Accurate identification of EEG recordings with interictal epileptiform discharges using a hybrid approach: Artificial intelligence supervised by human experts.

OBJECTIVE: To evaluate the diagnostic performance of artificial intelligence (AI)-based algorithms for identifying the presence of interictal epileptiform discharges (IEDs) in routine (20-min) electroencephalography (EEG) recordings. METHODS: We evaluated two approaches: a fully automated one and a hybrid approach, where three human raters applied an operational IED definition to assess the automated detections grouped into clusters by the algorithms. We used three previously developed AI algorithms: Encevis, SpikeNet, and Persyst. The diagnostic gold standard (epilepsy or not) was derived from video-EEG recordings of patients' habitual clinical episodes. We compared the algorithms with the gold standard at the recording level (epileptic or not). The independent validation data set (not used for training) consisted of 20-min EEG recordings containing sharp transients (epileptiform or not) from 60 patients: 30 with epilepsy (with a total of 340 IEDs) and 30 with nonepileptic paroxysmal events. We compared sensitivity, specificity, overall accuracy, and the review time-burden of the fully automated and hybrid approaches, with the conventional visual assessment of the whole recordings, based solely on unrestricted expert opinion. RESULTS: For all three AI algorithms, the specificity of the fully automated approach was too low for clinical implementation (16.67%; 63.33%; 3.33%), despite the high sensitivity (96.67%; 66.67%; 100.00%). Using the hybrid approach significantly increased the specificity (93.33%; 96.67%; 96.67%) with good sensitivity (93.33%; 56.67%; 76.67%). The overall accuracy of the hybrid methods (93.33%; 76.67%; 86.67%) was similar to the conventional visual assessment of the whole recordings (83.33%; 95% confidence interval [CI]: 71.48-91.70%; p > .5), yet the time-burden of review was significantly lower (p < .001). SIGNIFICANCE: The hybrid approach, where human raters apply the operational IED criteria to automated detections of AI-based algorithms, has high specificity, good sensitivity, and overall accuracy similar to conventional EEG reading, with a significantly lower time-burden. The hybrid approach is accurate and suitable for clinical implementation.

Involvement of distal sensory nerves in amyotrophic lateral sclerosis.

INTRODUCTION: The diagnostic criteria for amyotrophic lateral sclerosis (ALS) require normal sensory nerve conduction studies (NCS) or abnormal NCS only in the presence of neuropathy of identified etiology. In this study, we investigated the presence and extent of involvement of Aß sensory fibers in ALS. METHODS: Distal sensory NCS [antidromic dorsal sural (DS) and orthodromic medial plantar (MP)] and conventional sensory NCS (unilateral median sensory and bilateral sural nerves) were performed in 16 definite and 2 probable ALS patients (based on Awaji criteria) and 31 controls. RESULTS: Abnormal conventional sensory NCS were found in 8 (44.4%) ALS patients and 1 (3.2%) control subject (P = 0.002), whereas abnormal distal sensory NCS were found in 12 (66.7%) ALS patients and 3 (9.6%) controls (P < 0.0001). CONCLUSION: Distal sensory NCS were more often abnormal than conventional sensory NCS in ALS. Muscle Nerve 54: 1086-1092, 2016.

Contact Heat Evoked Potentials (CHEPs) in Patients with Mild-Moderate Alzheimer's Disease and Matched Control--A Pilot Study.

OBJECTIVE: Clinical studies have found that patients with Alzheimer's disease report pain of less intensity and with a lower affective response, which has been thought to be due to altered pain processing. The authors wished to examine the cerebral processing of non-painful and painful stimuli using somatosensory evoked potentials and contact heat evoked potentials in patients with Alzheimer's disease and in healthy elderly controls. DESIGN: Case-control study SETTING AND SUBJECTS: Twenty outpatients with mild-moderate Alzheimer's disease and in 17 age- and gender-matched healthy controls were included METHOD: Contact heat evoked potentials and somatosensory evoked potentials were recorded in all subjects. Furthermore, warmth detection threshold and heat pain threshold were assessed. Patients and controls also rated quality and intensity of the stimuli. RESULTS: The authors found no difference on contact heat evoked potential amplitude (P = 0.59) or latency of N2 or P2 wave (P = 0.62 and P = 0.75, respectively) between patients and controls. In addition, there was no difference in regard to pain intensity scores or pain quality. The patients and controls had similar warmth detection threshold and heat pain threshold. Somatosensory evoked potentials, amplitude, and latency were within normal range and similar for the two groups. CONCLUSIONS: The findings suggest that the processing of non-painful and painful stimuli is preserved in patients with mild to moderate Alzheimer's disease.

High-resolution ultrasound in ulnar neuropathy at the elbow: A prospective study.

INTRODUCTION: We examined the role of ultrasound in diagnosing and localizing ulnar neuropathy (UN) at the elbow. METHODS: Forty-one patients referred for UN and 43 healthy subjects were included. Cross-sectional area (CSA) of the ulnar nerve was measured by ultrasound at 5 locations across the elbow and at the wrist. Patients underwent near-nerve conduction studies (N-NCS). RESULTS: Ultrasound (US) had a sensitivity of 80.5%, N-NCS had a sensitivity of 70.7%, and combined sensitivity was 85.4% in diagnosing UN. Using N-NCS as a reference standard, US reached a sensitivity of 93.1% and a specificity of 50.0% in diagnosing UN, and a sensitivity of 77.3% and a specificity of 42.9% in localizing UN at the elbow. There was a reduction in mean CSA at 3-month follow-up. CONCLUSIONS: N-NCS and US are complementary methods and have the potential to increase the sensitivity for diagnosing and localizing UN at the elbow.

EEG Frequency Correlates with α2-Receptor Density in Parkinson's Disease.

INTRODUCTION: Increased theta and delta power and decreased alpha and beta power, measured with quantitative electroencephalography (EEG), have been demonstrated to have utility for predicting the development of dementia in patients with Parkinson's disease (PD). Noradrenaline modulates cortical activity and optimizes cognitive processes. We claim that the loss of noradrenaline may explain cognitive impairment and the pathological slowing of EEG waves. Here, we test the relationship between the number of noradrenergic α2 adrenoceptors and changes in the spectral EEG ratio in patients with PD. METHODS: We included nineteen patients with PD and thirteen healthy control (HC) subjects in the study. We used positron emission tomography (PET) with [11C]yohimbine to quantify α2 adrenoceptor density. We used EEG power in the delta (δ, 1.5-3.9 Hz), theta (θ, 4-7.9 Hz), alpha (α, 8-12.9 Hz) and beta (ß, 13-30 Hz) bands in regression analyses to test the relationships between α2 adrenoceptor density and EEG band power. RESULTS: PD patients had higher power in the theta and delta bands compared to the HC volunteers. Patients' theta band power was inversely correlated with α2 adrenoceptor density in the frontal cortex. In the HC subjects, age was correlated with, and occipital background rhythm frequency (BRF) was inversely correlated with, α2 adrenoceptor density in the frontal cortex, while occipital BRF was inversely correlated with α2 adrenoceptor density in the thalamus. CONCLUSIONS: The findings support the claim that the loss or dysfunction of noradrenergic neurotransmission may relate to the parallel processes of cognitive decline and EEG slowing.

Lifestyle risk factors for ulnar neuropathy and ulnar neuropathy-like symptoms.

INTRODUCTION: We examined whether lifestyle factors differ between patients with ulnar neuropathy confirmed by electroneurography (ENG) and those with ulnar neuropathy-like symptoms with normal ulnar nerve ENG. METHODS: Among patients examined by ENG for suspected ulnar neuropathy, we identified 546 patients with ulnar neuropathy and 633 patients with ulnar neuropathy-like symptoms. These groups were compared with 2 separate groups of matched community referents and to each other. Questionnaire information on lifestyle was obtained. The electrophysiological severity of neuropathy was also graded. We used conditional and unconditional logistic regression. RESULTS: Responses were obtained from 59%. Ulnar neuropathy was related to smoking, adjusted odds ratio (OR) 4.31 (95% confidence interval [CI] 2.43-7.64) for >24 pack-years. Ulnar neuropathy-like symptoms were related to body mass index ≥30 kg/m(2), OR 1.99 (95% CI 1.25-3.19). Smoking was associated with increased severity of ulnar neuropathy. CONCLUSIONS: Findings suggest that smoking specifically affects the ulnar nerve.

Differential effects of a 5% lidocaine medicated patch in peripheral nerve injury.

INTRODUCTION: We examined the effect of topical lidocaine on the function of small and large fibers in patients with peripheral neuropathic pain due to traumatic or postoperative nerve injury. METHODS: In an open-label study, 24 patients were treated with a 5% lidocaine patch for up to 12 weeks. We recorded contact heat evoked potentials (CHEPs) and performed quantitative sensory testing (QST) before and after treatment with the contralateral side as control. RESULTS: Twenty-one patients (mean age 47.6 ± 13.5 years) completed the study. Lidocaine increased cold pain threshold (P = 0.04) and reduced CHEP amplitude (P = 0.007) with no effect on other QST parameters. Patients responding to treatment had less cold detection deficit on the affected side and had a larger increase in cold pain detection threshold following treatment than nonresponders. CONCLUSIONS: Controlled trials are warranted to further understand the mechanisms mediating the effects of topical lidocaine.

Cortical somatosensory evoked potential amplitudes and clinical outcome after cardiac arrest: a retrospective multicenter study.

OBJECTIVE: Bilaterally absent cortical somatosensory evoked potentials (SSEPs) reliably predict poor outcome in comatose cardiac arrest (CA) patients. Cortical SSEP amplitudes are a recent prognostic extension; however, amplitude thresholds, inter-recording, and inter-rater agreement remain uncertain. METHODS: In a retrospective multicenter cohort study, we determined cortical SSEP amplitudes of comatose CA patients using a standardized evaluation pathway. We studied inter-recording agreement in repeated SSEPs and inter-rater agreement by four raters independently determining 100 cortical SSEP amplitudes. Primary outcome was assessed using the cerebral performance category (CPC) upon intensive care unit discharge dichotomized into good (CPC 1-3) and poor outcome (CPC 4-5). RESULTS: Of 706 patients with SSEPs with median 3 days after CA, 277 (39.2%) had good and 429 (60.8%) poor outcome. Of patients with bilaterally absent cortical SSEPs, one (0.8%) survived with CPC 3 and 130 (99.2%) had poor outcome. Otherwise, the lowest cortical SSEP amplitude in good outcome patients was 0.5 µV. 184 (42.9%) of 429 poor outcome patients had lower cortical SSEP amplitudes. In 106 repeated SSEPs, there were 6 (5.7%) with prognostication-relevant changes in SSEP categories. Following a standardized evaluation pathway, inter-rater agreement was almost perfect with a Fleiss' kappa of 0.88. INTERPRETATION: Bilaterally absent and cortical SSEP amplitudes below 0.5 µV predicted poor outcome with high specificity. A standardized evaluation pathway provided high inter-rater and inter-recording agreement. Regain of consciousness in patients with bilaterally absent cortical SSEPs rarely occurs. High-amplitude cortical SSEP amplitudes likely indicate the absence of severe brain injury.

Ulnar neuropathy and ulnar neuropathy-like symptoms in relation to biomechanical exposures assessed by a job exposure matrix: a triple case-referent study.

OBJECTIVES: We aimed to evaluate relations between occupational biomechanical exposures and (1) ulnar neuropathy confirmed by electroneurography (ENG) and (2) ulnar neuropathy-like symptoms with normal ENG. METHODS: In this triple case-referent study, we identified all patients aged 18-65 years, examined with ENG at a neurophysiological department on suspicion of ulnar neuropathy, 2001-2007. We mailed a questionnaire to 546 patients with ulnar neuropathy, 633 patients with ulnar neuropathy-like symptoms and two separate groups of community referents, matched on sex, age and primary care centre (risk set sampling). The two patient groups were also compared to each other directly. We constructed a Job Exposure Matrix to provide estimates of exposure to non-neutral postures, repetitive movements, hand-arm vibrations and forceful work. Conditional and unconditional logistic regressions were used. RESULTS: The proportion who responded was 59%. Ulnar neuropathy was related to forceful work with an exposure-response pattern reaching an OR of 3.85 (95% CI 2.04 to 7.24); non-neutral postures strengthened effects of forceful work. No relation was observed with repetitive movements. Ulnar neuropathy-like symptoms were related to repetitive movements with an OR of 1.89 (95% CI 1.01 to 3.52) in the highest-exposure category (≥2.5 h/day); forceful work was unrelated to the outcome. CONCLUSIONS: Ulnar neuropathy and ulnar neuropathy-like symptoms differed with respect to associations with occupational biomechanical exposures. Findings suggested specific effects of forceful work on the ulnar nerve. Thus, results corroborated the importance of an electrophysiological diagnosis when evaluating risk factors for ulnar neuropathy. Preventive effects may be achieved by reducing biomechanical exposures at work.

Common patterns of EEG reactivity in post-anoxic coma identified by quantitative analyses.

OBJECTIVE: Description of typical kinds of EEG reactivity (EEG-R) in post-anoxic coma using a quantitative method. METHODS: Study of 101 out-of-hospital cardiac arrest patients, 71 with good outcome (cerebral performance category scale ≤ 2). EEG was recorded 12-24 hours after cardiac arrest and four noxious, one auditory, and one visual stimulation were applied for 30 seconds each. Individual reference intervals for the power in the delta, theta, alpha, and beta bands were calculated based on six 2-second resting epochs just prior to stimulations. EEG-R in consecutive 2-second epochs after stimulation was expressed in Z-scores. RESULTS: EEG-R occurred roughly equally frequent as an increase or as a decrease in EEG activity. Sternal rub and sound stimulation were most provocative with the most pronounced changes as an increase in delta activity 4.5-8.5 seconds after stimulation and a decrease in theta activity 0.5-4.5 seconds after stimulation. These parameters predicted good outcome with an AUC of 0.852 (95 % CI: 0.771-0.932). CONCLUSIONS: Quantitative EEG-R is a feasible method for identification of common types of reactivity, for evaluation of stimulation methods, and for prognostication. SIGNIFICANCE: This method provides an objective measure of EEG-R revealing knowledge about the nature of EEG-R and its use as a diagnostic tool.

[EEG diagnosticsof non-convulsive status epilepticus in critically ill patients].

Critically ill patients are at high risk of non-convulsive status epilepticus (NCSE). As clinical signs of NCSE are subtle and unspecific, EEG is necessary to make the diagnosis. This is a review of the terminology for EEG reporting and the criteria for NCSE in critically ill patients. We discuss the newly proposed ictal-interictal continuum, and how caution is needed when assessing EEG criteria in order to avoid both over- and undertreatment. Finally, we discuss how specific EEG findings, in combination with clinical information, can help infer treatment decision and need for continuous EEG monitoring.

Comparison of ultrasound with electrodiagnosis of scapular winging: A prospective case control study.

OBJECTIVE: Compare high-resolution ultrasound (HRUS) and electrodiagnostic examination (EDX) in the diagnostic workup of patients with scapulae alatae. METHODS: 27 patients with scapulae alatae and 41 healthy subjects (HS) and underwent a standardized clinical examination (CEX), EDX and HRUS. We measured the thickness of the serratus anterior (SER), rhomboid major and trapezius muscles and the diameter of the long thoracic (LTN), dorsal scapular and spinal accessory nerves (SAN). RESULTS: Twenty patients showed medial winging and six patients showed lateral winging on CEX. One patient had both lateral and medial winging. In patients with medial winging, the SER muscle was thinner and the LTN diameter was larger on the symptomatic side compared with the asymptomatic side and with the dominant side in HS. In this group, both EDX and HRUS detected abnormalities of SER muscle/ LTN with sensitivity of 65%, and with specificity of 100% and 57%, respectively. EDX and HRUS detected abnormalities of the trapezius muscle/ SAN with sensitivity of 60% and 40%, and specificity of 91%, and 86 % a, respectively. There was no significant difference between the two methods. CONCLUSION: HRUS can contribute to the diagnostic workup of scapulae alatae by demonstrating atrophy of muscles and enlargement in nerve diameter. SIGNIFICANCE: HRUS supplements EDX in the diagnostic workup of scapulae alatae.

The influence of the abundance and morphology of epileptiform discharges on diagnostic accuracy: How many spikes you need to spot in an EEG.

OBJECTIVE: The operational definition of interictal epileptiform discharges (IEDs) of the International Federation of Clinical Neurophysiology (IFCN) described six morphological criteria. Our objective was to assess the impact of pattern-repetition in the EEG-recording, on the diagnostic accuracy of using the IFCN criteria. For clinical implementation, specificity over 95% was set as target. METHODS: Interictal EEG-recordings of 20-minutes, containing sharp-transients, from 60 patients (30 with epilepsy and 30 with non-epileptic paroxysmal events) were evaluated by three experts, who first marked IEDs solely based on expert opinion, and then, independently from the first session evaluated the presence of the IFCN criteria for each sharp-transient. The gold standard was derived from long-term video-EEG recordings of the patients habitual paroxysmal episodes. RESULTS: Presence of at least one discharge fulfilling five criteria provided a specificity of 100% (sensitivity: 70%). For discharges fulfilling fewer criteria, a higher number of discharges was needed to keep the specificity over 95% (5 discharges, when only 3 criteria were fulfilled). A sequential combination of these sets of criteria and thresholds provided a specificity of 97% and sensitivity of 80%. CONCLUSIONS: Pattern-repetition and IED morphology influence diagnostic accuracy. SIGNIFICANCE: Systematic application of these criteria will improve quality of clinical EEG interpretation.

Neuromuscular ultrasound of the scapular stabilisers in healthy subjects.

OBJECTIVES: To obtain normative high-resolution ultrasound (HRUS) data for thickness of the serratus anterior, the trapezius and the rhomboid major muscles and diameter of their corresponding nerves, the long thoracic, the spinal accessory and the dorsal scapular nerve. Moreover, we aimed to examine intra- and inter-examiner agreement of the HRUS measurements. METHODS: We included 41 healthy subjects. Muscle thickness and nerve diameter were measured bilaterally, resulting in 82 ultrasound measurements for each structure. Normative data were calculated using regression equations for the lower limit of muscle thickness and upper limit of nerve diameter, taking into account various variables. For intra- and inter-examiner agreement, ten subjects underwent two extra ultrasound examinations and Bland-Altman plots were calculated. RESULTS: This normative data set showed significant correlations between decreasing muscle thickness with increasing age and height and increasing muscle thickness with increasing weight and with male sex. Muscle thickness was larger on the dominant side compared to the non-dominant side for the trapezius and rhomboid muscles, whereas the opposite was found for the serratus anterior muscle. For all nerves, significant correlations were found between decreasing nerve diameter with increasing age and height. Intra-examiner agreement was acceptable in all sites. Inter-examiner agreement was acceptable for all sites but one site for the serratus anterior muscle and long thoracic nerve, and not acceptable for five out of six sites for the trapezius muscle. CONCLUSION: This study provides HRUS normative data and intra- and inter-examiner agreement data for muscle thickness and nerve diameter for the muscles stabilizing the scapulae and their corresponding nerves. SIGNIFICANCE: The normative HRUS data reported may be useful in future studies investigating neuromuscular disorders.

Gold Coast diagnostic criteria increase sensitivity in amyotrophic lateral sclerosis.

OBJECTIVE: This study evaluates diagnostic accuracy of the proposed 'Gold Coast' (GC) diagnostic criteria for amyotrophic lateral sclerosis (ALS). METHODS: Five European centres retrospectively sampled consecutive patients referred for electromyography on suspicion of ALS. Patients were classified according to the GC criteria, the revised El Escorial (rEE) criteria and the Awaji (AW) criteria without and with the 'Possible' category (+ Poss). Reference standard was ALS confirmed by disease progression at follow-up. RESULTS: Of 404 eligible patients 272 were diagnosed as ALS, 94 had mimicking disorders, 35 were lost for follow-up, and three had insufficient data. Sensitivity for the GC criteria was 88.2% (95% CI: 83.8-91.8%), which was higher than for previous criteria, of which the AW + Poss criteria reached the highest sensitivity of 77.6% (95% CI: 72.2-82.4%) (p < 0.001). Specificity was high for all criteria. The increase in sensitivity for the GC criteria was mainly due to the inclusion of 28 patients with progressive muscular atrophy (PMA). CONCLUSIONS: The simpler GC criteria increase the sensitivity, primarily due to considering PMA as a form of ALS with high specificity preserved. SIGNIFICANCE: This validation study supports that GC criteria should be used in clinical practice and may be used for inclusion in trials.

Automatic continuous EEG signal analysis for diagnosis of delirium in patients with sepsis.

OBJECTIVE: In critical care, continuous EEG (cEEG) monitoring is useful for delirium diagnosis. Although visual cEEG analysis is most commonly used, automatic cEEG analysis has shown promising results in small samples. Here we aimed to compare visual versus automatic cEEG analysis for delirium diagnosis in septic patients. METHODS: We obtained cEEG recordings from 102 septic patients who were scored for delirium six times daily. A total of 1252 cEEG blocks were visually analyzed, of which 805 blocks were also automatically analyzed. RESULTS: Automatic cEEG analyses revealed that delirium was associated with 1) high mean global field power (p < 0.005), mainly driven by delta activity; 2) low average coherence across all electrode pairs and all frequencies (p < 0.01); 3) lack of intrahemispheric (fronto-temporal and temporo-occipital regions) and interhemispheric coherence (p < 0.05); and 4) lack of cEEG reactivity (p < 0.005). Classification accuracy was assessed by receiver operating characteristic (ROC) curve analysis, revealing a slightly higher area under the curve for visual analysis (0.88) than automatic analysis (0.74) (p < 0.05). CONCLUSIONS: Automatic cEEG analysis is a useful supplement to visual analysis, and provides additional cEEG diagnostic classifiers. SIGNIFICANCE: Automatic cEEG analysis provides useful information in septic patients.

Variation in the neurophysiological examination of amyotrophic lateral sclerosis in Europe.

Our objective was to analyse how patients with amyotrophic lateral sclerosis (ALS) are examined neurophysiologically at different European centres in order to identify possible areas with variation or disagreement in the neurophysiological examination of ALS. Ninety-three prospectively collected examinations from six out of seven neurophysiologists in the European ESTEEM project were analysed. All examinations were peer reviewed with an electromyographic consensus diagnosis of motor neuron disease and the diagnosis of ALS confirmed by clinical follow-up. The examinations were analysed for differences among the physicians in EMG techniques and number and distribution of examined and abnormal muscles and nerve segments. Considerable variation was found among the physicians regarding the average numbers of performed and abnormal EMG and nerve conduction studies per patient, the EMG techniques used, and the topographical distribution of the examined muscles. The existence of two different examination approaches, one with quantitative EMG analyses and relatively few muscles studied, and one with more muscles studied using qualitative methods was clearly confirmed in the present study. The large variation among the physicians indicates that different criteria were used, or that criteria were used inconsistently.

Post resuscitation prognostication by EEG in 24 vs 48 h of targeted temperature management.

OBJECTIVE: To test if prognostic performance is affected by prolonged targeted temperature management (TTM) in comatose out-of-hospital cardiac arrest patients using two recently proposed EEG pattern classification models. METHODS: In this sub-study of the "Target Temperature Management for 48 vs. 24 hand Neurologic Outcome after Out-of-Hospital Cardiac Arrest: A Randomized Clinical Trial", EEGs of 20-30 min duration were collected 24 h and 48 h after reaching the target temperature of 33 ±â€¯1 °C. We classified EEGs according to two EEG classification models by Westhall et al. ("highly malignant", "malignant" and "benign") and Hofmeijer et al. ("unfavorable", "intermediate" and "favorable"). We tested prognostic ability against 6 months functional outcome using the Cerebral Performance Category score. RESULTS: We recorded EEGs in 120 patients at 24 h and in 44 patients at 48 h. We found no difference in specificities or sensitivities of the two models between the two TTM groups (all p-values >0.19) or in prognostication at 24 h compared to 48 h (all p-values >0.13), except for the presence of EEG reactivity favoring prognostication at 24 h (p < 0.001). Being classified in the "benign" or "favorable" category was strongly associated with good outcome with specificities of 100% (90-100) and 97% (85-100) for the Westhall and Hofmeijer models respectively. CONCLUSIONS: We found no difference in the prognostic performance of the two studied EEG classification models during prolonged TTM for 48 h compared to standard duration, nor between EEG classification performed at 24 h versus 48 h after reaching target temperature. The two models performed best in good outcome prediction.