Accuracy of Automated Computer-Aided Diagnosis for Stroke Imaging: A Critical Evaluation of Current Evidence

APIS: a paired CT-MRI dataset for ischemic stroke segmentation - methods and challenges

Stroke, the second leading cause of mortality globally, predominantly results from ischemic conditions. Immediate attention and diagnosis, related to the characterization of brain lesions, play a crucial role in patient prognosis. Standard stroke protocols include an initial evaluation from a non-contrast CT to discriminate between hemorrhage and ischemia. However, non-contrast CTs lack sensitivity in detecting subtle ischemic changes in this phase. Alternatively, diffusion-weighted MRI studies provide enhanced capabilities, yet are constrained by limited availability and higher costs. Hence, we idealize new approaches that integrate ADC stroke lesion findings into CT, to enhance the analysis and accelerate stroke patient management. This study details a public challenge where scientists applied top computational strategies to delineate stroke lesions on CT scans, utilizing paired ADC information. Also, it constitutes the first effort to build a paired dataset with NCCT and ADC studies of acute ischemic stroke patients. Submitted algorithms were validated with respect to the references of two expert radiologists. The best achieved Dice score was 0.2 over a test study with 36 patient studies. Despite all the teams employing specialized deep learning tools, results reveal limitations of computational approaches to support the segmentation of small lesions with heterogeneous density.

Extensive Multilabel Classification of Brain MRI Scans for Infarcts Using the Swin UNETR Architecture in Deep Learning Applications

OBJECTIVE

To distinguish infarct location and type with the utmost precision using the advantages of the Swin UNEt TRansformers (Swin UNETR) architecture.

METHODS

The research employed a two-phase training approach. In the first phase, the Swin UNETR model was trained using the Ischemic Stroke Lesion Segmentation Challenge (ISLES) 2022 dataset, which included cases of acute and subacute infarcts. The second phase involved training with data from 309 patients. The 110 categories result from classifying infarcts based on 22 specific brain regions. Each region is divided into right and left sides, and each side includes four types of infarcts (acute, acute lacunar, subacute, subacute lacunar). The unique architecture of Swin UNETR, integrating elements of both the transformer and u-net designs with a hierarchical transformer computed with shifted windows, played a crucial role in the study.

RESULTS

During Swin UNETR training with the ISLES 2022 dataset, batch loss decreased to 0.8885±0.1897, with training and validation dice scores reaching 0.4224±0.0710 and 0.4827±0.0607, respectively. The optimal model weight had a validation dice score of 0.5747. In the patient data model, batch loss decreased to 0.0565±0.0427, with final training and validation accuracies of 0.9842±0.0005 and 0.9837±0.0010.

CONCLUSION

The results of this study surpass the accuracy of similar studies, but they involve the issue of overfitting, highlighting the need for future efforts to improve generalizability. Such detailed classifications could significantly aid physicians in diagnosing infarcts in clinical settings.

Reducing the global burden of cerebral venous thrombosis: An international research agenda

BACKGROUND

Due to the rarity of cerebral venous thrombosis (CVT), performing high-quality scientific research in this field is challenging. Providing answers to unresolved research questions will improve prevention, diagnosis, and treatment, and ultimately translate to a better outcome of patients with CVT. We present an international research agenda, in which the most important research questions in the field of CVT are prioritized.

AIMS

This research agenda has three distinct goals: (1) to provide inspiration and focus to research on CVT for the coming years, (2) to reinforce international collaboration, and (3) to facilitate the acquisition of research funding.

SUMMARY OF REVIEW

This international research agenda is the result of a research summit organized by the International Cerebral Venous Thrombosis Consortium in Amsterdam, the Netherlands, in June 2023. The summit brought together 45 participants from 15 countries including clinical researchers from various disciplines, patients who previously suffered from CVT, and delegates from industry and non-profit funding organizations. The research agenda is categorized into six pre-specified themes: (1) epidemiology and clinical features, (2) life after CVT, (3) neuroimaging and diagnosis, (4) pathophysiology, (5) medical treatment, and (6) endovascular treatment. For each theme, we present two to four research questions, followed by a brief substantiation per question. The research questions were prioritized by the participants of the summit through consensus discussion.

CONCLUSIONS

This international research agenda provides an overview of the most burning research questions on CVT. Answering these questions will advance our understanding and management of CVT, which will ultimately lead to improved outcomes for CVT patients worldwide.

Assessment of Deep Learning-Based Triage Application for Acute Ischemic Stroke on Brain MRI in the ER

RATIONALE AND OBJECTIVES

To assess a deep learning application (DLA) for acute ischemic stroke (AIS) detection on brain magnetic resonance imaging (MRI) in the emergency room (ER) and the effect of T2-weighted imaging (T2WI) on its performance.

MATERIALS AND METHODS

We retrospectively analyzed brain MRIs taken through the ER from March to October 2021 that included diffusion-weighted imaging (DWI) and fluid-attenuated inversion recovery (FLAIR) sequences. MRIs were processed by the DLA, and sensitivity, specificity, accuracy, and area under the receiver operating characteristic curve (AUROC) were evaluated, with three neuroradiologists establishing the gold standard for detection performance. In addition, we examined the impact of axial T2WI, when available, on the accuracy and processing time of DLA.

RESULTS

The study included 947 individuals (mean age ± standard deviation, 64 years ± 16; 461 men, 486 women), with 239 (25%) positive for AIS. The overall performance of DLA was as follows: sensitivity, 90%; specificity, 89%; accuracy, 89%; and AUROC, 0.95. The average processing time was 24 s. In the subgroup with T2WI, T2WI did not significantly impact MRI assessments but did result in longer processing times (35 s without T2WI compared to 48 s with T2WI, p < 0.001).

CONCLUSION

The DLA successfully identified AIS in the ER setting with an average processing time of 24 s. The absence of performance acquire with axial T2WI suggests that the DLA can diagnose AIS with just axial DWI and FLAIR sequences, potentially shortening the exam duration in the ER.

A Narrative Review of Interhospital Transfers for Intracerebral Hemorrhage

Of the 750,000 strokes in the United States every year, 15% patients suffer from hemorrhagic stroke. Intracerebral hemorrhage (ICH) is a subtype of hemorrhagic stroke. Despite advances in acute management, patients with hemorrhagic stroke continue to suffer from high mortality and survivors suffer from multidomain impairments in the physical, cognitive, and mental health domains which could last for months to years from their index stroke. Long-term prognosis after ICH is critically dependent on the quality and efficacy of care a patient receives during the acute phase of care. With ongoing care consolidation in stroke systems of care, the number of ICH patients who need to undergo interhospital transfers (IHTs) is increasing. However, the associations between IHT and ICH outcomes have not been well described in literature. In this review, we describe the epidemiology of IHT for ICH, the relationship between IHT and ICH patient outcomes, and proposed improvements to the IHT process to ensure better long-term patient outcomes. Our review indicates that evidence regarding the safety and benefit of IHT for ICH patients is conflicting, with some studies reporting poorer outcomes for transferred patients compared to direct admissions via emergency rooms and other studies showing no effect on outcomes. The American Heart Association guidelines for ICH provide recommendations for timely blood pressure control and anticoagulation reversal to improve patient outcomes. The American Heart Association stroke systems of care guidelines provide recommendations for transfer agreements and but do not provide details on how patients should be managed while undergoing IHT. Large, prospective, and multicenter studies comparing outcomes of IHT patients to direct admissions are necessary to provide more definitive guidance to optimize IHT protocols and aid clinical decision-making.

The Role of Artificial Intelligence-Powered Imaging in Cerebrovascular Accident Detection

Cerebrovascular accidents (CVAs) often occur suddenly and abruptly, leaving patients with long-lasting disabilities that place a huge emotional and economic burden on everyone involved. CVAs result when emboli or thrombi travel to the brain and impede blood flow; the subsequent lack of oxygen supply leads to ischemia and eventually tissue infarction. The most important factor determining the prognosis of CVA patients is time, specifically the time from the onset of disease to treatment. Artificial intelligence (AI)-assisted neuroimaging alleviates the time constraints of analysis faced using traditional diagnostic imaging modalities, thus shortening the time from diagnosis to treatment. Numerous recent studies support the increased accuracy and processing capabilities of AI-assisted imaging modalities. However, the learning curve is steep, and huge barriers still exist preventing a full-scale implementation of this technology. Thus, the potential for AI to revolutionize medicine and healthcare delivery demands attention. This paper aims to elucidate the progress of AI-powered imaging in CVA diagnosis while considering traditional imaging techniques and suggesting methods to overcome adoption barriers in the hope that AI-assisted neuroimaging will be considered normal practice in the near future. There are multiple modalities for AI neuroimaging, all of which require collecting sufficient data to establish inclusive, accurate, and uniform detection platforms. Future efforts must focus on developing methods for data harmonization and standardization. Furthermore, transparency in the explainability of these technologies needs to be established to facilitate trust between physicians and AI-powered technology. This necessitates considerable resources, both financial and expertise wise which are not available everywhere.

Strengthening deep-learning models for intracranial hemorrhage detection: strongly annotated computed tomography images and model ensembles

Background and purpose

Multiple attempts at intracranial hemorrhage (ICH) detection using deep-learning techniques have been plagued by clinical failures. We aimed to compare the performance of a deep-learning algorithm for ICH detection trained on strongly and weakly annotated datasets, and to assess whether a weighted ensemble model that integrates separate models trained using datasets with different ICH improves performance.

Methods

We used brain CT scans from the Radiological Society of North America (27,861 CT scans, 3,528 ICHs) and AI-Hub (53,045 CT scans, 7,013 ICHs) for training. DenseNet121, InceptionResNetV2, MobileNetV2, and VGG19 were trained on strongly and weakly annotated datasets and compared using independent external test datasets. We then developed a weighted ensemble model combining separate models trained on all ICH, subdural hemorrhage (SDH), subarachnoid hemorrhage (SAH), and small-lesion ICH cases. The final weighted ensemble model was compared to four well-known deep-learning models. After external testing, six neurologists reviewed 91 ICH cases difficult for AI and humans.

Results

InceptionResNetV2, MobileNetV2, and VGG19 models outperformed when trained on strongly annotated datasets. A weighted ensemble model combining models trained on SDH, SAH, and small-lesion ICH had a higher AUC, compared with a model trained on all ICH cases only. This model outperformed four deep-learning models (AUC [95% C.I.]: Ensemble model, 0.953[0.938-0.965]; InceptionResNetV2, 0.852[0.828-0.873]; DenseNet121, 0.875[0.852-0.895]; VGG19, 0.796[0.770-0.821]; MobileNetV2, 0.650[0.620-0.680]; p < 0.0001). In addition, the case review showed that a better understanding and management of difficult cases may facilitate clinical use of ICH detection algorithms.

Conclusion

We propose a weighted ensemble model for ICH detection, trained on large-scale, strongly annotated CT scans, as no model can capture all aspects of complex tasks.

Real-world evaluation of Brainomix e-Stroke software

BACKGROUND AND PURPOSE

Brainomix e-Stroke is an artificial intelligence-based decision support tool that aids the interpretation of CT imaging in the context of acute stroke. While e-Stroke has the potential to improve the speed and accuracy of diagnosis, real-world validation is essential. The aim of this study was to prospectively evaluate the performance of Brainomix e-Stroke in an unselected cohort of patients with suspected acute ischaemic stroke.

METHODS

The study cohort included all patients admitted to the University College London Hospital Hyperacute Stroke Unit between October 2021 and April 2022. For e-ASPECTS and e-CTA, the ground truth was determined by a neuroradiologist with access to all clinical and imaging data. For e-CTP, the values of the core infarct and ischaemic penumbra were compared with those derived from syngo.via, an alternate software used at our institution.

RESULTS

1163 studies were performed in 551 patients admitted during the study period. Of these, 1130 (97.2%) were successfully processed by e-Stroke in an average of 4 min. For identifying acute middle cerebral artery territory ischaemia, e-ASPECTS had an accuracy of 77.0% and was more specific (83.5%) than sensitive (58.6%). The accuracy for identifying hyperdense thrombus was lower (69.1%), which was mainly due to many false positives (positive predictive value of 22.9%). Identification of acute haemorrhage was highly accurate (97.8%) with a sensitivity of 100% and a specificity of 97.6%; false positives were typically caused by areas of calcification. The accuracy of e-CTA for large vessel occlusions was 91.5%. The core infarct and ischaemic penumbra volumes provided by e-CTP strongly correlated with those provided by syngo.via (ρ=0.804-0.979).

CONCLUSION

Brainomix e-Stroke software provides rapid and reliable analysis of CT imaging in the acute stroke setting although, in line with the manufacturer's guidance, it should be used as an adjunct to expert interpretation rather than a standalone decision-making tool.

Automated advanced imaging in acute ischemic stroke. Certainties and uncertainties

The purpose of this is study was to review pearls and pitfalls of advanced imaging, such as computed tomography perfusion and diffusion-weighed imaging and perfusion-weighted imaging in the selection of acute ischemic stroke (AIS) patients suitable for endovascular treatment (EVT) in the late time window (6-24 h from symptom onset). Advanced imaging can quantify infarct core and ischemic penumbra using specific threshold values and provides optimal selection parameters, collectively called target mismatch. More precisely, target mismatch criteria consist of core volume and/or penumbra volume and mismatch ratio (the ratio between total hypoperfusion and core volumes) with precise cut-off values. The parameters of target mismatch are automatically calculated with dedicated software packages that allow a quick and standardized interpretation of advanced imaging. However, this approach has several limitations leading to a misclassification of core and penumbra volumes. In fact, automatic software platforms are affected by technical artifacts and are not interchangeable due to a remarkable vendor-dependent variability, resulting in different estimate of target mismatch parameters. In addition, advanced imaging is not completely accurate in detecting infarct core, that can be under- or overestimated. Finally, the selection of candidates for EVT remains currently suboptimal due to the high rates of futile reperfusion and overselection caused by the use of very stringent inclusion criteria. For these reasons, some investigators recently proposed to replace advanced with conventional imaging in the selection for EVT, after the demonstration that non-contrast CT ASPECTS and computed tomography angiography collateral evaluation are not inferior to advanced images in predicting outcome in AIS patients treated with EVT. However, other authors confirmed that CTP and PWI/DWI postprocessed images are superior to conventional imaging in establishing the eligibility of patients for EVT. Therefore, the routine application of automatic assessment of advanced imaging remains a matter of debate. Recent findings suggest that the combination of conventional and advanced imaging might improving our selection criteria.

Radiological features of brain hemorrhage through automated segmentation from computed tomography in stroke and traumatic brain injury

Introduction

Radiological assessment is necessary to diagnose spontaneous intracerebral hemorrhage (ICH) and traumatic brain injury intracranial hemorrhage (TBI-bleed). Artificial intelligence (AI) deep learning tools provide a means for decision support. This study evaluates the hemorrhage segmentations produced from three-dimensional deep learning AI model that was developed using non-contrast computed tomography (CT) imaging data external to the current study.

Methods

Non-contrast CT imaging data from 1263 patients were accessed across seven data sources (referred to as sites) in Norway and Sweden. Patients were included based on ICH, TBI-bleed, or mild TBI diagnosis. Initial non-contrast CT images were available for all participants. Hemorrhage location frequency maps were generated. The number of estimated haematoma clusters was correlated with the total haematoma volume. Ground truth expert annotations were available for one ICH site; hence, a comparison was made with the estimated haematoma volumes. Segmentation volume estimates were used in a receiver operator characteristics (ROC) analysis for all samples (i.e., bleed detected) and then specifically for one site with few TBI-bleed cases.

Results

The hemorrhage frequency maps showed spatial patterns of estimated lesions consistent with ICH or TBI-bleed presentations. There was a positive correlation between the estimated number of clusters and total haematoma volume for each site (correlation range: 0.45-0.74; each p-value < 0.01) and evidence of ICH between-site differences. Relative to hand-drawn annotations for one ICH site, the VIOLA-AI segmentation mask achieved a median Dice Similarity Coefficient of 0.82 (interquartile range: 0.78 and 0.83), resulting in a small overestimate in the haematoma volume by a median of 0.47 mL (interquartile range: 0.04 and 1.75 mL). The bleed detection ROC analysis for the whole sample gave a high area-under-the-curve (AUC) of 0.92 (with sensitivity and specificity of 83.28% and 95.41%); however, when considering only the mild head injury site, the TBI-bleed detection gave an AUC of 0.70.

Discussion

An open-source segmentation tool was used to visualize hemorrhage locations across multiple data sources and revealed quantitative hemorrhage site differences. The automated total hemorrhage volume estimate correlated with a per-participant hemorrhage cluster count. ROC results were moderate-to-high. The VIOLA-AI tool had promising results and might be useful for various types of intracranial hemorrhage.

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.

Artificial Intelligence in Neuroradiology: A Review of Current Topics and Competition Challenges

There is an expanding body of literature that describes the application of deep learning and other machine learning and artificial intelligence methods with potential relevance to neuroradiology practice. In this article, we performed a literature review to identify recent developments on the topics of artificial intelligence in neuroradiology, with particular emphasis on large datasets and large-scale algorithm assessments, such as those used in imaging AI competition challenges. Numerous applications relevant to ischemic stroke, intracranial hemorrhage, brain tumors, demyelinating disease, and neurodegenerative/neurocognitive disorders were discussed. The potential applications of these methods to spinal fractures, scoliosis grading, head and neck oncology, and vascular imaging were also reviewed. The AI applications examined perform a variety of tasks, including localization, segmentation, longitudinal monitoring, diagnostic classification, and prognostication. While research on this topic is ongoing, several applications have been cleared for clinical use and have the potential to augment the accuracy or efficiency of neuroradiologists.

Deep learning-based segmentation of brain parenchyma and ventricular system in CT scans in the presence of anomalies

INTRODUCTION

The automatic segmentation of brain parenchyma and cerebrospinal fluid-filled spaces such as the ventricular system is the first step for quantitative and qualitative analysis of brain CT data. For clinical practice and especially for diagnostics, it is crucial that such a method is robust to anatomical variability and pathological changes such as (hemorrhagic or neoplastic) lesions and chronic defects. This study investigates the increase in overall robustness of a deep learning algorithm that is gained by adding hemorrhage training data to an otherwise normal training cohort.

METHODS

A 2D U-Net is trained on subjects with normal appearing brain anatomy. In a second experiment the training data includes additional subjects with brain hemorrhage on image data of the RSNA Brain CT Hemorrhage Challenge with custom reference segmentations. The resulting networks are evaluated on normal and hemorrhage test casesseparately, and on an independent test set of patients with brain tumors of the publicly available GLIS-RT dataset.

RESULTS

Adding data with hemorrhage to the training set significantly improves the segmentation performance over an algorithm trained exclusively on normally appearing data, not only in the hemorrhage test set but also in the tumor test set. The performance on normally appearing data is stable. Overall, the improved algorithm achieves median Dice scores of 0.98 (parenchyma), 0.91 (left ventricle), 0.90 (right ventricle), 0.81 (third ventricle), and 0.80 (fourth ventricle) on the hemorrhage test set. On the tumor test set, the median Dice scores are 0.96 (parenchyma), 0.90 (left ventricle), 0.90 (right ventricle), 0.75 (third ventricle), and 0.73 (fourth ventricle).

CONCLUSION

Training on an extended data set that includes pathologies is crucial and significantly increases the overall robustness of a segmentation algorithm for brain parenchyma and ventricular system in CT data, also for anomalies completely unseen during training. Extension of the training set to include other diseases may further improve the generalizability of the algorithm.

Accuracy of artificial intelligence software for CT angiography in stroke

OBJECTIVE

Software developed using artificial intelligence may automatically identify arterial occlusion and provide collateral vessel scoring on CT angiography (CTA) performed acutely for ischemic stroke. We aimed to assess the diagnostic accuracy of e-CTA by Brainomix™ Ltd by large-scale independent testing using expert reading as the reference standard.

METHODS

We identified a large clinically representative sample of baseline CTA from 6 studies that recruited patients with acute stroke symptoms involving any arterial territory. We compared e-CTA results with masked expert interpretation of the same scans for the presence and location of laterality-matched arterial occlusion and/or abnormal collateral score combined into a single measure of arterial abnormality. We tested the diagnostic accuracy of e-CTA for identifying any arterial abnormality (and in a sensitivity analysis compliant with the manufacturer's guidance that software only be used to assess the anterior circulation).

RESULTS

We include CTA from 668 patients (50% female; median: age 71 years, NIHSS 9, 2.3 h from stroke onset). Experts identified arterial occlusion in 365 patients (55%); most (343, 94%) involved the anterior circulation. Software successfully processed 545/668 (82%) CTAs. The sensitivity, specificity and diagnostic accuracy of e-CTA for detecting arterial abnormality were each 72% (95% CI = 66-77%). Diagnostic accuracy was non-significantly improved in a sensitivity analysis excluding occlusions from outside the anterior circulation (76%, 95% CI = 72-80%).

INTERPRETATION

Compared to experts, the diagnostic accuracy of e-CTA for identifying acute arterial abnormality was 72-76%. Users of e-CTA should be competent in CTA interpretation to ensure all potential thrombectomy candidates are identified.

Predicting DWI-FLAIR mismatch on NCCT: the role of artificial intelligence in hyperacute decision making

Background

The presence of diffusion-weighted imaging (DWI) and fluid-attenuated inversion recovery (FLAIR) mismatch was used to determine eligibility for intravenous thrombolysis in clinical trials. However, due to the restricted availability of MRI and the ambiguity of image assessment, it is not widely implemented in clinical practice.

Methods

A total of 222 acute ischemic stroke patients underwent non-contrast computed tomography (NCCT), DWI, and FLAIR within 1 h of one another. Human experts manually segmented ischemic lesions on DWI and FLAIR images and independently graded the presence of DWI-FLAIR mismatch. Deep learning (DL) models based on the nnU-net architecture were developed to predict ischemic lesions visible on DWI and FLAIR images using NCCT images. Inexperienced neurologists evaluated the DWI-FLAIR mismatch on NCCT images without and with the model's results.

Results

The mean age of included subjects was 71.8 ± 12.8  years, 123 (55%) were male, and the baseline NIHSS score was a median of 11 [IQR, 6-18]. All images were taken in the following order: NCCT - DWI - FLAIR, starting after a median of 139 [81-326] min after the time of the last known well. Intravenous thrombolysis was administered in 120 patients (54%) after NCCT. The DL model's prediction on NCCT images revealed a Dice coefficient and volume correlation of 39.1% and 0.76 for DWI lesions and 18.9% and 0.61 for FLAIR lesions. In the subgroup with 15  mL or greater lesion volume, the evaluation of DWI-FLAIR mismatch from NCCT by inexperienced neurologists improved in accuracy (from 0.537 to 0.610) and AUC-ROC (from 0.493 to 0.613).

Conclusion

The DWI-FLAIR mismatch may be reckoned using NCCT images through advanced artificial intelligence techniques.

What's the Future of Vascular Neurology?

The field of vascular neurology has made tremendous advances over the last several decades, with major shifts in diagnosis, treatment, prevention, and rehabilitation of patients with stroke. Furthermore, the individuals who are providing the care represent a different cohort than those who were caring for stroke patients 30 years ago, with the increasing need for rapid decision-making for acute interventions and a larger workforce being needed to provide the many complicated aspects of care of stroke patients. Understanding the history of the field is critical before one can speculate about its future directions. In summarizing some of the past massive shifts in the past few decades, this review will discuss future opportunities and future challenges and will introduce the rest of this special issue focusing on vascular neurology in a post-thrombectomy era. Although thrombolysis and thrombectomy remain a major part of ischemic stroke management and care, in the coming years, there will likely be further modifications in how we provide the care, who provides it, how we train those individuals who provide it, where it is provided, and what data inform early management decisions.

Diagnostic Performance of Artificial Intelligence-Based Computer-Aided Detection and Diagnosis in Pediatric Radiology: A Systematic Review

Artificial intelligence (AI)-based computer-aided detection and diagnosis (CAD) is an important research area in radiology. However, only two narrative reviews about general uses of AI in pediatric radiology and AI-based CAD in pediatric chest imaging have been published yet. The purpose of this systematic review is to investigate the AI-based CAD applications in pediatric radiology, their diagnostic performances and methods for their performance evaluation. A literature search with the use of electronic databases was conducted on 11 January 2023. Twenty-three articles that met the selection criteria were included. This review shows that the AI-based CAD could be applied in pediatric brain, respiratory, musculoskeletal, urologic and cardiac imaging, and especially for pneumonia detection. Most of the studies (93.3%, 14/15; 77.8%, 14/18; 73.3%, 11/15; 80.0%, 8/10; 66.6%, 2/3; 84.2%, 16/19; 80.0%, 8/10) reported model performances of at least 0.83 (area under receiver operating characteristic curve), 0.84 (sensitivity), 0.80 (specificity), 0.89 (positive predictive value), 0.63 (negative predictive value), 0.87 (accuracy), and 0.82 (F1 score), respectively. However, a range of methodological weaknesses (especially a lack of model external validation) are found in the included studies. In the future, more AI-based CAD studies in pediatric radiology with robust methodology should be conducted for convincing clinical centers to adopt CAD and realizing its benefits in a wider context.

AI Improves Nodule Detection on Chest Radiographs in a Health Screening Population: A Randomized Controlled Trial

Background The impact of artificial intelligence (AI)-based computer-aided detection (CAD) software has not been prospectively explored in real-world populations. Purpose To investigate whether commercial AI-based CAD software could improve the detection rate of actionable lung nodules on chest radiographs in participants undergoing health checkups. Materials and Methods In this single-center, pragmatic, open-label randomized controlled trial, participants who underwent chest radiography between July 2020 and December 2021 in a health screening center were enrolled and randomized into intervention (AI group) and control (non-AI group) arms. One of three designated radiologists with 13-36 years of experience interpreted each radiograph, referring to the AI-based CAD results for the AI group. The primary outcome was the detection rate, that is, the number of true-positive radiographs divided by the total number of radiographs, of actionable lung nodules confirmed on CT scans obtained within 3 months. Actionable nodules were defined as solid nodules larger than 8 mm or subsolid nodules with a solid portion larger than 6 mm (Lung Imaging Reporting and Data System, or Lung-RADS, category 4). Secondary outcomes included the positive-report rate, sensitivity, false-referral rate, and malignant lung nodule detection rate. Clinical outcomes were compared between the two groups using univariable logistic regression analyses. Results A total of 10 476 participants (median age, 59 years [IQR, 50-66 years]; 5121 men) were randomized to an AI group (n = 5238) or non-AI group (n = 5238). The trial met the predefined primary outcome, demonstrating an improved detection rate of actionable nodules in the AI group compared with the non-AI group (0.59% [31 of 5238 participants] vs 0.25% [13 of 5238 participants], respectively; odds ratio, 2.4; 95% CI: 1.3, 4.7; P = .008). The detection rate for malignant lung nodules was higher in the AI group compared with the non-AI group (0.15% [eight of 5238 participants] vs 0.0% [0 of 5238 participants], respectively; P = .008). The AI and non-AI groups showed similar false-referral rates (45.9% [56 of 122 participants] vs 56.0% [56 of 100 participants], respectively; P = .14) and positive-report rates (2.3% [122 of 5238 participants] vs 1.9% [100 of 5238 participants]; P = .14). Conclusion In health checkup participants, artificial intelligence-based software improved the detection of actionable lung nodules on chest radiographs. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Auffermann in this isssue.