Clinical decision support systems for intensive care units: using artificial neural networks.

The paper provides an overview of applications of artificial neural networks (ANNs) to various medical problems, with a particular focus on the intensive care unit environment (ICU). Several technical approaches were tested to see whether they improve the ANN performance in estimating medical outcomes and resource utilization in adult ICUs. These experiments include: (1) use of the weight-elimination cost function; (2) use of 'high' and 'low' nodes for input variables; (3) verifying the effect of the total number of input variables on the results; (4) testing the impact of the value of the constant predictor on the performance of the ANNs. The developments presented intend to help medical and nursing personnel to assess patient status, assist in making a diagnosis, and facilitate the selection of a course of therapy.     

Rating organ failure via adverse events using data mining in the intensive care unit

OBJECTIVE: The main intensive care unit (ICU) goal is to avoid or reverse the organ failure process by adopting a timely intervention. Within this context, early identification of organ impairment is a key issue. The sequential organ failure assessment (SOFA) is an expert-driven score that is widely used in European ICUs to quantify organ disorder. This work proposes a complementary data-driven approach based on adverse events, defined from commonly monitored biometrics. The aim is to study the impact of these events when predicting the risk of ICU organ failure. MATERIALS AND METHODS: A large database was considered, with a total of 25,215 daily records taken from 4425 patients and 42 European ICUs. The input variables include the case mix (i.e. age, diagnosis, admission type and admission from) and adverse events defined from four bedside physiologic variables (i.e. systolic blood pressure, heart rate, pulse oximeter oxygen saturation and urine output). The output target is the organ status (i.e. normal, dysfunction or failure) of six organ systems (respiratory, coagulation, hepatic, cardiovascular, neurological and renal), as measured by the SOFA score. Two data mining (DM) methods were compared: multinomial logistic regression (MLR) and artificial neural networks (ANNs). These methods were tested in the R statistical environment, using 20 runs of a 5-fold cross-validation scheme. The area under the receiver operator characteristic (ROC) curve and Brier score were used as the discrimination and calibration measures. RESULTS: The best performance was obtained by the ANNs, outperforming the MLR in both discrimination and calibration criteria. The ANNs obtained an average (over all organs) area under the ROC curve of 64, 69 and 74% and Brier scores of 0.18, 0.16 and 0.09 for the dysfunction, normal and failure organ conditions, respectively. In particular, very good results were achieved when predicting renal failure (ROC curve area of 76% and Brier score of 0.06). CONCLUSION: Adverse events, taken from bedside monitored data, are important intermediate outcomes, contributing to a timely recognition of organ dysfunction and failure during ICU length of stay. The obtained results show that it is possible to use DM methods to get knowledge from easy obtainable data, thus making room for the development of intelligent clinical alarm monitoring.     

Pathological voice quality assessment using artificial neural networks

This paper describes a prototype system for the objective assessment of voice quality in patients recovering from various stages of laryngeal cancer. A large database of male subjects steadily phonating the vowel /i/ was used in the study, and the quality of their voices was independently assessed by a speech and language therapist (SALT) according to their seven-point ranking of subjective voice quality. The system extracts salient short-term and long-term time-domain and frequency-domain parameters from impedance (EGG) signals and these are used to train and test an artificial neural network (ANN). Multi-layer perceptron (MLP) ANNs were investigated using various combinations of these parameters, and the best results were obtained using a combination of short-term and long-term parameters, for which an accuracy of 92% was achieved. It is envisaged that this system could be used as an assessment tool, providing a valuable aid to the SALT during clinical evaluation of voice quality.     

EEG dipole source localization using artificial neural networks

Localization of focal electrical activity in the brain using dipole source analysis of the electroencephalogram (EEG), is usually performed by iteratively determining the location and orientation of the dipole source, until optimal correspondence is reached between the dipole source and the measured potential distribution on the head. In this paper, we investigate the use of feed-forward layered artificial neural networks (ANNs) to replace the iterative localization procedure, in order to decrease the calculation time. The localization accuracy of the ANN approach is studied within spherical and realistic head models. Additionally, we investigate the robustness of both the iterative and the ANN approach by observing the influence on the localization error of both noise in the scalp potentials and scalp electrode mislocalizations. Finally, after choosing the ANN structure and size that provides a good trade off between low localization errors and short computation times, we compare the calculation times involved with both the iterative and ANN methods. An average localization error of about 3.5 mm is obtained for both spherical and realistic head models. Moreover, the ANN localization approach appears to be robust to noise and electrode mislocations. In comparison with the iterative localization, the ANN provides a major speed-up of dipole source localization. We conclude that an artificial neural network is a very suitable alternative for iterative dipole source localization in applications where large numbers of dipole localizations have to be performed, provided that an increase of the localization errors by a few millimetres is acceptable.     

Realistic animation of human figures using artificial neural networks

We describe a new approach to the animation of human figures which can produce realistic animation and based on artificial neural networks (ANN). A fully connected ANN is trained with inputs and outputs of key frames obtained from image analysis and key postures and parameters of standing, walking and running. A behaviour index is introduced as an input to the ANN. Each index is unique to each behaviour. Other inputs include speed, cycle history and subsytem index. The subsystem index refers to the different subsytem of the human figure e.g. the right leg is a subsystem referred to by an index. The outputs are the joints displacements. The ANN is trained using the back propagation method. The ANN was able to generate realistic animations of walking and running and could merge three different behaviours; standing, walking and running. The proposed method should enable design evaluations, human factors analysis, task simulation and motion understanding easier for non-animation experts.     

Feature-based classification of myoelectric signals using artificial neural networks

A pattern classification system, designed to separate myoelectric signal records based on contraction tasks, is described. The amplitude of the myoelectric signal during the first 200 ms following the onset of a contraction has a non-random structure that is specific to the task performed. This permits the application of advanced pattern recognition techniques to separate these signals. The pattern classification system described consists of a spectrographic preprocessor, a feature extraction stage and a classifier stage. The preprocessor creates a spectrogram by generating a series of power spectral densities over adjacent time segments of the input signal. The feature extraction stage reduces the dimensionality of the spectrogram by identifying features that correspond to subtle underlying structures in the input signal data. This is realised by a self-organising artificial neural network (ANN) that performs an advanced statistical analysis procedure known as exploratory projection pursuit. The extracted features are then classified by a supervised-learning ANN. An evaluation of the system, in terms of system performance and the complexity of the ANNs, is presented.     

Clinical decision-support systems for intensive care units using case-based reasoning

The artificial intelligence approach used in this work focusses on case-based reasoning techniques for the estimation of medical outcomes and resource utilization. The systems were designed with a view to help medical and nursing personnel to assess patient status, assist in making a diagnosis, and facilitate the selection of a course of therapy. The initial prototype provided information on the closest-matching patient cases to the newest patient admission in an adult intensive care unit (ICU). The system was subsequently re-designed for use in a neonatal ICU. The results of a short clinical pilot evaluation performed in both adult and neonatal units are reported and have led to substantial improvement of the prototype. Future work will include longer-term clinical trials for both adult and neonatal ICUs, once all the software changes have been made to both prototypes in response to the comments of the users made during the preliminary evaluations. To date, the results are very encouraging and physician interest in the potential clinical usefulness of these two systems remains high, and particularly so in the new testing environment in Ottawa.     

Computer-based test-bed for clinical assessment of hand/wrist feed-forward neuroprosthetic controllers using artificial neural networks

Neuroprosthestic systems can be used to restore hand grasp and wrist control in individuals with C5/C6 spinal cord injury. A computer-based system was developed for the implementation, tuning and clinical assessment of neuroprosthetic controllers, using off-the-shelf hardware and software. The computer system turned a Pentium III PC running Windows NT into a non-dedicated, real-time system for the control of neuroprostheses. Software execution (written using the high-level programming languages LabVIEW and MATLAB) was divided into two phases: training and real-time control. During the training phase, the computer system collected input/output data by stimulating the muscles and measuring the muscle outputs in real-time, analysed the recorded data, generated a set of training data and trained an artificial neural network (ANN)-based controller. During real-time control, the computer system stimulated the muscles using stimulus pulsewidths predicted by the ANN controller in response to a sampled input from an external command source, to provide independent control of hand grasp and wrist posture. System timing was stable, reliable and capable of providing muscle stimulation at frequencies up to 24 Hz. To demonstrate the application of the test-bed, an ANN-based controller was implemented with three inputs and two independent channels of stimulation. The ANN controller's ability to control hand grasp and wrist angle independently was assessed by quantitative comparison of the outputs of the stimulated muscles with a set of desired grasp or wrist postures determined by the command signal. Controller performance results were mixed, but the platform provided the tools to implement and assess future controller designs.     

Critical events in intensive care unit

This prospective study was designed to have an insight into critical events occurring in the 13-bedded multidisciplinary intensive care unit (ICU) of our hospital and to report the critical events to evaluate the avoidable/iatrogenic problems so as to improve patient outcome and keep a self-check in the ICU. The errors reported were due to wrong mechanical or human performance. Repeated performance errors of the same kind pointed to the problem area, to which was paid proper attention in the required manner. Some malfunctioning equipments were abandoned and the need for adequate availability of staff was emphasized. Reporting of critical events was done keeping the patients'' and doctor''s identities anonymous through a proforma designed to report the event.     

Automatic modeling of pectus excavatum corrective prosthesis using artificial neural networks

Pectus excavatum is the most common deformity of the thorax. Pre-operative diagnosis usually includes Computed Tomography (CT) to successfully employ a thoracic prosthesis for anterior chest wall remodeling. Aiming at the elimination of radiation exposure, this paper presents a novel methodology for the replacement of CT by a 3D laser scanner (radiation-free) for prosthesis modeling.The complete elimination of CT is based on an accurate determination of ribs position and prosthesis placement region through skin surface points. The developed solution resorts to a normalized and combined outcome of an artificial neural network (ANN) set. Each ANN model was trained with data vectors from 165 male patients and using soft tissue thicknesses (STT) comprising information from the skin and rib cage (automatically determined by image processing algorithms). Tests revealed that ribs position for prosthesis placement and modeling can be estimated with an average error of 5.0 ± 3.6 mm. One also showed that the ANN performance can be improved by introducing a manually determined initial STT value in the ANN normalization procedure (average error of 2.82 ± 0.76 mm). Such error range is well below current prosthesis manual modeling (approximately 11 mm), which can provide a valuable and radiation-free procedure for prosthesis personalization.     

Prediction of protein functional domains from sequences using artificial neural networks

An artificial neural network (ANN) solution is described for the recognition of domains in protein sequences. A query sequence is first compared to a reference database of domain sequences by use of and the output data, encoded in the form of six parameters, are forwarded to feed-forward artificial neural networks with six input and six hidden units with sigmoidal transfer function. The recognition is based on the distribution of scores precomputed for the known domain groups in a database versus database comparison. Applications to the prediction of function are discussed.     

Antibiotic sensitivity of enterobacteria in intensive care units

Antibiotic therapy of intensive care patients is usually undocumented. The treatment is chosen according to epidemiologic and susceptibility data from microbiological laboratories. The aim of our study is to determine antibiotic susceptibility of enterobacteria isolated from intensive care patients during a five-month multicenter study in 18 French hospitals. Numerous (n = 1,113) strains were studied: 447 enterobacteria isolated from urine (n = 229), blood cultures (n = 106), respiratory tract specimens (n = 72), peritoneal fluids (n = 22), pus (n = 15) and catheters (n = 2). MICs of group 2 and group 3 enterobacteria were determined using the dilution agar method and were interpreted according to the CASFM (Comité de l'antibiogramme de la société fran?aise de microbiology) recommendations. Group 1 enterobacteria were most frequently isolated (67%). Only one Escherichia coli strain produced ESBL (0.3%). Among group 2 enterobacteria, one Citrobacter koseri strain produced ESBL. We did not isolate Klebsiella pneumoniae ESBL. Isolation of group 3 enterobacteria was frequent (24%). Thirty-five percent of group 3 enterobacteria were resistant to cefotaxime, 26% to ceftazidime and 16% to cefepime and cefpirome. Fourteen strains of this group produced ESBL: 13 Enterobacter aerogenes and one E. amnigenus.     

Strategies for selecting antibiotics in intensive care units

Crowding of severely ill patients in intensive care units has led worldwide to important increases in nosocomial (ICU-related) infections. Moreover, the nature of these hospital-acquired infections is shifting towards Gram-positive microorganisms, yeast and Gram-negative rods, possessing important resistance genes (e.g. extended spectrum beta-lactamases and inducible Enterobacteriaceae). Ceftazidime and aztreonam are loosing their activity against the Gram-negative microorganisms. The fourth generation cephalosporins have an intrinsic high activity against the inducible Enterobacteriaceae. On our Hematology and Intensive Care units, the introduction of cefepime for nosocomial infections led to a remarkable drop in the number of Enterobacter isolates combined with important decreases in Enterobacter resistance towards several antibiotics.     

Candida glabrata fungaemia in intensive care units.

Candidaemia is increasingly important in intensive care units (ICUs). Compared with Candida albicans fungaemia, the impact of C. glabrata fungaemia on ICU patients is not well-known. The aim of this study was to investigate the clinical features, the antifungal susceptibility and the treatment outcomes of C. glabrata fungaemia in ICU patients. The medical records of ICU patients with candidaemia between 2000 and 2005 were reviewed retrospectively, and antifungal susceptibility testing was performed for isolates of C. glabrata. Among 147 episodes of candidaemia occurring in adult ICUs, C. glabrata was the second most common species and accounted for 45 (30%) episodes of candidaemia. The incidence of C. glabrata fungaemia was 1.3/1000 ICU admissions. Fluconazole resistance was found in 11% of C. glabrata isolates. The 30-day all-cause mortality rate was 58%. Therapeutic regimens containing amphotericin B were associated with better outcome. Despite higher fluconazole resistance, C. glabrata candidaemia was not associated with greater mortality than non-glabrata candidaemia in the ICU setting.     

Modes of death in neonatal intensive care units

With the ever-increasing availability of aggressive medical treatment and technical support, neonatologists are offered an increasing ability to prolong life. While "end-of-life" decisions within NICUs have been studied internationally, there is limited data available for Ireland. Through the auspices of the Irish Faculty of Paediatrics 2002 Neonatal Mortality Ouestionnaire, decisions made around the time of death in Irish Neonatal Intensive Care Units were examined. The overall response rate to the questionnaire was 96% (n=25). One hundred and eighty seven deaths were reported for 2002. Information pertaining to the mode of death was available in 53% of cases. Seventy seven percent of those paediatricians who answered this question, reported either withdrawing or withholding treatment in babies thought to have a hopeless outcome, with the greatest proportion of these deaths occurring in premature infants (n=30) and babies with congenital defects (n=40).     

Parent needs assessment instruments in neonatal intensive care units: Implications for parent education interventions

ObjectivesPsychosocial and educational interventions based on standardized needs assessment can help alleviate distress among parents of premature infants. This study aims to (1) provide an overview of standardized instruments used to assess parental needs in neonatal intensive care units (NICUs) and (2) discuss their potential to facilitate the provision of appropriate support to parents of premature babies.MethodsA systematic literature review was conducted. PubMed, CiNAHL, PsychARTICLES, PsychINFO, and Medline were searched for studies reporting on the use of validated parental needs assessment instruments in the NICU.ResultsFollowing the analysis of 33 publications, 6 instruments designed to assess the needs of premature infants’ parents were identified. Based on their good psychometric properties and practicality, the NICU Family Needs Inventory, the Critical Care Maternal Needs Inventory, and the Nurse Parent Support Tool were considered particularly relevant for use in clinical and research settings.ConclusionsValidated parent needs assessment instruments are available for use in the NICU setting. Further research evaluating the benefits and usability of standardized parental needs assessment in the NICU is needed.Practice implicationsValidated needs assessment instruments should be consistently used to facilitate the development of targeted psychosocial and educational interventions for parents in the NICU.     

WeAidU-a decision support system for myocardial perfusion images using artificial neural networks

This paper presents a computer-based decision support system for automated interpretation of diagnostic heart images (called WeAidU), which is made available via the Internet. The system is based on image processing techniques, artificial neural networks (ANNs) and large well-validated medical databases. We present results using artificial neural networks, and compare with two other classification methods, on a retrospective data set containing 1320 images from the clinical routine. The performance of the artificial neural networks detecting infarction and ischemia in different parts of the heart, measured as areas under the receiver operating characteristic curves, is in the range 0.83-0.96. These results indicate a high potential for the tool as a clinical decision support system.