A Robust Blood-based Signature of Cerebrospinal Fluid Aß42 Status.

Early detection of Alzheimer's disease (AD) is of vital importance in the development of disease-modifying therapies. This necessitates the use of early pathological indicators of the disease such as amyloid abnormality to identify individuals at early disease stages where intervention is likely to be most effective. Recent evidence suggests that cerebrospinal fluid (CSF) amyloid ß1-42 (Aß42) level may indicate AD risk earlier compared to amyloid positron emission tomography (PET). However, the method of collecting CSF is invasive. Blood-based biomarkers indicative of CSF Aß42 status may remedy this limitation as blood collection is minimally invasive and inexpensive. In this study, we show that APOE4 genotype and blood markers comprising EOT3, APOC1, CGA, and Aß42 robustly predict CSF Aß42 with high classification performance (0.84 AUC, 0.82 sensitivity, 0.62 specificity, 0.81 PPV and 0.64 NPV) using machine learning approach. Due to the method employed in the biomarker search, the identified biomarker signature maintained high performance in more than a single machine learning algorithm, indicating potential to generalize well. A minimally invasive and cost-effective solution to detecting amyloid abnormality such as proposed in this study may be used as a first step in a multi-stage diagnostic workup to facilitate enrichment of clinical trials and population-based screening.

Identification of Optimum Panel of Blood-based Biomarkers for Alzheimer's Disease Diagnosis Using Machine Learning.

With the increasing number of people living with Alzheimer's disease (AD), there is a need for low-cost and easy to use methods to detect AD early to facilitate access to appropriate care pathways. Neuroimaging biomarkers (such as those based on PET and MRI) and biochemical biomarkers (such as those based on CSF) are recommended by international guidelines to facilitate diagnosis. However, neuroimaging is expensive and may not be widely available and CSF testing is invasive. Bloodbased biomarkers offer the potential for the development of a low-cost and more time efficient tool to detect AD to complement CSF and neuroimaging as blood is much easier to obtain. Although no single blood biomarker is yet able to detect AD, combinations of biomarkers (also called panels) have shown good results. However, a large number of biomarkers are often needed to achieve a satisfactory detection performance. In addition, it is difficult to reproduce reported results within and across different study cohorts because of data overfitting and lack of access to the datasets used in the studies. In this study, our focus is to identify an optimum panel (in terms of the least number of blood biomarkers to meet the specified diagnostic performance of 80% sensitivity and specificity) based on a widely accessible data set, and to demonstrate a testing methodology that reinforces reproducibility of results. Realizing a panel with reduced number of markers will have significant impact on the complexity and cost of diagnosis and potential development of cost-effective point of care devices.

Identification of blood biomarkers for use in point of care diagnosis tool for Alzheimer's disease.

Early diagnosis of Alzheimer's Disease (AD) is widely regarded as necessary to allow treatment to be started before irreversible damage to the brain occur and for patients to benefit from new therapies as they become available. Low-cost point-of-care (PoC) diagnostic tools that can be used to routinely diagnose AD in its early stage would facilitate this, but such tools require reliable and accurate biomarkers. However, traditional biomarkers for AD use invasive cerebrospinal fluid (CSF) analysis and/or expensive neuroimaging techniques together with neuropsychological assessments. Blood-based PoC diagnostics tools may provide a more cost and time efficient way to assess AD to complement CSF and neuroimaging techniques. However, evidence to date suggests that only a panel of biomarkers would provide the diagnostic accuracy needed in clinical practice and that the number of biomarkers in such panels can be large. In addition, the biomarkers in a panel vary from study to study. These issues make it difficult to realise a PoC device for diagnosis of AD. An objective of this paper is to find an optimum number of blood biomarkers (in terms of number of biomarkers and sensitivity/specificity) that can be used in a handheld PoC device for AD diagnosis. We used the Alzheimer's disease Neuroimaging Initiative (ADNI) database to identify a small number of blood biomarkers for AD. We identified a 6-biomarker panel (which includes A1Micro, A2Macro, AAT, ApoE, complement C3 and PPP), which when used with age as covariate, was able to discriminate between AD patients and normal subjects with a sensitivity of 85.4% and specificity of 78.6%.

Double-blind evaluation and benchmarking of survival models in a multi-centre study.

Accurate modelling of time-to-event data is of particular importance for both exploratory and predictive analysis in cancer, and can have a direct impact on clinical care. This study presents a detailed double-blind evaluation of the accuracy in out-of-sample prediction of mortality from two generic non-linear models, using artificial neural networks benchmarked against a partial logistic spline, log-normal and COX regression models. A data set containing 2880 samples was shared over the Internet using a purpose-built secure environment called GEOCONDA (www.geoconda.com). The evaluation was carried out in three parts. The first was a comparison between the predicted survival estimates for each of the four survival groups defined by the TNM staging system, against the empirical estimates derived by the Kaplan-Meier method. The second approach focused on the accurate prediction of survival over time, quantified with the time dependent C index (C(td)). Finally, calibration plots were obtained over the range of follow-up and tested using a generalization of the Hosmer-Lemeshow test. All models showed satisfactory performance, with values of C(td) of about 0.7. None of the models showed a systematic tendency towards over/under estimation of the observed survival at tau=3 and 5 years. At tau=10 years, all models underestimated the observed survival, except for COX regression which returned an overestimate. The study presents a robust and unbiased benchmarking methodology using a bespoke web facility. It was concluded that powerful, recent flexible modelling algorithms show a comparative predictive performance to that of more established methods from the medical and biological literature, for the reference data set.

Bioprofiling over grid for eHealthcare.

A trend in modern medicine is towards individualization of healthcare and, potentially, grid computing can play an important role in this by allowing sharing of resources and expertise to improve the quality of care. In this paper, we present a new test bed, the BIOPATTERN Grid, which aims to fulfil this role in the long term. The main objectives in this paper are 1) to report the development of the BIOPATTERN Grid, for biopattern analysis and bioprofiling in support of individualization of healthcare. The BIOPATTERN Grid is designed to facilitate secure and seamless sharing of geographically distributed bioprofile databases and to support the analysis of bioprofiles to combat major diseases such as brain diseases and cancer within a major EU project, BIOPATTERN (www.biopattern.org); 2) to illustrate how the BIOPATTERN Grid could be used for biopattern analysis and bioprofiling for early detection of dementia and for brain injury assessment on an individual basis. We highlight important issues that would arise from the mobility of citizens in the EU, such as those associated with access to medical data, ethical and security; and 3) to describe two grid services which aim to integrate BIOPATTERN Grid with existing grid projects on crawling service and remote data acquisition which is necessary to underpin the use of the test bed for biopattern analysis and bioprofiling.

Biopattern analysis and subject-specific diagnosis and care of dementia.

Worldwide, the number of people that develop Alzheimer's disease and other types of dementia is rapidly rising and will create a considerable financial burden on the health and social services. The availability of new drugs that may slow or even halt the disease progression makes accurate early detection crucial. Objective methods are needed to support clinical diagnosis and care for patients; to quantify severity, monitor progression and response to new treatments. Electrophysiological markers have an important role to play in the objective assessment and care for dementia. The EEG provides a measure of brain dysfunction and EEG changes could be detected fairly early in the dementing process. Subject-specific EEG analysis offers the possibility of using objective methods to assess and care for dementia on an individual basis. The main objectives of this paper are: (i) to introduce the concepts of subject-specific EEG analysis as a basis for improving diagnosis and care for dementia; and (ii) present two novel methods for deriving suitable subject-specific electrophysiological markers analysis of fractal dimension and zero crossing interval density of the EEG. We present findings that indicate that the methods are potentially good candidates for the development of individualized, low-cost, easy to administer and reasonably accurate methods for detecting dementia within the growing at risk population.

Nonlinear methods for biopattern analysis: role and challenges.

An important trend in medical technology is towards support for personalised healthcare, fuelled by developments in genomic-based medicine. New computational intelligent techniques for biodata analysis will be needed to fully exploit the vast amounts of data that are being generated. Non-linear signal processing methods will form an important part of such computational intelligent techniques. This paper introduces some non-linear methods which are likely to play a role in the emerging area of biopattern and bioprofile analysis that will underpin personalized healthcare. We highlight their application to clinical problems involving EEG and fetal ECG and heart rate analysis, and issues that arise when they are applied to real world problems. The clinical problems include dementia assessment, drug administration and fetal monitoring. The potential role and challenges in the application of non-linear signal analysis of biopattern and bioprofile are highlighted within the context of a major EU project, BIOPATTERN.

Receiver operating characteristic analysis for intelligent medical systems--a new approach for finding confidence intervals.

Intelligent systems are increasingly being deployed in medicine and healthcare, but there is a need for a robust and objective methodology for evaluating such systems. Potentially, receiver operating characteristic (ROC) analysis could form a basis for the objective evaluation of intelligent medical systems. However, it has several weaknesses when applied to the types of data used to evaluate intelligent medical systems. First, small data sets are often used, which are unsatisfactory with existing methods. Second, many existing ROC methods use parametric assumptions which may not always be valid for the test cases selected. Third, system evaluations are often more concerned with particular, clinically meaningful, points on the curve, rather than on global indexes such as the more commonly used area under the curve. A novel, robust and accurate method is proposed, derived from first principles, which calculates the probability density function (pdf) for each point on a ROC curve for any given sample size. Confidence intervals are produced as contours on the pdf. The theoretical work has been validated by Monte Carlo simulations. It has also been applied to two real-world examples of ROC analysis, taken from the literature (classification of mammograms and differential diagnosis of pancreatic diseases), to investigate the confidence surfaces produced for real cases, and to illustrate how analysis of system performance can be enhanced. We illustrate the impact of sample size on system performance from analysis of ROC pdf's and 95% confidence boundaries. This work establishes an important new method for generating pdf's, and provides an accurate and robust method of producing confidence intervals for ROC curves for the small sample sizes typical of intelligent medical systems. It is conjectured that, potentially, the method could be extended to determine risks associated with the deployment of intelligent medical systems in clinical practice.

The evaluation of an expert system for the analysis of umbilical cord blood.

An assessment of neonatal outcome may be obtained from analysis of blood in the umbilical cord of an infant immediately after delivery. This can provide information on the health of the new-born infant, guide requirements for neonatal care, but there are problems with the technique. Samples frequently contain errors in one or more of the important parameters, preventing accurate interpretation and many clinical staff lack the expert knowledge required to interpret error-free results. The development and implementation of an expert system to overcome these difficulties has previously been described. This expert system validates the raw data, provides an interpretation of the results for clinicians and archives all the results, including the quality control and calibration data, for permanent storage. Issues regarding the clinical evaluation of this system are now detailed further, along with some clinical results illustrating the potential of such a system.

The development and implementation of an expert system for the analysis of umbilical cord blood.

An assessment of neonatal outcome may be obtained from analysis of blood in the umbilical cord of the infant immediately after delivery. This can provide information on the health of the newborn infant, guide requirements for neonatal care, and is recommended practice of the Royal College of Obstetricians and Gynaecologists. However, there are problems with the technique. Samples frequently contain errors in one or more of the important parameters, preventing accurate interpretation and many clinical staff lack the expert knowledge required to interpret error-free results. In this paper the development and implementation of an expert system to overcome these difficulties is described. The expert system validates results, provides a textual interpretation and archives all results to database for audit, research and medico-legal purposes. The system has now been in routine clinical use for over 3 years in Plymouth, and has also been installed in several other hospitals in the UK. Results are presented in which the types and frequency of errors are established and the user acceptance of the system is determined.

A multicentre comparative study of 17 experts and an intelligent computer system for managing labour using the cardiotocogram.

OBJECTIVES: To investigate 1. whether an intelligent computer system could obtain a performance in labour management comparable with experts when using cardiotocograms (CTGs), patient information, and fetal blood sampling and 2. whether experts could be consistent and agree in their management of labour. SUBJECTS: An intelligent computer system and 17 clinicians experienced in fetal monitoring from 16 centres in the UK. DESIGN: Fifty cases with complete intrapartum CTGs and clinical data were reviewed by each expert and the system independently on two occasions, at least one month apart. Each CTG was scored in 15 min segments according to a protocol and estimates of the cervical dilatation and fetal scalp blood pH were given when requested. MAIN OUTCOME MEASURES: Consistency and agreement in the recorded scores, agreement and timing of cases recommended for caesarean sections, fetal blood sampling rates, intervention in cases with poor outcome and intervention in cases with good clinical outcome. RESULTS: The system: 1. Agreed with experts well and significantly better than chance (67.33%, kappa = 0.31, P << 0.001). 2. Was highly consistent (99.16%, kappa = 0.98, P << 0.001) when used by two operators independently. 3. Recommended no unnecessary intervention in cases with normal delivery and good condition (cord artery pH > 7.15, vein pH > 7.20, 5 min Apgar > or = 9 and no resuscitation). This was better than all but two of the experts. 4. Recommended delivery by caesarean section in 11 cases; at least 15 of the 17 experts in each review also recommended caesarean section delivery in these cases. The majority did so within 15 min of the system and two-thirds did so within 30 min. 5. Identified as many of the birth asphyxiated cases (cord arterial pH < 7.05 and BDecf > or = 12, and Apgar score at 5 min < or = 7 with neonatal morbidity) as the majority of experts and one more than was acted upon clinically. The experts were found to be consistent and to agree. There was good agreement in the cases and the timing of caesarean section recommendations. The majority of experts did not recommend operative intervention in cases which had a normal delivery and good outcome, but did recommend operative interventions in 10 of 12 cases delivered with cord arterial pH < 7.05. However, in one of the cases delivered with birth asphyxia, 14 of the 17 experts and the system failed to recommend intervention. CONCLUSIONS: The system's performance was found to be indistinguishable from the experts' in the 50 cases examined, but it was more consistent. This demonstrates the potential for an intelligent computer system to improve the interpretation of the CTG and decrease intervention. Furthermore, the good performance of most experts in this study demonstrates the potential effectiveness of the CTG and raises important questions regarding why the CTG has fallen short of expectations in current practice.

Suitability of artificial neural networks for feature extraction from cardiotocogram during labour.

Fetal condition during labour is inferred from a continuous display of fetal heart rate and uterine contractions called the cardiotocogram (CTG). The CTG requires a considerable expertise for correct interpretation, which is not always available. We are developing an intelligent system to support clinical decision-making during labour. The system's performance depends on its ability to classify features from the CTG similarly to experts. Artificial neural networks (NNs) can be taught by experts for such tasks, and so may be particularly suitable. We found NNs suitable for feature extraction when the problem was reduced to small well defined tasks, and numerical algorithms were used to pre-process the raw data before application to the NNs. A NN with optimised dimensions was used in this way to classify the magnitude of decelerations, a feature clinicians find particularly difficult. The NN was compared with the algorithm used in a commercial antenatal monitor and six reviewers which included two CTG experts. The experts were consistent (89.7% and 97.0%) and agreed well with each other (81.0%), whereas the non-experts were less consistent and agreed less well. The NN agreed well with the experts (75.0% and 81.9%) but the algorithm agreed poorly (56.5% and 68.9%). It was found that the algorithm's performance could be improved (72.1% and 76.7%) when modified to use additional information. Our earlier attempts to fully classify the raw CTG using a single NN were unsuccessful because of the large number of data patterns. A simplified approach to classify the magnitude and timing of decelerations was also unsuitable when contraction data was of poor quality or absent.(ABSTRACT TRUNCATED AT 250 WORDS)

Preliminary evaluation of an intelligent system for the management of labour.

Over the past 10-15 years, workers using conventional computing approaches have attempted to provide an accurate assessment of fetal condition during labour based on the cardiotocogram (CTG) alone. These have not proved successful perhaps because the correct interpretation of fetal condition also requires physiological knowledge, specific patient information, knowledge of events during labour and considerable practical experience. An intelligent system which considers all the relevant information and embodies expertise may better diagnose fetal condition and support decision making. This study reports the preliminary evaluation of such a system and investigates whether this approach can attain a performance comparable with experienced local clinicians. From a database of 200 high risk labour records, 30 cases were selected; the 9 cases which received clinical intervention for 'fetal-distress' and a further 21 cases selected randomly. The management specified by the system, 3 experienced clinicians (A, B and C) and the actual clinical management were compared in a retrospective blinded review. The experts were found to agree well with each other. Expert A reviewed the cases five months later and was found to be entirely consistent in the management of 28 of the 30 cases. The system's actions were indistinguishable from the experts' and in no case did the system recommend an action not also recommended by at least one experienced reviewer. This study demonstrates the potential of an intelligent system to assist in the management of labour.

Suitability of fetal scalp electrodes for monitoring the fetal electrocardiogram during labour.

As the limitations of heart-rate based intrapartum monitoring have become apparent, there is renewed interest in analysis of the fetal electrocardiographic waveform as obtained from a fetal scalp electrode. A high quality ECG signal is necessary for waveform analysis. This study examined the suitability of five commonly available scalp electrodes for collecting this signal by examining their physical and electrical characteristics, together with a randomised clinical trial in which the ECG trace quality was assessed in 50 patients. The frequency response of Copeland electrodes was such that they attenuate the ECG signal more than the baseline noise. Difficulties were experienced in obtaining optimum attachment and the long, semi-rigid design increased movement artefact resulting in significantly poorer quality ECG signals. Whilst the Hewlett-Packard double spiral electrode had a near ideal frequency response, certain design features made it difficult to apply and remain secure so the clinical signals were of intermediate quality. The Corometrics and Cetro single spirals had the most stable attachment to the scalp and a near ideal frequency response, so produced significantly better signal quality in the clinical trial. Currently, single spiral electrodes are the most suitable for electrocardiographic data collection.