Construction and validation of an algorithm to separate focal and generalised epilepsy using clinical variables: A comparison of machine learning approaches.

PURPOSE: Epilepsy type, whether focal or generalised, is important in deciding anti-seizure medication (ASM). In resource-limited settings, investigations are usually not available, so a clinical separation is required. We used a naïve Bayes approach to devise an algorithm to do this, and compared its accuracy with algorithms devised by five other machine learning methods. METHODS: We used data on 28 clinical variables from 503 patients attending an epilepsy clinic in India with defined epilepsy type, as determined by an epileptologist with access to clinical, imaging, and EEG data. We adopted a machine learning approach to select the most relevant variables based on mutual information, to train the model on part of the data, and then to evaluate it on the remaining data (testing set). We used a naïve Bayes approach and compared the results in the testing set with those obtained by several other machine learning algorithms by measuring sensitivity, specificity, accuracy, area under the curve, and Cohen's kappa. RESULTS: The six machine learning methods produced broadly similar results. The best naïve Bayes algorithm contained eleven variables, and its accuracy was 92.2% in determining epilepsy type (sensitivity 92.0%, specificity 92.7%). An algorithm incorporating the best eight of these variables was only slightly less accurate - 91.0% (sensitivity 89.6%, and specificity 95.1%) - and easier for clinicians to use. CONCLUSION: A clinical algorithm with eight variables is effective and accurate at separating focal from generalised epilepsy. It should be useful in resource-limited settings, by epilepsy-inexperienced doctors, to help determine epilepsy type and therefore optimal ASMs for individual patients, without the need for EEG or neuroimaging.

An epilepsy type algorithm developed in India is accurate in Sudan: A prospective validation study.

PURPOSE: The effects of epilepsy are worse in lower- and middle-income countries (LMICs) where most people with epilepsy live, and where most are untreated. Correct treatment depends on determining whether focal or generalised epilepsy is present. EEG and MRI are usually not available to help so an entirely clinical method is required. We applied an eight-variable algorithm, which had been derived from 503 patients from India using naïve-Bayesian methods, to an adult Sudanese cohort with epilepsy. METHODS: There were 150 consecutive adult patients with known epilepsy type as defined by two neurologists who had access to clinical information, EEG and neuroimaging ("the gold standard"). We used seven of the eight variables, together with their likelihood ratios, to calculate the probability of focal as opposed to generalised epilepsy in each patient and compared that to the "gold standard". Sensitivity, specificity, accuracy, and Cohen's kappa statistic were calculated. RESULTS: Mean age was 28 years (range 17-49) and 53% were female. The accuracy of an algorithm comprising seven of the eight variables was 92%, with sensitivity of 99% and specificity of 72% for focal epilepsy. Cohen's kappa was 0.773, indicating substantial agreement. Ninety-four percent of patients had probability scores either less than 0.1 (generalised) or greater than 0.9 (focal). CONCLUSION: The results confirm the high accuracy of this algorithm in determining epilepsy type in Sudan. They suggest that, in a clinical condition like epilepsy, where a history is crucial, results in one continent can be applied to another. This is especially important as untreated epilepsy and the epilepsy treatment gap are so widespread. The algorithm can be applied to patients giving an individual probability score which can help determine the appropriate anti-seizure medication. It should give epilepsy-inexperienced doctors confidence in managing patients with epilepsy.

How good is an explanation?

How good is an explanation and when is one explanation better than another? In this paper, I address these questions by exploring probabilistic measures of explanatory power in order to defend a particular Bayesian account of explanatory goodness. Critical to this discussion is a distinction between weak and strong measures of explanatory power due to Good (Br J Philos Sci 19:123-143, 1968). In particular, I argue that if one is interested in the overall goodness of an explanation, an appropriate balance needs to be struck between the weak explanatory power and the complexity of a hypothesis. In light of this, I provide a new defence of a strong measure proposed by Good by providing new derivations of it, comparing it with other measures and exploring its connection with information, confirmation and explanatory virtues. Furthermore, Good really presented a family of strong measures, whereas I draw on a complexity criterion that favours a specific measure and hence provides a more precise way to quantify explanatory goodness.

Optimal COVID-19 lockdown strategies in an age-structured SEIR model of Northern Ireland.

An age-structured SEIR model simulates the propagation of COVID-19 in the population of Northern Ireland. It is used to identify optimal timings of short-term lockdowns that enable long-term pandemic exit strategies by clearing the threshold for herd immunity or achieving time for vaccine development with minimal excess deaths.

An Early Stage Researcher's Primer on Systems Medicine Terminology.

Background: Systems Medicine is a novel approach to medicine, that is, an interdisciplinary field that considers the human body as a system, composed of multiple parts and of complex relationships at multiple levels, and further integrated into an environment. Exploring Systems Medicine implies understanding and combining concepts coming from diametral different fields, including medicine, biology, statistics, modeling and simulation, and data science. Such heterogeneity leads to semantic issues, which may slow down implementation and fruitful interaction between these highly diverse fields. Methods: In this review, we collect and explain more than100 terms related to Systems Medicine. These include both modeling and data science terms and basic systems medicine terms, along with some synthetic definitions, examples of applications, and lists of relevant references. Results: This glossary aims at being a first aid kit for the Systems Medicine researcher facing an unfamiliar term, where he/she can get a first understanding of them, and, more importantly, examples and references for digging into the topic.

European and US lockdowns and second waves during the COVID-19 pandemic.

This paper investigates the lockdowns to contain the spread of the SARS-CoV-2 coronavirus in France, Germany, Italy, Spain, the UK and the US and also recent developments since these lockdowns have been relaxed. The analysis employs a two-stage SEIR model with different reproductive numbers pre- and post-lockdown. These parameters are estimated from data on the daily number of confirmed cases in a process that automatically detects the time at which the lockdown became effective. The model is evaluated by considering its predictive accuracy on current data and is then extended to a three-stage version to explore relaxations. The results show the extent to which each country was successful in reducing the reproductive number and demonstrate how the approach is able to model recent increases in the number of cases in all six countries, including the second peak in the US. The results also indicate that the current levels of relaxation in all five European countries could lead to significant second waves that last longer than the corresponding first waves. While there is uncertainty about the implications of these findings at this stage, they do suggest that a lot of vigilance is needed.

A practical computerized decision support system for predicting the severity of Alzheimer's disease of an individual.

Computerized clinical decision support systems can help to provide objective, standardized, and timely dementia diagnosis. However, current computerized systems are mainly based on group analysis, discrete classification of disease stages, or expensive and not readily accessible biomarkers, while current clinical practice relies relatively heavily on cognitive and functional assessments (CFA). In this study, we developed a computational framework using a suite of machine learning tools for identifying key markers in predicting the severity of Alzheimer's disease (AD) from a large set of biological and clinical measures. Six machine learning approaches, namely Kernel Ridge Regression (KRR), Support Vector Regression, and k-Nearest Neighbor for regression and Support Vector Machine (SVM), Random Forest, and k-Nearest Neighbor for classification, were used for the development of predictive models. We demonstrated high predictive power of CFA. Predictive performance of models incorporating CFA was shown to consistently have higher accuracy than those based solely on biomarker modalities. We found that KRR and SVM were the best performing regression and classification methods respectively. The optimal SVM performance was observed for a set of four CFA test scores (FAQ, ADAS13, MoCA, MMSE) with multi-class classification accuracy of 83.0%, 95%CI = (72.1%, 93.8%) while the best performance of the KRR model was reported with combined CFA and MRI neuroimaging data, i.e., R 2 = 0.874, 95%CI = (0.827, 0.922). Given the high predictive power of CFA and their widespread use in clinical practice, we then designed a data-driven and self-adaptive computerized clinical decision support system (CDSS) prototype for evaluating the severity of AD of an individual on a continuous spectrum. The system implemented an automated computational approach for data pre-processing, modelling, and validation and used exclusively the scores of selected cognitive measures as data entries. Taken together, we have developed an objective and practical CDSS to aid AD diagnosis.

A hybrid computational approach for efficient Alzheimer's disease classification based on heterogeneous data.

There is currently a lack of an efficient, objective and systemic approach towards the classification of Alzheimer's disease (AD), due to its complex etiology and pathogenesis. As AD is inherently dynamic, it is also not clear how the relationships among AD indicators vary over time. To address these issues, we propose a hybrid computational approach for AD classification and evaluate it on the heterogeneous longitudinal AIBL dataset. Specifically, using clinical dementia rating as an index of AD severity, the most important indicators (mini-mental state examination, logical memory recall, grey matter and cerebrospinal volumes from MRI and active voxels from PiB-PET brain scans, ApoE, and age) can be automatically identified from parallel data mining algorithms. In this work, Bayesian network modelling across different time points is used to identify and visualize time-varying relationships among the significant features, and importantly, in an efficient way using only coarse-grained data. Crucially, our approach suggests key data features and their appropriate combinations that are relevant for AD severity classification with high accuracy. Overall, our study provides insights into AD developments and demonstrates the potential of our approach in supporting efficient AD diagnosis.

Integration of genomic data for inferring protein complexes from global protein-protein interaction networks.

Protein-protein interactions (PPIs) play crucial roles in virtually every aspect of cellular function within an organism. One important objective of modern biology is the extraction of functional modules, such as protein complexes from global protein interaction networks. This paper describes how seven genomic features and four experimental interaction data sets were combined using a Bayesian-networks-based data integration approach to infer PPI networks in yeast. Greater coverage and higher accuracy were achieved than in previous high-throughput studies of PPI networks in yeast. A Markov clustering algorithm was then used to extract protein complexes from the inferred protein interaction networks. The quality of the computed complexes was evaluated using the hand-curated complexes from the Munich Information Center for Protein Sequences database and gene-ontology-driven semantic similarity. The results indicated that, by integrating multiple genomic information sources, a better clustering result was obtained in terms of both statistical measures and biological relevance.