Suitability of PSA-detected localised prostate cancers for focal therapy: experience from the ProtecT study.

BACKGROUND: Contemporary screening for prostate cancer frequently identifies small volume, low-grade lesions. Some clinicians have advocated focal prostatic ablation as an alternative to more aggressive interventions to manage these lesions. To identify which patients might benefit from focal ablative techniques, we analysed the surgical specimens of a large sample of population-detected men undergoing radical prostatectomy as part of a randomised clinical trial. METHODS: Surgical specimens from 525 men who underwent prostatectomy within the ProtecT study were analysed to determine tumour volume, location and grade. These findings were compared with information available in the biopsy specimen to examine whether focal therapy could be provided appropriately. RESULTS: Solitary cancers were found in prostatectomy specimens from 19% (100 out of 525) of men. In addition, 73 out of 425 (17%) men had multiple cancers with a solitary significant tumour focus. Thus, 173 out of 525 (33%) men had tumours potentially suitable for focal therapy. The majority of these were small, well-differentiated lesions that appeared to be pathologically insignificant (38-66%). Criteria used to select patients for focal prostatic ablation underestimated the cancer's significance in 26% (34 out of 130) of men and resulted in overtreatment in more than half. Only 18% (24 out of 130) of men presumed eligible for focal therapy, actually had significant solitary lesions. CONCLUSION: Focal therapy appears inappropriate for the majority of men presenting with prostate-specific antigen-detected localised prostate cancer. Unifocal prostate cancers suitable for focal ablation are difficult to identify pre-operatively using biopsy alone. Most lesions meeting criteria for focal ablation were either more aggressive than expected or posed little threat of progression.

Intelligent model-based advisory system for the management of ventilated intensive care patients. Part II: Advisory system design and evaluation.

The optimisation of ventilatory support is a crucial issue for the management of respiratory failure in critically ill patients, aiming at improving gas exchange while preventing ventilator-induced dysfunction of the respiratory system. Clinicians often rely on their knowledge/experience and regular observation of the patient's response for adjusting the level of respiratory support. Using a similar data-driven decision-making methodology, an adaptive model-based advisory system has been designed for the clinical monitoring and management of mechanically ventilated patients. The hybrid blood gas patient model SOPAVent developed in Part I of this paper and validated against clinical data for a range of patients lung abnormalities is embedded into the advisory system to predict continuously and non-invasively the patient's respiratory response to changes in the ventilator settings. The choice of appropriate ventilator settings involves finding a balance among a selection of fundamentally competing therapeutic decisions. The design approach used here is based on a goal-directed multi-objective optimisation strategy to determine the optimal ventilator settings that effectively restore gas exchange and promote improved patient's clinical conditions. As an initial step to its clinical validation, the advisory system's closed-loop stability and performance have been assessed in a series of simulations scenarios reconstructed from real ICU patients data. The results show that the designed advisory system can generate good ventilator-setting advice under patient state changes and competing ventilator management targets.

Intelligent model-based advisory system for the management of ventilated intensive care patients: Hybrid blood gas patient model.

Arterial blood gas (ABG) analyses are essential for assessing the acid-base status and guiding the adjustment of mechanical ventilation in critically ill patients. Conventional ABG sampling requires repeated arterial punctures or the insertion of an arterial catheter causing pain, haemorrhage and thrombosis to the patients. Less invasive and non-invasive blood gas analysers, with a technology still in transition, have offered some promise in the recent years. SOPAVent (Simulation of Patients under Artificial Ventilation) is a five compartment blood gas model which captures the basic features of respiratory physiology and gas exchange in the human lungs. It uses ventilator settings and routinely monitored physiological parameters as inputs to produce steady-state estimates of the patient's ABG. This paper overviews the original SOPAVent model and presents an improved data-driven hybrid model that is patient-specific and gives continuous and totally non-invasive ABG predictions. The model has been comprehensively tested in simulations and validated using recorded measurements of ABG and ventilator parameters from ICU patients.

Artificial intelligence and bladder cancer arrays.

Non-muscle invasive bladder cancer is a heterogenous disease whose management is dependent upon the risk of progression to muscle invasion. Although the recurrence rate is high, the majority of tumors are indolent and can be managed by endoscopic means alone. The prognosis of muscle invasion is poor and radical treatment is required if cure is to be obtained. Progression risk in non-invasive tumors is hard to determine at tumor diagnosis using current clinicopathological means. To improve the accuracy of progression prediction various biomarkers have been evaluated. To discover novel biomarkers several authors have used gene expression microarrays. Various statistical methods have been described to interpret array data, but to date no biomarkers have entered clinical practice. Here, we describe a new method of microarray analysis using neurofuzzy modeling (NFM), a form of artificial intelligence, and integrate it with artificial neural networks (ANN) to investigate non-muscle invasive bladder cancer array data (n=66 tumors). We develop a predictive panel of 11 genes, from 2800 expressed genes, that can significantly identify tumor progression (average Logrank p = 0.0288) in the analyzed cancers. In comparison, this panel appears superior to those genes chosen using traditional analyses (average Logrank p = 0.3455) and tumor grade (Logrank, p = 0.2475) in this non-muscle invasive cohort. We then analyze panel members in a new non-muscle invasive bladder cancer cohort (n=199) using immunohistochemistry with six commercially available antibodies. The combination of 6 genes (LIG3, TNFRSF6, KRT18, ICAM1, DSG2 and BRCA2) significantly stratifies tumor progression (Logrank p = 0.0096) in the new cohort. We discuss the benefits of the transparent NFM approach with respect to other reported methods.

Comparative study of intelligent models for the prediction of bladder cancer progression.

New techniques for the prediction of tumour behaviour are needed since statistical analysis has low accuracy and is not applicable to the individual. Artificial intelligence (AI) may provide suitable methods. We have compared the predictive accuracies of neuro-fuzzy modelling (NFM), artificial neural networks (ANN) and traditional statistical methods for the prediction of bladder cancer. Experimental molecular biomarkers, including p53 expression and gene methylation, and conventional clinicopathological data were studied in a cohort of 117 patients with bladder cancer. For all 3 methods, models were produced to predict the presence and timing of tumour progression. Both methods of AI predicted progression with an accuracy ranging from 88-100%, which was superior to logistic regression, and NFM appeared to be better than ANN at predicting the timing of progression.

Adaptive ventilator FiO2 advisor: use of non-invasive estimations of shunt.

A non-invasive and simple method of parameter estimation has been developed for the model-based decision support of the artificial ventilation in intensive care units. The parameter concerned was the respiratory shunt. Originally, the shunt had to be estimated using a numerical algorithm, which was slow and unreliable. The estimation process also required the knowledge of other parameters, whose values could only be obtained using invasive monitoring equipment. In this paper, the respiratory index is used for the shunt estimation. A linear regression model and a non-linear adaptive neuro-fuzzy inference system (ANFIS) model were used to describe the relationship between the respiratory index and the shunt. The shunts estimated using these models were then used to calculate the fractional inspired oxygen needed to attain the target arterial oxygen level of the model patient. The advisor also utilises population median values of the cardiac index and oxygen consumption index. This alleviates the need for invasive monitoring. In a simulation study, the mean squared error of the control using the ANFIS model was 0.75 kPa2 compared to 2.06 kPa2 using the linear regression model. Therefore, the performance of the FiO2 advisor was better when the shunt was estimated using the non-linear ANFIS model.

Migrating from target-controlled infusion to closed-loop control in general anaesthesia.

The target-controlled infusion (TCI) technique has been successfully and commercially used in clinical general anaesthesia with the intravenous anaesthetic agent propofol. The technique is based on a population pharmacokinetic model and is an open-loop control system. Closed-loop control requires a reliable and consistent signal for feedback utilisation. With all anaesthetic agents the somatosensory evoked potentials (SEP) have been shown to give increased latency as anaesthetic depth is increased. Using infusion rate and SEP response data from rats anaesthetised with propofol a mathematical model was derived to describe the anaesthetic process. This model was used as a design reference to develop a proportional integral (PI) closed-loop control system using SEP as the feedback measure. A serials of 10 trials were conducted to investigate the difference between continuous bolus injection and infusion, all under closed-loop control. The trials showed that the use of SEPs in closed-loop control of anaesthesia is feasible.

Rule-base derivation for intensive care ventilator control using ANFIS.

In recent years, much research has been done on the use of fuzzy systems in medicine. The fuzzy rule-bases have usually been derived after extensive discussion with the clinical experts. This takes a lot of time from the clinical experts and the knowledge engineers. This paper presents the use of the adaptive neuro-fuzzy inference system (ANFIS) in rule-base derivation for ventilator control. The change of the inspired fraction of oxygen (FiO(2)) advised by eight clinical experts responding to 71 clinical scenarios was recorded. ANFIS and a multilayer perceptron (MLP) were then used to model the relationship between the inputs (the arterial oxygen tension (PaO(2)), FiO(2) and the positive end-expiratory pressure (PEEP) level) and the change in FiO(2) suggested. Compared to a previous fuzzy advisor (FAVeM), both the ANFIS and the MLP were found to correlate with the clinicians' decision better (correlation coefficient of 0.694 and 0.701, respectively compared to 0.630). A formerly developed model-based radial basis network advisor (RBN-MB) was used for comparison. Closed-loop simulations showed that the ANFIS, MLP and the RBN-MB's performance were comparable to the clinicians' performance (correlation coefficients of 0.852, 0.962 and 0.787, respectively). The FAVeM's performance differed from the clinicians' performance (correlation coefficient of 0.332) but the resulting PaO(2) was still within safety limits.

Neuronal network modelling of the effects of anaesthetic agents on somatosensory pathways.

The whole question of consciousness, awareness and depth of anaesthesia is both timely, little understood and deeply challenging. Models of the underlying neural pathway mechanisms/dynamics are necessary for understanding the interactions involved and their structure and function. A neuronal network of the somatosensory pathways is proposed in this paper based on experimental information and physiological investigation into anaesthesia. Existing mathematical neuronal models from the literature have been modified and then employed to describe the dynamics of the proposed pathway network. Effects of anaesthetic agents on the cortex were simulated in the model which describes the evoked cortical responses. By comparison with responses from anaesthetised rats, the model's responses are able to describe the dynamics of typical responses. Thus, the proposed model promises to be valuable for investigating the mechanisms of anaesthesia on the cortex and the effects of brain lesions.

Online elicitation of Mamdani-type fuzzy rules via TSK-based generalized predictive control.

Many synergies have been proposed between soft-computing techniques, such as neural networks (NNs), fuzzy logic (FL), and genetic algorithms (GAs), which have shown that such hybrid structures can work well and also add more robustness to the control system design. In this paper, a new control architecture is proposed whereby the on-line generated fuzzy rules relating to the self-organizing fuzzy logic controller (SOFLC) are obtained via integration with the popular generalized predictive control (GPC) algorithm using a Takagi-Sugeno-Kang (TSK)-based controlled autoregressive integrated moving average (CARIMA) model structure. In this approach, GPC replaces the performance index (PI) table which, as an incremental model, is traditionally used to discover, amend, and delete the rules. Because the GPC sequence is computed using predicted future outputs, the new hybrid approach rewards the time-delay very well. The new generic approach, named generalized predictive self-organizing fuzzy logic control (GPSOFLC), is simulated on a well-known nonlinear chemical process, the distillation column, and is shown to produce an effective fuzzy rule-base in both qualitative (minimum number of generated rules) and quantitative (good rules) terms.

Survey on the use of smart and adaptive engineering systems in medicine.

In this paper, the current published knowledge about smart and adaptive engineering systems in medicine is reviewed. The achievements of frontier research in this particular field within medical engineering are described. A multi-disciplinary approach to the applications of adaptive systems is observed from the literature surveyed. The three modalities of diagnosis, imaging and therapy are considered to be an appropriate classification method for the analysis of smart systems being applied to specified medical sub-disciplines. It is expected that future research in biomedicine should identify subject areas where more advanced intelligent systems could be applied than is currently evident. The literature provides evidence of hybridisation of different types of adaptive and smart systems with applications in different areas of medical specifications.

Analysis of solvent central nervous system toxicity and ethanol interactions using a human population physiologically based kinetic and dynamic model.

The effect of acute ethanol-mediated inhibition of m-xylene metabolism on central nervous system (CNS) depression in the human worker population was investigated using physiologically based pharmacokinetic (PBPK) models and probabilistic random (Monte Carlo) sampling. PBPK models of inhaled m-xylene and orally ingested ethanol were developed and combined by a competitive enzyme (CYP2E1) inhibition model. Human interindividual variability was modeled by combining estimated statistical distributions of model parameters with the deterministic PBPK models and multiple random or Monte Carlo simulations. A simple threshold pharmacodynamic model was obtained by simulating m-xylene kinetics in human studies where CNS effects were observed and assigning the peak venous blood m-xylene concentration (C(V,max)) as the dose surrogate of toxicity. Probabilistic estimates of an individual experiencing CNS disturbances given exposure to the current UK occupational exposure standard (100 ppm time-weighted average over 8 h), with and without ethanol ingestion, were obtained. The probability of experiencing CNS effects given this scenario increases markedly and nonlinearly with ethanol dose. As CYP2E1-mediated metabolism of other occupationally relevant organic compounds may be inhibited by ethanol, simulation studies of this type should have an increasingly significant role in the chemical toxicity risk assessment.

A survey of fuzzy logic monitoring and control utilisation in medicine.

Intelligent systems have appeared in many technical areas, such as consumer electronics, robotics and industrial control systems. Many of these intelligent systems are based on fuzzy control strategies which describe complex systems mathematical models in terms of linguistic rules. Since the 1980s new techniques have appeared from which fuzzy logic has been applied extensively in medical systems. The justification for such intelligent systems driven solutions is that biological systems are so complex that the development of computerised systems within such environments is not always a straightforward exercise. In practice, a precise model may not exist for biological systems or it may be too difficult to model. In most cases fuzzy logic is considered to be an ideal tool as human minds work from approximate data, extract meaningful information and produce crisp solutions. This paper surveys the utilisation of fuzzy logic control and monitoring in medical sciences with an analysis of its possible future penetration.

The intelligent systems in biomedicine laboratory in the Department of Automatic Control and Systems Engineering at the University of Sheffield, UK.

The group has been engaged in research into modelling and control in biomedicine for many years. Initially, this used quantitative approaches but increasingly this has moved towards qualitative techniques, in particular that of fuzzy logic. The current emphasis is on hybrid models (quantitative/qualitative) and hybrid intelligent algorithms (fuzzy logic, neural networks, evolutionary computing) applied into anaesthesia.

A systematic neuro-fuzzy modeling framework with application to material property prediction.

A systematic neural-fuzzy modeling framework that includes the initial fuzzy model self-generation, significant input selection, partition validation, parameter optimization, and rule-base simplification is proposed in this paper. In this framework, the structure identification and parameter optimization are carried out automatically and efficiently by the combined use of a sell-organization network, fuzzy clustering, adaptive back-propagation learning, and similarity analysis-based model simplification. The proposed neuro-fuzzy modeling approach has been used for nonlinear system identification and mechanical property prediction in hot-rolled steels from construct composition and microstructure data. Experimental studies demonstrate that the predicted mechanical properties have a good agreement with the measured data by using the elicited fuzzy model with a small number of rules.

Somatosensory evoked potentials for closed-loop control of anaesthetic depth using propofol in the urethane-anaesthetized rat.

Primary somatosensory cortical mass responses have been shown to exhibit dose-dependent changes in latency when general anaesthetics are administered. Here we describe a system in which the latency of evoked responses was measured automatically in real time in five animals. Latency changes were used to operate a closed-loop control of propofol delivery by intravenous infusion. The system attempted to induce and maintain a 1 ms increase in evoked response latency; this was reversed when infusion was discontinued. Allowing for the rapid and large biological fluctuations in the evoked response, this was achieved successfully. The system maintained a mean increase in latency of 1.27 (SD 0.42) ms. The mean statistical dispersion index of data obtained during the controlled period was 1.23 (0.3); in an ideal controllable system it approximates to 1. Such a system may provide a means for the automatic delivery of anaesthetics.

Intelligent control of a cryogenic cooling plant based on blackboard system architecture.

Intelligent system techniques have been rapidly assimilating into process control engineering, with many applications reported in the last decade. Intelligent control is bringing a new perspective as well as new challenges to process control. In this paper, a software architecture for a Blackboard for Integrated Intelligent Control Systems (BIICS) is described. The system is designed to simultaneously support multiple heterogeneous intelligent methodologies, such as neural networks. expert systems, fuzzy logic, neural networks and genetic algorithms. It will be shown how such methodologies can be readily assimilated into the software architecture. The BIICS system represents a multi-purpose platform for design and simulation of intelligent control paradigms for different kinds of processes. Currently the system utilizes intelligent control techniques (neuro-fuzzy and genetic optimization) for controlling a cryogenic plant used for superconductor testing at temperatures below 100 K.

Self-learning fuzzy control with temporal knowledge for atracurium-induced neuromuscular block during surgery.

Self-learning fuzzy logic control has the important property of accommodating uncertain, nonlinear, and time-varying process characteristics. This intelligent control scheme starts with no fuzzy control rules and learns how to control each process presented to it in real time without the need for detailed process modeling. In this study we utilize temporal knowledge of generated rules to improve control performance. A suitable medical application to investigate this control strategy is atracurium-induced neuromuscular block of patients in the operating theater where the patient response exhibits high nonlinearity and individual patient dose requirements may vary fivefold during an operating procedure. We developed a computer control system utilizing Relaxograph (Datex) measurements to assess the clinical performance of a self-learning fuzzy controller in this application. Using a T1 setpoint of 10% of baseline in 10 patients undergoing general surgery, we found a mean T1 error of 0.28% (SD = 0.39%) while accommodating a 0.25 to 0.38 mg/kg/h range in the mean atracurium infusion rate. This result compares favorably with more complex and computationally intensive model-based control strategies for atracurium infusion.

Internal representation in neural networks used for classification of patient anaesthetic states and dosage.

In this study we aimed to explore the ability of artificial neural networks (ANN) to classify patient anaesthetic states and dosage. Surgical data obtained under different states of anaesthesia and dose levels were modelled via this approach. It is shown that inferential parameters can be used to determine the patient anaesthetic states and drug dosage. In addition to demonstrating the capability of ANN for classification we were interested in the internal representations that are developed automatically by networks while they are learning their processing task. An unsupervised learning procedure of clustering via which the classes are inferred from the data and a supervised learning technique of discrimination via which to construct a classification of the known categories were applied to analyse the performance of the ANN. Discriminant analysis (DA) was also utilised to optimise the network architecture.

Dynamic monitoring and control of patient anaesthetic and dose levels: time-delay, moving-average neural networks, and principal components analysis.

The goal of this study was to examine the capabilities of neural network models for dynamic monitoring and control of patient anaesthetic and dose levels. The network models that we considered are split into two basic groups: static networks and dynamic networks. Static networks are characterised by equations that are memoryless. On the other hand, dynamic networks are systems with memory. Additionally, principal components analysis was used to introduce a further improvement to network design by reducing the dimensionality of the encoded temporal information. Principal components analysis was applied as both pre-processing and post-processing techniques. In the first instance it was used to reduce the dimensionality of the data to more manageable intrinsic information. In the second instance it was employed to understand how the hidden layers separate the data, in order to optimise the network architecture.