Artificial intelligence for MRI stroke detection: a systematic review and meta-analysis.

OBJECTIVES: This systematic review and meta-analysis aimed to assess the stroke detection performance of artificial intelligence (AI) in magnetic resonance imaging (MRI), and additionally to identify reporting insufficiencies. METHODS: PRISMA guidelines were followed. MEDLINE, Embase, Cochrane Central, and IEEE Xplore were searched for studies utilising MRI and AI for stroke detection. The protocol was prospectively registered with PROSPERO (CRD42021289748). Sensitivity, specificity, accuracy, and area under the receiver operating characteristic (ROC) curve were the primary outcomes. Only studies using MRI in adults were included. The intervention was AI for stroke detection with ischaemic and haemorrhagic stroke in separate categories. Any manual labelling was used as a comparator. A modified QUADAS-2 tool was used for bias assessment. The minimum information about clinical artificial intelligence modelling (MI-CLAIM) checklist was used to assess reporting insufficiencies. Meta-analyses were performed for sensitivity, specificity, and hierarchical summary ROC (HSROC) on low risk of bias studies. RESULTS: Thirty-three studies were eligible for inclusion. Fifteen studies had a low risk of bias. Low-risk studies were better for reporting MI-CLAIM items. Only one study examined a CE-approved AI algorithm. Forest plots revealed detection sensitivity and specificity of 93% and 93% with identical performance in the HSROC analysis and positive and negative likelihood ratios of 12.6 and 0.079. CONCLUSION: Current AI technology can detect ischaemic stroke in MRI. There is a need for further validation of haemorrhagic detection. The clinical usability of AI stroke detection in MRI is yet to be investigated. CRITICAL RELEVANCE STATEMENT: This first meta-analysis concludes that AI, utilising diffusion-weighted MRI sequences, can accurately aid the detection of ischaemic brain lesions and its clinical utility is ready to be uncovered in clinical trials. KEY POINTS: There is a growing interest in AI solutions for detection aid. The performance is unknown for MRI stroke assessment. AI detection sensitivity and specificity were 93% and 93% for ischaemic lesions. There is limited evidence for the detection of patients with haemorrhagic lesions. AI can accurately detect patients with ischaemic stroke in MRI.

Emergency Medical Services dispatcher recognition of stroke: A systematic review.

PURPOSE: Stroke treatments are time-sensitive, and thus early and correct recognition of stroke by Emergency Medical Services is essential for outcomes. This is particularly important with the adaption of mobile stroke units. In this systematic review, we therefore aimed to provide a comprehensive overview of Emergency Medical Services dispatcher recognition of stroke. METHODS: The review was registered on PROSPERO and the PRISMA guidelines were applied. We searched PubMed, Embase, and Cochrane Review Library. Screening and data extraction were performed by two observers. Risk of bias was assessed using the QUADAS-2 instrument. FINDINGS: Of 1200 papers screened, 24 fulfilled the inclusion criteria. Data on sensitivity was reported in 22 papers and varied from 17.9% to 83.0%. Positive predictive values were reported in 12 papers and ranged from 24.0% to 87.7%. Seven papers reported specificity, which ranged from 20.0% to 99.1%. Six papers reported negative predictive value, ranging from 28.0% to 99.4%. In general, the risk of bias was low. DISCUSSION: Stroke recognition by dispatchers varied greatly, but overall many patients with stroke are not recognised, despite the initiatives taken to improve stroke literacy. The available data are of high quality, however Asian, African, and South American populations are underrepresented. CONCLUSION: While the data are heterogenous, this review can serve as a reference for future research in emergency medical dispatcher stroke recognition and initiatives to improve prehospital stroke recognition.

School performance and psychiatric comorbidity in childhood absence epilepsy: A Danish cohort study.

The aim was to determine school performance and psychiatric comorbidity in children with childhood absence epilepsy (CAE). We reviewed the medical records in children with ICD-10 codes for idiopathic generalized epilepsy before 18 years of age, and pediatric neurologists confirmed the International League Against Epilepsy criteria for CAE were met. Control groups were the general pediatric population or children with non-neurological chronic disease. Outcomes were from nationwide and population-based registers on school performance and psychiatric comorbidity. We compared the mean grade point average using linear regression and estimated hazard ratios (HR) using Cox regression for the other outcomes. Analyses were adjusted for the child's sex, and year of birth, and parental highest education, receipt of cash benefits or early disability pension. We included 114 children with CAE with a median age at onset of 5.9 years (interquartile range = 4.5-7.3 years). Compared with both population controls and non-neurological chronically ill children, children with CAE had increased hazard of special needs education (HR = 2.7, 95% confidence interval (CI) = 1.8-4.1, p < 0.0001), lower grade point average at 9th grade by 1.7 grade points (95% CI = -2.5 to -1.0, p < 0.001), increased ADHD medicine use (HR = 4.4, 95% CI = 2.7-7.2, p < 0.001), increased sleep medicine use (HR = 2.7, 95% CI = 1.7-4.3, p < 0.001), and increased psychiatry visits (HR = 2.1, 95% CI = 1.1-4.0, p = 0.03). In conclusion, children with CAE have increased psychiatric comorbidity and a considerable proportion of these children receive special needs education in primary/secondary school, albeit insufficient to normalize their considerably lower grade point average in the 9th grade.

School performance and psychiatric comorbidity in juvenile absence epilepsy and juvenile myoclonic epilepsy: a Danish population-based cohort study.

BACKGROUND: We aimed to determine school performance and psychiatric comorbidity in juvenile absence epilepsy (JAE), juvenile myoclonic epilepsy (JME), and generalized tonic-clonic seizures (GTCS) alone. METHODS: All children (< 18 years) fulfilled International League Against Epilepsy criteria after review of their medical records. Control groups were the pediatric background population or children with non-neurological chronic disease. Outcomes were on school performance and psychiatric comorbidity. We compared mean grade point averages using linear regression and estimated hazard ratios using Cox regression in the remaining analyses. We adjusted for the child's sex, age, and year of birth; and parental highest education, receipt of cash benefits or early retirement. RESULTS: We included 92 JAE, 190 JME, 27 GTCS alone, 15,084 non-neurological chronic disease controls, and population controls. JAE had two times increased hazard for special needs education compared with age-matched population controls (hazard ratio 2.2, 95% CI = 1.1‒4.6, p = 0.03); this was not seen in JME. Compared with population controls, both JAE and JME had lower grade point average in secondary and high school (JME: 9th grade: - 0.5 points, 95% CI = -0.9 to -0.06, p = 0.03; high school: - 0.6 points, 95% CI = -1.3 to -0.1, p = 0.04), and 8% fewer JME and 15% fewer JAE attended high school. Both JME and JAE had higher hazard for redeeming sleep medication compared with non-neurological chronic disease; additionally, JAE had increased hazard for ADHD medicine redemptions. CONCLUSIONS: Both JAE and JME had marginally poorer school performance; performance seemed worse in JAE than in JME. Both JAE and JME had increased use of sleep medication.

Validation of Pediatric Idiopathic Generalized Epilepsy Diagnoses from the Danish National Patient Register During 1994‒2019.

Objective: To identify pediatric idiopathic generalized epilepsy (IGE) during 1994-2019 using ICD-10 codes in the Danish National Patient Register and anti-seizure prescriptions in the Danish Prescription Database. Study Design and Setting: We reviewed the medical records in children with ICD-10 codes for IGE before 18 years of age, and pediatric neurologists confirmed that the International League Against Epilepsy criteria were met. We estimated positive predictive values (PPV) and sensitivity for ICD-10 alone, including combinations of codes, anti-seizure prescription, and age at first code registration using medical record-validated diagnoses as gold standard. Results: We validated the medical record in 969 children with an ICD-10 code of IGE, and 431 children had IGE (115 childhood absence epilepsy, 97 juvenile absence epilepsy, 192 juvenile myoclonic epilepsy, 27 generalized tonic-clonic seizures alone). By combining ICD-10 codes with antiseizure prescription and age at epilepsy code registration, we found a PPV for childhood absence epilepsy at 44% (95% confidence interval [CI]=34%‒54%) and for juvenile absence epilepsy at 44% (95% CI=36%-52%). However, ethosuximide prescription, age at ethosuximide code registration before age 8 years and a combination of ICD-10 codes yielded a PPV of 59% (95% CI=42%‒75%) for childhood absence epilepsy but the sensitivity was only 17% (20/115 children identified). For juvenile myoclonic epilepsy the highest PPV was 68% (95% CI=62%‒74%) using the code G40.3F plus antiseizure prescription and age at epilepsy code registration after age 8 years, with sensitivity of 85% (164/192 children identified). For generalized tonic-clonic seizures alone the highest PPV was 31% (95% CI=15%‒51%) using G40.3G during 2006-2019 plus antiseizure prescription and age at code registration after age 5 years. Conclusion: The Danish National Patient Register and the Danish Prescription Database are not suitable for identifying children with IGE subtypes, except for juvenile myoclonic epilepsy which can be identified with caution.