessel occlusion (LVO).

METHOD

MEDLINE and Embase were searched for studies investigating the diagnostic performance of single-phase CTA in detecting LVO. Study quality was assessed. Sensitivity and specificity were calculated and meta-analyzed with a bivariate random-effects model. Heterogeneity was assessed with a chi-squared test.

RESULTS

Eleven studies were included. High risk of bias with regard to "patient selection", "reference standard", and "flow and timing" was present in 4, 1, and 2 studies, respectively. In 7 studies, it was unclear whether reference tests were interpreted blinded to CTA readings. There was variability in types of vessel segments analyzed, resulting in heterogeneous sensitivity and specificity (P < 0.05). Two studies provided data for the proximal anterior circulation (distal intracranial carotid artery, A1-, A2-, M1- and M2-segments), with pooled sensitivity of 88.4 % (95 % CI: 62.2-97.2 %) and pooled specificity of 98.5 % (95 % CI: 33.2-100 %). One study suggested that multiphase CTA improved agreement between nonexperts and an expert in detecting A1-, A2-, M1-, M2-, and M3-segment occlusions compared to single-phase CTA (ĸ = 0.72-0.76 vs. ĸ = 0.32-0.45). No other included study reported added value of advanced CTA (CT perfusion, 4D-CTA, or multiphase CTA) compared to single-phase CTA in detecting proximal anterior circulation LVO.

CONCLUSION

There is lack of high-quality studies on the diagnostic performance of single-phase CTA for LVO detection in the proximal anterior circulation. The added value of advanced CTA techniques in detecting proximal anterior circulation LVO is not completely clear yet.

Artificial Intelligence Applications for Workflow, Process Optimization and Predictive Analytics

There is great potential for artificial intelligence (AI) applications, especially machine learning and natural language processing, in medical imaging. Much attention has been garnered by the image analysis tasks for diagnostic decision support and precision medicine, but there are many other potential applications of AI in radiology and have potential to enhance all levels of the radiology workflow and practice, including workflow optimization and support for interpretation tasks, quality and safety, and operational efficiency. This article reviews the important potential applications of informatics and AI related to process improvement and operations in the radiology department.

An East Coast Perspective on Artificial Intelligence and Machine Learning: Part 2: Ischemic Stroke Imaging and Triage

Acute ischemic stroke constitutes approximately 85% of strokes. Most strokes occur in community settings; thus, automatic algorithms techniques are attractive for managing these cases. This article reviews the use of deep learning convolutional neural networks in the management of ischemic stroke. Artificial intelligence-based algorithms may be used in patient triage to detect and sound the alarm based on early imaging, alert care teams, and assist in treatment selection. This article reviews algorithms for artificial intelligence techniques that may be used to detect and localize acute ischemic stroke. We describe artificial intelligence algorithms for these tasks and illustrate them with examples.

Deep Learning Based Software to Identify Large Vessel Occlusion on Noncontrast Computed Tomography

Background and Purpose:

Reliable recognition of large vessel occlusion (LVO) on noncontrast computed tomography (NCCT) may accelerate identification of endovascular treatment candidates. We aim to validate a machine learning algorithm (MethinksLVO) to identify LVO on NCCT.

Methods:

Patients with suspected acute stroke who underwent NCCT and computed tomography angiography (CTA) were included. Software detection of LVO (MethinksLVO) on NCCT was tested against the CTA readings of 2 experienced radiologists (NR-CTA). We used a deep learning algorithm to identify clot signs on NCCT. The software image output trained a binary classifier to determine LVO on NCCT. We studied software accuracy when adding National Institutes of Health Stroke Scale and time from onset to the model (MethinksLVO+).

Results:

From 1453 patients, 823 (57%) had LVO by NR-CTA. The area under the curve for the identification of LVO with MethinksLVO was 0.87 (sensitivity: 83%, specificity: 71%, positive predictive value: 79%, negative predictive value: 76%) and improved to 0.91 with MethinksLVO+ (sensitivity: 83%, specificity: 85%, positive predictive value: 88%, negative predictive value: 79%).

Conclusions:

In patients with suspected acute stroke, MethinksLVO software can rapidly and reliably predict LVO. MethinksLVO could reduce the need to perform CTA, generate alarms, and increase the efficiency of patient transfers in stroke networks.

Radiology workload in clinical implementation of thrombectomy for acute ischemic stroke: experience from The Netherlands

PURPOSE

To investigate the number of acute stroke patients undergoing CT angiography (CTA) for suspected large vessel occlusion (LVO) and those eligible for thrombectomy in relation to the population.

METHODS

Consecutive patients in a Western population who underwent CTA for suspected LVO of the proximal anterior circulation between January and August 2019 were included. The date and time of CTA and the number of patients eligible for thrombectomy were assessed. Our hospital's service area population was estimated using the Central Bureau for Statistics data. One-way analysis of variance with post-hoc tests and chi-squared tests were used for statistical analyses.

RESULTS

Of 520 patients (49% males, mean age of 72 years) undergoing CTA, 84 (16.2%) were eligible for thrombectomy. Our hospital's service area population was estimated at 420,000. Therefore, 3.6 CTA scans were performed and 0.6 patients were eligible for thrombectomy per 100,000 people per week. The number of patients undergoing CTA and the number of patients eligible for thrombectomy both did not significantly differ between any days of the week (P > 0.05). A total of 236 (45%) and 284 patients (55%) underwent CTA during office and on-call hours, respectively. The percentage of patients eligible for thrombectomy did not significantly differ between office and on-call hours (P = 0.834).

CONCLUSION

Our study estimated the number of stroke patients undergoing CTA for suspected LVO and those eligible for thrombectomy in relation to the population. Numbers were essentially the same throughout the week, and during office and on-call hours. Our data can be used to make adequate staffing plans.