Artificial neural networks

First Principles Study of Electronic and Optical Properties of Al-P Co-Doped ZnO in the Presence of Zn Vacancies

The BP neural network optimized by the Adam algorithm was used to predict the defect formation energy of Al-P co-doped ZnO systems with different concentrations of P replacing O under the presence of different concentrations of VZn. It was found that the easily formed AlZnPo-1VZn, AlZnPO-2VZn, and AlZn2PO-1VZn systems. The first principles of density function were used to study the geometric, electronic, and optical properties of each system. The simulation results show that the bandgap values of the three systems have decreased relative to the intrinsic ZnO, among which AlZnPO-1VZn and AlZnPO-2VZn is still a p-type conductive system, AlZnPO-2VZn has the highest conductivity. From the analysis of reflectivity, absorption rate, and light transmittance, AlZn2PO-1VZn has the most relatively excellent optical properties, followed by AlznPo-2VZn.

Improving ADMET Prediction Accuracy for Candidate Drugs: Factors to Consider in QSPR Modeling Approaches

Quantitative Structure-Property Relationship (QSPR) employs mathematical and statistical methods to reveal quantitative correlations between the pharmacokinetics of compounds and their molecular structures, as well as their physical and chemical properties. QSPR models have been widely applied in the prediction of drug absorption, distribution, metabolism, excretion, and toxicity (ADMET). However, the accuracy of QSPR models for predicting drug ADMET properties still needs improvement. Therefore, this paper comprehensively reviews the tools employed in various stages of QSPR predictions for drug ADMET. It summarizes commonly used approaches to building QSPR models, systematically analyzing the advantages and limitations of each modeling method to ensure their judicious application. We provide an overview of recent advancements in the application of QSPR models for predicting drug ADMET properties. Furthermore, this review explores the inherent challenges in QSPR modeling while also proposing a range of considerations aimed at enhancing model prediction accuracy. The objective is to enhance the predictive capabilities of QSPR models in the field of drug development and provide valuable reference and guidance for researchers in this domain.

Prediction of diagnosis results of rheumatoid arthritis patients based on autoantibodies and cost-sensitive neural network

OBJECTIVES

To analyze and evaluate the effectiveness of the detection of single autoantibody and combined autoantibodies in patients with rheumatoid arthritis (RA) and related autoimmune diseases and establish a machine learning model to predict the disease of RA.

METHODS

A total of 309 patients with joint pain as the first symptom were retrieved from the database. The effectiveness of single and combined antibodies tests was analyzed and evaluated in patients with RA, a cost-sensitive neural network (CSNN) model was used to integrate multiple autoantibodies and patient symptoms to predict the diagnosis of RA, and the ROC curve was used to analyze the diagnosis performance and calculate the optimal cutoff value.

RESULTS

There are differences in the seropositive rate of autoimmune diseases, the sensitivity and specificity of single or multiple autoantibody tests were insufficient, and anti-CCP performed best in RA diagnosis and had high diagnostic value. The cost-sensitive neural network prediction model had a sensitivity of up to 0.90 and specificity of up to 0.86, which was better than a single antibody and combined multiple antibody detection.

CONCLUSION

In-depth analysis of autoantibodies and reliable early diagnosis based on the neural network could guide specialized physicians to develop different treatment plans to prevent deterioration and enable early treatment with antirheumatic drugs for remission. Key Points • There are differences in the seropositive rate of autoimmune diseases. • This is the first study to use a cost-sensitive neural network model to diagnose RA disease in patients. • The diagnosis effect of the cost-sensitive neural network model is better than a single antibody and combined multiple antibody detection.

A Review of the Modeling of Adsorption of Organic and Inorganic Pollutants from Water Using Artificial Neural Networks

The application of artificial neural networks on adsorption modeling has significantly increased during the last decades. These artificial intelligence models have been utilized to correlate and predict kinetics, isotherms, and breakthrough curves of a wide spectrum of adsorbents and adsorbates in the context of water purification. Artificial neural networks allow to overcome some drawbacks of traditional adsorption models especially in terms of providing better predictions at different operating conditions. However, these surrogate models have been applied mainly in adsorption systems with only one pollutant thus indicating the importance of extending their application for the prediction and simulation of adsorption systems with several adsorbates (i.e., multicomponent adsorption). This review analyzes and describes the data modeling of adsorption of organic and inorganic pollutants from water with artificial neural networks. The main developments and contributions on this topic have been discussed considering the results of a detailed search and interpretation of more than 250 papers published on Web of Science ® database. Therefore, a general overview of the training methods, input and output data, and numerical performance of artificial neural networks and related models utilized for adsorption data simulation is provided in this document. Some remarks for the reliable application and implementation of artificial neural networks on the adsorption modeling are also discussed. Overall, the studies on adsorption modeling with artificial neural networks have focused mainly on the analysis of batch processes (87%) in comparison to dynamic systems (13%) like packed bed columns. Multicomponent adsorption has not been extensively analyzed with artificial neural network models where this literature review indicated that 87% of references published on this topic covered adsorption systems with only one adsorbate. Results reported in several studies indicated that this artificial intelligence tool has a significant potential to develop reliable models for multicomponent adsorption systems where antagonistic, synergistic, and noninteraction adsorption behaviors can occur simultaneously. The development of reliable artificial neural networks for the modeling of multicomponent adsorption in batch and dynamic systems is fundamental to improve the process engineering in water treatment and purification.

THE IMPACT OF SOCIAL CAPITAL ON OPEN INNOVATION: THE TUNISIAN SMEs CASE

Given that few studies have focused on the relationship between social capital, open innovation and firm performance, particularly in small- and medium-sized enterprises (SMEs), we decided to extend this line of research. Thus, we tried to focus on social capital as a key element of informal cultural norms and to study its impact on open innovation in these firms and consequently on their performance. Through a review of the literature, we were able to develop the existing relationships between these three concepts and design our theoretical model. In accordance with a positivist posture, we adopted a hypothetical-deductive approach and a quantitative empirical validation method. In order to test our conceptual model and to analyse the results of the study carried out with 67 Tunisian SMEs of the mechanical and metallurgical industries, we used the Structural Equations Method and the Neural Networks Method. The results of this research indicate that social capital significantly influences open innovation which in turn positively influences the performance of SMEs. Specifically, this study contributes to resource theory and social capital theory by first confirming that intangible resources and capabilities, associated with social capital and open innovation are valuable to SME performance. In addition to the theoretical implications, the results of this study also provide important guidance to practitioners regarding strategic direction, in that SMEs should deliberately use their limited administrative capabilities and intangible resources to support business performance. As SMEs play an important role in building, maintaining, and evolving innovation ecosystems and contribute significantly to a country’s economic growth, our research can offer significant policy implications for Tunisian SMEs to improve their innovation performance through OI strategies.

WSES project on decision support systems based on artificial neural networks in emergency surgery

The article is a scoping review of the literature on the use of decision support systems based on artificial neural networks in emergency surgery. The authors present modern literature data on the effectiveness of artificial neural networks for predicting, diagnosing and treating abdominal emergency conditions: acute appendicitis, acute pancreatitis, acute cholecystitis, perforated gastric or duodenal ulcer, acute intestinal obstruction, and strangulated hernia. The intelligent systems developed at present allow a surgeon in an emergency setting, not only to check his own diagnostic and prognostic assumptions, but also to use artificial intelligence in complex urgent clinical cases. The authors summarize the main limitations for the implementation of artificial neural networks in surgery and medicine in general. These limitations are the lack of transparency in the decision-making process; insufficient quality educational medical data; lack of qualified personnel; high cost of projects; and the complexity of secure storage of medical information data. The development and implementation of decision support systems based on artificial neural networks is a promising direction for improving the forecasting, diagnosis and treatment of emergency surgical diseases and their complications.

Can machine learning improve mortality prediction following cardiac surgery?

OBJECTIVES

Interest in the clinical usefulness of machine learning for risk prediction has bloomed recently. Cardiac surgery patients are at high risk of complications and therefore presurgical risk assessment is of crucial relevance. We aimed to compare the performance of machine learning algorithms over traditional logistic regression (LR) model to predict in-hospital mortality following cardiac surgery.

METHODS

A single-centre data set of prospectively collected information from patients undergoing adult cardiac surgery from 1996 to 2017 was split into 70% training set and 30% testing set. Prediction models were developed using neural network, random forest, naive Bayes and retrained LR based on features included in the EuroSCORE. Discrimination was assessed using area under the receiver operating characteristic curve, and calibration analysis was undertaken using the calibration belt method. Model calibration drift was assessed by comparing Goodness of fit χ2 statistics observed in 2 equal bins from the testing sample ordered by procedure date.

RESULTS

A total of 28 761 cardiac procedures were performed during the study period. The in-hospital mortality rate was 2.7%. Retrained LR [area under the receiver operating characteristic curve 0.80; 95% confidence interval (CI) 0.77-0.83] and random forest model (0.80; 95% CI 0.76-0.83) showed the best discrimination. All models showed significant miscalibration. Retrained LR proved to have the weakest calibration drift.

CONCLUSIONS

Our findings do not support the hypothesis that machine learning methods provide advantage over LR model in predicting operative mortality after cardiac surgery.

Machine learning improves mortality risk prediction after cardiac surgery: Systematic review and meta-analysis

BACKGROUND

Interest in the usefulness of machine learning (ML) methods for outcomes prediction has continued to increase in recent years. However, the advantage of advanced ML model over traditional logistic regression (LR) remains controversial. We performed a systematic review and meta-analysis of studies comparing the discrimination accuracy between ML models versus LR in predicting operative mortality following cardiac surgery.

METHODS

The present systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analysis statement. Discrimination ability was assessed using the C-statistic. Pooled C-statistics and its 95% credibility interval for ML models and LR were obtained were obtained using a Bayesian framework. Pooled estimates for ML models and LR were compared to inform on difference between the 2 approaches.

RESULTS

We identified 459 published citations of which 15 studies met inclusion criteria and were used for the quantitative and qualitative analysis. When the best ML model from individual study was used, meta-analytic estimates showed that ML were associated with a significantly higher C-statistic (ML, 0.88; 95% credibility interval, 0.83-0.93 vs LR, 0.81; 95% credibility interval, 0.77-0.85; P = .03). When individual ML algorithms were instead selected, we found a nonsignificant trend toward better prediction with each of ML algorithms. We found no evidence of publication bias (P = .70).

CONCLUSIONS

The present findings suggest that when compared with LR, ML models provide better discrimination in mortality prediction after cardiac surgery. However, the magnitude and clinical influence of such an improvement remains uncertain.

Forecasting the Structure of Energy Production from Renewable Energy Sources and Biofuels in Poland

The world’s economic development depends on access to cheap energy sources. So far, energy has been obtained mainly from conventional sources like coal, gas and oil. Negative climate changes related to the high emissions of the economy based on the combustion of hydrocarbons and the growing public awareness have made it necessary to look for new ecological energy sources. This condition can be met by renewable energy sources. Both social pressure and international activities force changes in the structure of sources from which energy is produced. This also applies to the European Union countries, including Poland. There are no scientific studies in the area of forecasting energy production from renewable energy sources for Poland. Therefore, it is reasonable to investigate this subject since such a forecast can have a significant impact on investment decisions in the energy sector. At the same time, it must be as reliable as possible. That is why a modern method was used for this purpose, which undoubtedly involves artificial neural networks. The following article presents the results of the analysis of energy production from renewable energy sources in Poland and the forecasts for this production until 2025. Artificial neural networks were used to make the forecast. The analysis covered eight main sources from which this energy is produced in Poland. Based on the production volume since 1990, predicted volumes of renewable energy sources until 2025 were determined. These forecasts were prepared for all studied renewable energy sources. Renewable energy production plans and their share in total energy consumption in Poland were also examined and included in climate plans. The research was carried out using artificial neural networks. The results should be an important source of information on the effects of implementing climate policies in Poland. They should also be utilized to develop action plans to achieve the objectives of the European Green Deal strategy.

Machine Learning techniques in breast cancer prognosis prediction: A primary evaluation

More than 750 000 women in Italy are surviving a diagnosis of breast cancer. A large body of literature tells us which characteristics impact the most on their prognosis. However, the prediction of each disease course and then the establishment of a therapeutic plan and follow‐up tailored to the patient is still very complicated. In order to address this issue, a multidisciplinary approach has become widely accepted, while the Multigene Signature Panels and the Nottingham Prognostic Index are still discussed options. The current technological resources permit to gather many data for each patient. Machine Learning (ML) allows us to draw on these data, to discover their mutual relations and to esteem the prognosis for the new instances. This study provides a primary evaluation of the application of ML to predict breast cancer prognosis. We analyzed 1021 patients who underwent surgery for breast cancer in our Institute and we included 610 of them. Three outcomes were chosen: cancer recurrence (both loco‐regional and systemic) and death from the disease within 32 months. We developed two types of ML models for every outcome (Artificial Neural Network and Support Vector Machine). Each ML algorithm was tested in accuracy (=95.29%‐96.86%), sensitivity (=0.35‐0.64), specificity (=0.97‐0.99), and AUC (=0.804‐0.916). These models might become an additional resource to evaluate the prognosis of breast cancer patients in our daily clinical practice. Before that, we should increase their sensitivity, according to literature, by considering a wider population sample with a longer period of follow‐up. However, specificity, accuracy, minimal additional costs, and reproducibility are already encouraging.

Automatic Lung Nodule Detection Combined With Gaze Information Improves Radiologists' Screening Performance

Early diagnosis of lung cancer via computed tomography can significantly reduce the morbidity and mortality rates associated with the pathology. However, searching lung nodules is a high complexity task, which affects the success of screening programs. Whilst computer-aided detection systems can be used as second observers, they may bias radiologists and introduce significant time overheads. With this in mind, this study assesses the potential of using gaze information for integrating automatic detection systems in the clinical practice. For that purpose, 4 radiologists were asked to annotate 20 scans from a public dataset while being monitored by an eye tracker device, and an automatic lung nodule detection system was developed. Our results show that radiologists follow a similar search routine and tend to have lower fixation periods in regions where finding errors occur. The overall detection sensitivity of the specialists was 0.67±0.07, whereas the system achieved 0.69. Combining the annotations of one radiologist with the automatic system significantly improves the detection performance to similar levels of two annotators. Filtering automatic detection candidates only for low fixation regions still significantly improves the detection sensitivity without increasing the number of false-positives.

Artificial Neural Networks for Prediction of Tuberculosis Disease

Background: The global burden of tuberculosis (TB) and antibiotic resistance is attracting the attention of researchers to develop some novel and rapid diagnostic tools. Although, the conventional methods like culture are considered as the gold standard, they are time consuming in diagnostic procedure, during which there are more chances in the transmission of disease. Further, the Xpert MTB/RIF assay offers a fast diagnostic facility within 2 h, but due to low sensitivity in some sample types may lead to more serious state of the disease. The role of computer technologies is now increasing in the diagnostic procedures. Here, in the current study we have applied the artificial neural network (ANN) that predicted the TB disease based on the TB suspect data. Methods: We developed an approach for prediction of TB, based on an ANN. The data was collected from the TB suspects, guardians or care takers along with samples, referred by TB units and health centers. All the samples were processed and cultured. Data was trained on 12,636 records of TB patients, collected during the years 2016 and 2017 from the provincial TB reference laboratory, Khyber Pakhtunkhwa, Pakistan. The training and test set of the suspect data were kept as 70 and 30%, respectively, followed by validation and normalization. The ANN takes the TB suspect's information such as gender, age, HIV-status, previous TB history, sample type, and signs and symptoms for TB prediction. Results: Based on TB patient data, ANN accurately predicted the Mycobacterium tuberculosis (MTB) positive or negative with an overall accuracy of >94%. Further, the accuracy of the test and validation were found to be >93%. This increased accuracy of ANN in the detection of TB suspected patients might be useful for early management of disease to adopt some control measures in further transmission and reduce the drug resistance burden. Conclusion: ANNs algorithms may play an effective role in the early diagnosis of TB disease that might be applied as a supportive tool. Modern computer technologies should be trained in diagnostics for rapid disease management. Delays in TB diagnosis and initiation treatment may allow the emergence of new cases by transmission, causing high drug resistance in countries with a high TB burden.

Cytotoxicity of chitosan/streptokinase nanoparticles as a function of size: An artificial neural networks study

Predicting the size and toxicity of chitosan/streptokinase nanoparticles at various values of processing parameters was the aim of this study. For the first time, a comprehensive model could be developed to determine the cytotoxicity of the nanoparticles as a function of their size. Then, artificial neural networks were used for identifying main factors influencing self-assembly prepared nanoparticles size and cytotoxicity. Three variables included polymer concentration; pH and stirring time were used for a modeling study. A second modeling was performed to evaluate the influence of particles' size on toxicity. Experimentally data modeled using ANNs was validated against unseen data. The response surfaces generated from the software demonstrated that chitosan concentration is the dominant factor with a direct effect on size. Results also showed that the most important factor in determining the particles' toxicity is size--smaller particles showed more toxic effects, regardless of the effect of other input parameters. From the Clinical Editor: The understanding of toxicity of nanoparticles is of prime importance. In this article, the authors generated a model to visualize the relationship between nanoparticle size and its cellular toxicity, using chitosan/streptokinase nanoparticles. The data generated here would help the design of future nanoparticles of appropriate sizes for the application in the clinical setting.

An Automated System for Skeletal Maturity Assessment by Extreme Learning Machines

Assessing skeletal age is a subjective and tedious examination process. Hence, automated assessment methods have been developed to replace manual evaluation in medical applications. In this study, a new fully automated method based on content-based image retrieval and using extreme learning machines (ELM) is designed and adapted to assess skeletal maturity. The main novelty of this approach is it overcomes the segmentation problem as suffered by existing systems. The estimation results of ELM models are compared with those of genetic programming (GP) and artificial neural networks (ANNs) models. The experimental results signify improvement in assessment accuracy over GP and ANN, while generalization capability is possible with the ELM approach. Moreover, the results are indicated that the ELM model developed can be used confidently in further work on formulating novel models of skeletal age assessment strategies. According to the experimental results, the new presented method has the capacity to learn many hundreds of times faster than traditional learning methods and it has sufficient overall performance in many aspects. It has conclusively been found that applying ELM is particularly promising as an alternative method for evaluating skeletal age.

The use of intelligent database systems in acute pancreatitis--a systematic review

INTRODUCTION

Acute pancreatitis (AP) is a complex disease with multiple aetiological factors, wide ranging severity, and multiple challenges to effective triage and management. Databases, data mining and machine learning algorithms (MLAs), including artificial neural networks (ANNs), may assist by storing and interpreting data from multiple sources, potentially improving clinical decision-making.

AIMS

1) Identify database technologies used to store AP data, 2) collate and categorise variables stored in AP databases, 3) identify the MLA technologies, including ANNs, used to analyse AP data, and 4) identify clinical and non-clinical benefits and obstacles in establishing a national or international AP database.

METHODS

Comprehensive systematic search of online reference databases. The predetermined inclusion criteria were all papers discussing 1) databases, 2) data mining or 3) MLAs, pertaining to AP, independently assessed by two reviewers with conflicts resolved by a third author.

RESULTS

Forty-three papers were included. Three data mining technologies and five ANN methodologies were reported in the literature. There were 187 collected variables identified. ANNs increase accuracy of severity prediction, one study showed ANNs had a sensitivity of 0.89 and specificity of 0.96 six hours after admission--compare APACHE II (cutoff score ≥8) with 0.80 and 0.85 respectively. Problems with databases were incomplete data, lack of clinical data, diagnostic reliability and missing clinical data.

CONCLUSION

This is the first systematic review examining the use of databases, MLAs and ANNs in the management of AP. The clinical benefits these technologies have over current systems and other advantages to adopting them are identified.

Development of a new CBR-based platform for human contamination emergency situations

In the case of a radiological emergency situation, involving accidental human exposure, it is necessary to establish as soon as possible a dosimetry evaluation. In most cases, this evaluation is based on numerical representations and models of the victims. Unfortunately, personalised and realistic human representations are often unavailable for the exposed subjects. Hence, existing models like the 'Reference Man' representative of the average male individual are used. However, the accuracy of the treatment depends on the similarity of the phantom to the victim. The EquiVox platform (Research of Equivalent Voxel phantom) developed in this work uses the case-based reasoning principles to retrieve, from a set of existing phantoms, the most adapted one to represent the victim. This paper introduces the EquiVox platform and gives the example of in vivo lung monitoring optimisation to prove its efficiency in choosing the right model. It also presents the artificial neural network tools being developed to adapt the model to the victim.

Novel solutions for an old disease: diagnosis of acute appendicitis with random forest, support vector machines, and artificial neural networks

BACKGROUND

Diagnosing acute appendicitis clinically is still difficult. We developed random forests, support vector machines, and artificial neural network models to diagnose acute appendicitis.

METHODS

Between January 2006 and December 2008, patients who had a consultation session with surgeons for suspected acute appendicitis were enrolled. Seventy-five percent of the data set was used to construct models including random forest, support vector machines, artificial neural networks, and logistic regression. Twenty-five percent of the data set was withheld to evaluate model performance. The area under the receiver operating characteristic curve (AUC) was used to evaluate performance, which was compared with that of the Alvarado score.

RESULTS

Data from a total of 180 patients were collected, 135 used for training and 45 for testing. The mean age of patients was 39.4 years (range, 16-85). Final diagnosis revealed 115 patients with and 65 without appendicitis. The AUC of random forest, support vector machines, artificial neural networks, logistic regression, and Alvarado was 0.98, 0.96, 0.91, 0.87, and 0.77, respectively. The sensitivity, specificity, positive, and negative predictive values of random forest were 94%, 100%, 100%, and 87%, respectively. Random forest performed better than artificial neural networks, logistic regression, and Alvarado.

CONCLUSION

We demonstrated that random forest can predict acute appendicitis with good accuracy and, deployed appropriately, can be an effective tool in clinical decision making.

[Evaluation of severity in acute pancreatitis]

Acute pancreatitis has a variable etiology and natural history, and some patients have severe complications with a significant risk of death. The prediction of severe disease should be achieved by careful ongoing clinical assessment coupled with the use of a multiple factor scoring system and imaging studies. Over the past 30 years several scoring systems have been developed to predict the severity of acute pancreatitis. However, there are no complete scoring index with high sensitivity and specificity till now. The interest in new biological markers and predictive models for identifying severe acute pancreatitis testifies to the continued clinical importance of early severity prediction. Among them, IL-6, IL-10, procalcitonin, and trypsinogen activation peptide are most likely to be used in clinical practice as predictors of severity. Even if contrast-enhanced CT has been considered the gold standard for diagnosing pancreatic necrosis, early scanning for the prediction of severity is limited because the full extent of pancreatic necrosis may not develop within the first 48 hour of presentation.

Prediction of conversion of laparoscopic cholecystectomy to open surgery with artificial neural networks

Background

The intent of this study was to predict conversion of laparoscopic cholecystectomy (LC) to open surgery employing artificial neural networks (ANN).

Methods

The retrospective data of 793 patients who underwent LC in a teaching university hospital from 1997 to 2004 was collected. We employed linear discrimination analysis and ANN models to examine the predictability of the conversion. The models were validated using prospective data of 100 patients who underwent LC at the same hospital.

Results

The overall conversion rate was 9%. Conversion correlated with experience of surgeons, emergency LC, previous abdominal surgery, fever, leukocytosis, elevated bilirubin and alkaline phosphatase levels, and ultrasonographic detection of common bile duct stones. In the validation group, discriminant analysis formula diagnosed the conversion in 5 cases out of 9 (sensitivity: 56%; specificity: 82%); the ANN model diagnosed 6 cases (sensitivity: 67%; specificity: 99%).

Conclusion

The conversion of LC to open surgery is effectively predictable based on the preoperative health characteristics of patients using ANN.

Risk assessment in acute pancreatitis

BACKGROUND

Acute pancreatitis has a variable natural history and in a proportion of patients is associated with severe complications and a significant risk of death. The various tools available for risk assessment in acute pancreatitis are reviewed.

METHODS

Relevant medical literature from PubMed, Ovid, Embase, Web of Science and The Cochrane Library websites to May 2008 was reviewed.

RESULTS AND CONCLUSION

Over the past 30 years several scoring systems have been developed to predict the severity of acute pancreatitis in the first 48-72 h. Biochemical and immunological markers, imaging modalities and novel predictive models may help identify patients at high risk of complications or death. Recently, there has been a recognition of the importance of the systemic inflammatory response syndrome and organ dysfunction.

Identification of severe acute pancreatitis using an artificial neural network

BACKGROUND

The aim of this study was to construct and validate an artificial neural network (ANN) model to identify severe acute pancreatitis (AP) and predict fatal outcome.

METHODS

All patients who presented with AP from January 2000 to September 2004 were reviewed. Presentation data on admission and at 48 hours were collected. Acute Physiology and Chronic Health Evaluation (APACHE) II and Glasgow severity (GS) score were calculated. A feed-forward ANN was created and trained to predict development of severe AP and mortality from AP; 25% of the data set was withheld from training and was used to evaluate the accuracy of the ANN. Accuracy of the ANN in predicting severity of AP was compared with APACHE II and GS scores.

RESULTS

A total of 664 patients with AP were identified of whom 181 (27.3%) fulfilled the clinical and radiologic criteria for severe pancreatitis and 42 patients died (6.3%). Median APACHE II score at 48 hours was 4 (range, 0 to 23). ANN was more accurate than APACHE II or GS scoring systems at predicting progression to a severe course (P < .05 and P < .01, respectively), predicting development of multiorgan dysfunction syndrome (P < .05 and P < .01) and at predicting death from AP (P < .05).

CONCLUSIONS

An ANN was able to predict progression to severe disease, development of organ failure and mortality from acute pancreatitis with considerable accuracy and outperformed other clinical risk scoring systems. Further studies are required to assess its utility in aiding management decisions in patients with AP.

Prediction of survival from carcinoma of oesophagus and oesophago-gastric junction following surgical resection using an artificial neural network

AIM

The aim of this study was to assess the ability of artificial neural network (ANN) in predicting survival in patients undergoing surgical resection for carcinoma of oesophagus and oesophago-gastric junction.

METHODS

From January 1995 to August 2004 patients who underwent surgery for oesophageal and gastric carcinoma were identified. Biographical data, body mass index and pathological minimal cancer dataset were used to design an ANN. Post-operative survival was assessed at 1 and 3 years. Sixty percent of data was used to train and validate the ANN and 40% was used to evaluate the accuracy of trained ANN in predicting survival. This was compared with Union Internacional Contra la Cancrum UICC TNM classification system.

RESULTS

Two hundred and sixteen patients underwent resectional surgery for oesophageal and OGJ carcinoma. The accuracy of the ANN in predicting survival at 1 and 3 years was 88% (sensitivity: 92.3%, specificity: 84.5%, DP = 2.3) and 91.5% (sensitivity of 94.61%, specificity: 88%, DP = 2.72), respectively. These figures were significantly better than 1- and 3-year survival predictions using the UICC TNM classification system 71.6% (sensitivity of 66.4%, specificity: 75.5%, and DP < 1) and 74.7% (sensitivity of 70.5%, specificity: 74.9%, DP < 1), respectively (P < 0.01) (P < 0.05).

CONCLUSION

ANNs are superior to the UICC TNM classification system in correlating with survival following resection of carcinoma of oesophagus and OG junction and can become valuable tools in the management of patients with oesophageal carcinoma.

Optimal prediction of mortality after abdominal aortic aneurysm repair with statistical models

OBJECTIVE

To identify the best method for the prediction of postoperative mortality in individual abdominal aortic aneurysm surgery (AAA) patients by comparing statistical modelling with artificial neural networks' (ANN) and clinicians' estimates.

METHODS

An observational multicenter study was conducted of prospectively collected postoperative Acute Physiology and Chronic Health Evaluation II data for a 9-year period from 24 intensive care units (ICU) in the Thames region of the United Kingdom. The study cohort consisted of 1205 elective and 546 emergency AAA patients. Four independent physiologic variables-age, acute physiology score, emergency operation, and chronic health evaluation-were used to develop multiple regression and ANN models to predict in-hospital mortality. The models were developed on 75% of the patient population and their validity tested on the remaining 25%. The results from these two models were compared with the observed outcome and clinicians' estimates by using measures of calibration, discrimination, and subgroup analysis.

RESULTS

Observed in-hospital mortality for elective surgery was 9.3% (95% confidence interval [CI], 7.7% to 11.1%) and for emergency surgery, 46.7% (95% CI, 42.5 to 51.0%). The ANN and the statistical models were both more accurate than the clinicians' predictions. Only the statistical model was internally valid, however, when applied to the validation set of observations, as evidenced by calibration (Hosmer-Lemeshow C statistic, 14.97; P = .060), discrimination properties (area under receiver operating characteristic curve, 0.869; 95% CI, 0.824 to 0.913), and subgroup analysis.

CONCLUSIONS

The prediction of in-hospital mortality in AAA patients by multiple regression is more accurate than clinicians' estimates or ANN modelling. Clinicians can use this statistical model as an objective adjunct to generate informed prognosis.

Prediction of the Exact Degree of Internal Carotid Artery Stenosis Using an Artificial Neural Network Based on Duplex Velocity Measurements

Duplex ultrasound criteria use a combination of velocity measurements to evaluate internal carotid artery (ICA) stenosis. These evaluations divide ICA stenosis into broad categories. The aim of this study was to design an artificial neural network (ANN) capable of predicting the exact degree of ICA stenosis based on duplex velocity measurements. Consecutive patients with significant carotid atherosclerosis underwent carotid duplex ultrasound and angiography. Peak systolic and end-diastolic velocities in the ICA and common carotid artery were measured. Multilayered perceptron ANNs were constructed and trained to predict the degree of ICA stenosis and band the degree of ICA stenosis into 10% intervals based on these measurements. The accuracy of the ANN models in predicting the degree of ICA stenosis and classifying the ICA stenosis was compared with the angiographic degree of ICA stenosis and duplex velocity criteria. A total of 208 carotid bifurcations were studied. ANNs were able to accurately predict the degree of angiographic ICA stenosis (R2 = 0.9374, p < 0.0001) and band the ICA stenosis into the predefined 10% intervals [sensitivity 97.3% (95% CI 90.7-99.3), specificity 97.7 % (95% CI 93.6-99.2), accuracy 97.5%]. The ANN model was more accurate [discriminant power (DP) = 4.11] in banding the degree of ICA stenosis than duplex velocity criteria (DP = 1.67) (p < 0.05). The accuracy of the ANN in correctly identifying >70% ICA stenosis was 98.4% [sensitivity 96.4% (95% CI 93.8-99.3), specificity 98.7% (95% CI 93.4-99.8), DP = 4.21]. ANNs can accurately predict the degree of ICA stenosis. With further refinement, ANNs could replace velocity criteria in the assessment of ICA stenosis using duplex ultrasound.

Bullet trajectory predicts the need for damage control: an artificial neural network model

BACKGROUND

Effective use of damage control in trauma hinges on an early decision to use it. Bullet trajectory has never been studied as a marker for damage control. We hypothesize that this decision can be predicted by an artificial neural network (ANN) model based on the bullet trajectory and the patient's blood pressure.

METHODS

A multilayer perceptron ANN predictive model was developed from a data set of 312 patients with single abdominal gunshot injuries. Input variables were the bullet path, trajectory patterns, and admission systolic pressure. The output variable was either a damage control laparotomy or intraoperative death. The best performing ANN was implemented on prospectively collected data from 34 patients.

RESULTS

The model achieved a correct classification rate of 0.96 and area under the receiver operating characteristic curve of 0.94. External validation showed the model to have a sensitivity of 88% and specificity of 96%. Model implementation on the prospectively collected data had a correct classification rate of 0.91. Sensitivity analysis showed that systolic pressure, bullet path across the midline, and trajectory involving the right upper quadrant were the three most important input variables.

CONCLUSION

Bullet trajectory is an important, hitherto unrecognized, factor that should be incorporated into the decision to use damage control.

A review of evidence of health benefit from artificial neural networks in medical intervention

The purpose of this review is to assess the evidence of healthcare benefits involving the application of artificial neural networks to the clinical functions of diagnosis, prognosis and survival analysis, in the medical domains of oncology, critical care and cardiovascular medicine. The primary source of publications is PUBMED listings under Randomised Controlled Trials and Clinical Trials. The rĵle of neural networks is introduced within the context of advances in medical decision support arising from parallel developments in statistics and artificial intelligence. This is followed by a survey of published Randomised Controlled Trials and Clinical Trials, leading to recommendations for good practice in the design and evaluation of neural networks for use in medical intervention.