Voltage-based automated detection of postictal generalized electroencephalographic suppression: Algorithm development and validation

ObjectivePostictal generalized electroencephalographic suppression (PGES) is a pattern of low-voltage scalp electroencephalographic (EEG) activity following termination of generalized seizures. PGES has been associated with both sudden unexplained death in patients with epilepsy and therapeutic efficacy of electroconvulsive therapy (ECT). Automated detection of PGES epochs may aid in reliable quantification of this phenomenon.MethodsWe developed a voltage-based algorithm for detecting PGES. This algorithm applies existing criteria to simulate expert epileptologist readings. Validation relied on postictal EEG recording from patients undergoing ECT (NCT02761330), assessing concordance among the algorithm and four clinical epileptologists.ResultsWe observed low-to-moderate concordance among epileptologist ratings of PGES. Despite this, the algorithm displayed high discriminability in comparison to individual epileptologists (C-statistic range: 0.86–0.92). The algorithm displayed high discrimination (C-statistic: 0.91) and substantial peak agreement (Cohen’s Kappa: 0.65) in comparison to a consensus of clinical ratings. Interrater agreement between the algorithm and individual epileptologists was on par with that among expert epileptologists.ConclusionsAn automated voltage-based algorithm can be used to detect PGES following ECT, with discriminability nearing that of experts.SignificanceAlgorithmic detection may support clinical readings of PGES and improve precision when correlating this marker with clinical outcomes following generalized seizures.     

Concordancia interevaluador en la elección de tratamiento profiláctico para la migraña vestibular y papel de un algoritmo asistido para la elección de dicho tratamiento

Background and objectiveVestibular migraine (VM) is a cause of episodic vestibular syndrome. There are many drugs available for its prophylaxis and currently the choice is mainly made according to the patient's comorbidities. The aim of this article was to measure the agreement of a group of otorhinolaryngologists in the choice of a prophylactic treatment and to evaluate the role of an assisted algorithm in the choice of this prophylaxis.Material and methodThe medical records of 10 patients with VM were offered to 10 otolaryngologists who were asked to select for each patient the drug they considered most appropriate among five possible options. The Fleiss’ κ index was calculated among the 10 otolaryngologists alone, recalculating it including the algorithm as the eleventh evaluator, and Cohen's κ index was calculated between each otolaryngologist and the answers of the algorithm. The otolaryngologists were offered the option to change their responses after knowing the responses of the algorithm and then both indexes were calculated again.ResultsThe Fleiss’ κ index was .302. This index was raised to .343 after introducing the algorithm as an evaluator. After offering the responses proposed by the algorithm, Cohen's κ was improved in 9 of the 10 evaluators, and Fleiss’ κ rose to .711.ConclusionsThe agreement between otorhinolaryngologists in choosing prophylaxis for MV can be defined as “fair”. The responses of the algorithm for the choice of prophylaxis were close to the average opinion of the otolaryngologists, raising the agreement between them to “substantial”.     

Forest classification trees and forest support vector machines algorithms: Demonstration using microarray data

Classification into multiple classes when the measured variables are outnumbered is a major methodological challenge in -omics studies. Two algorithms that overcome the dimensionality problem are presented: the forest classification tree (FCT) and the forest support vector machines (FSVM). In FCT, a set of variables is randomly chosen and a classification tree (CT) is grown using a forward classification algorithm. The process is repeated and a forest of CTs is derived. Finally, the most frequent variables from the trees with the smallest apparent misclassification rate (AMR) are used to construct a productive tree. In FSVM, the CTs are replaced by SVMs. The methods are demonstrated using prostate gene expression data for classifying tissue samples into four tumor types. For threshold split value 0.001 and utilizing 100 markers the productive CT consisted of 29 terminal nodes and achieved perfect classification (AMR=0). When the threshold value was set to 0.01, a tree with 17 terminal nodes was constructed based on 15 markers (AMR=7%). In FSVM, reducing the fraction of the forest that was used to construct the best classifier from the top 80% to the top 20% reduced the misclassification to 25% (when using 200 markers). The proposed methodologies may be used for identifying important variables in high dimensional data. Furthermore, the FCT allows exploring the data structure and provides a decision rule.     

Effect of air leak on the performance of auto-PAP devices: a bench study

Automatic positive airway pressure (auto-PAP) devices, used in the treatment of patients with obstructive sleep apnea (OSA), may not function optimally in the presence of an air leak. We set out to determine the magnitude of air leak that prevents auto-PAP devices from responding to respiratory events of OSA in a bench model. We simulated apnea, flow limitation, obstructive hypopnea, nonobstructive hypopnea, and snoring events of OSA with an artificial airway and a loudspeaker in a bench model connected to a commercially available auto-PAP device. Four auto-PAP devices were tested, but two of the “older-generation” devices (Tranquility and Virtuoso) did not respond to events of OSA that involved changes in flow contour; hence, we studied the effects of air leak and humidifier in the two “newer-generation” auto-PAP devices only (GoodKnight 418P, Autoset-T). When the air leak was progressively increased from baseline levels recommended by the manufacturer to levels seen clinically—5 to 7, 10, and 30 l/min—the GoodKnight 418P decreased pressure response by 56% (5.6±1.8 cm H₂O, p=0.04). The pressure response of the Autoset-T, however, did not change from baseline during similar levels of air leak. The GoodKnight performed appropriately when the air leak was within 20 l/min, but the corresponding value for the Autoset was higher at 50 l/min. For both devices combined, air leak caused the pressure to drop between the device and the airway: 2.8±0.3 cm H₂O at 30 l/min of air leak (p<0.001). Air leaks cause auto-PAP devices to underestimate the pressure required to treat events of OSA and to overestimate the pressure delivered at the upper airway. Physicians should be aware of performance limitations of auto-PAP devices in the presence of an air leak.     

Algorithms for challenging scenarios encountered in transradial intervention

Transradial intervention (TRI) was first introduced by Lucien Campeau in 1989 and since then has created a lasting impact in the field of interventional cardiology. Several studies have demonstrated that TRI is associated with fewer vascular site complications, offer earlier ambulation and greater post-procedural comfort. Patients presenting with ST Segment Elevation Myocardial Infarction (STEMI) have experienced survival benefit and higher quality-of-life metrics as well with TRI. While both the updated scientific statement by the American Heart Association and the 2017 European Society of Cardiology guidelines recommend a “radial first” approach there appears to be a lag in physicians adapting TRI as the preferred vascular access. We present a review focusing on identification and management of TRA related challenges and complications using a systematic algorithmic approach.     

Use of the University of California Los Angeles Integrated Staging System (UISS) to predict survival in localized renal cell carcinoma in an Asian population

Objectives To evaluate the applicability of the University of California Los Angeles Integrated Staging System (UISS) in predicting the prognosis of Chinese patients with localized renal cell carcinoma after radical nephrectomy, with reference to that reported by Patard et al in an international multicenter study (J Clin Oncol 2004, 22:3316–3322). Methodology One hundred and twenty-eight Chinese patients with localized renal cell carcinoma were stratified into low risk (LR), intermediate risk (IR) and high risk (HR) groups according to the UISS, based on the TMN staging and Fuhrman grading of the tumor and the Eastern Cooperative Oncology Group performance status of the patients. The survival curves of each risk group were then calculated. Results The number of patients in the LR, IR and HR was 24 (18.8%), 94 (73.4%) and 10 (7.8%) respectively. The estimated 2-year survival rates were 100%, 89.9% and 100% for the LR, IR and HR groups respectively. Whereas the estimated 5-year survival rates were 93.3%, 72.4% and 80% for the LR, IR and HR groups respectively. The LR and IR patients had comparable 2-year and 5-year estimated survival rates with those reported by Patard et al. However, the estimated survival rate for HR patients was better than that reported. Conclusions UISS provided a valuable tool in predicting the survival of Chinese patients with localized renal cell carcinoma of LR and IR groups, as reported in other international centers. Further large scale study may be needed to confirm the applicability in HR population.