Understanding artificial neural networks and exploring their potential applications for the practicing urologist

Opioid Nonadherence Risk Prediction of Patients with Cancer-Related Pain Based on Five Machine Learning Algorithms

Objectives

Opioid nonadherence represents a significant barrier to cancer pain treatment efficacy. However, there is currently no effective prediction method for opioid adherence in patients with cancer pain. We aimed to develop and validate a machine learning (ML) model and evaluate its feasibility to predict opioid nonadherence in patients with cancer pain.

Methods

This was a secondary analysis from a cross-sectional study that included 1195 patients from March 1, 2018, to October 31, 2019. Five ML algorithms, such as logistic regression (LR), random forest, eXtreme Gradient Boosting, multilayer perceptron, and support vector machine, were used to predict opioid nonadherence in patients with cancer pain using 43 demographic and clinical factors as predictors. The predictive effects of the models were compared by the area under the receiver operating characteristic curve (AUC_ROC), accuracy, precision, sensitivity, specificity, and F1 scores. The value of the best model for clinical application was assessed using decision curve analysis (DCA).

Results

The best model obtained in this study, the LR model, had an AUC_ROC of 0.82, accuracy of 0.82, and specificity of 0.71. The DCA showed that clinical interventions for patients at high risk of opioid nonadherence based on the LR model can benefit patients. The strongest predictors for adherence were, in order of importance, beliefs about medicines questionnaire (BMQ)-harm, time since the start of opioid, and BMQ-necessity. Discussion. ML algorithms can be used as an effective means of predicting adherence to opioids in patients with cancer pain, which allows for proactive clinical intervention to optimize cancer pain management. This trial is registered with ChiCTR2000033576.

Artificial Intelligence in Urology: Current Status and Future Perspectives

Surgical fields, especially urology, have shifted increasingly toward the use of artificial intelligence (AI). Advancements in AI have created massive improvements in diagnostics, outcome predictions, and robotic surgery. For robotic surgery to progress from assisting surgeons to eventually reaching autonomous procedures, there must be advancements in machine learning, natural language processing, and computer vision. Moreover, barriers such as data availability, interpretability of autonomous decision-making, Internet connection and security, and ethical concerns must be overcome.

Developing an Efficient Cancer Detection and Prediction Tool Using Convolution Neural Network Integrated with Neural Pattern Recognition

The application of computational approaches in medical science for diagnosis is made possible by the development in technical advancements connected to computer and biological sciences. The current cancer diagnosis system is becoming outmoded due to the new and rapid growth in cancer cases, and new, effective, and efficient methodologies are now required. Accurate cancer-type prediction is essential for cancer diagnosis and treatment. Understanding, diagnosing, and identifying the various types of cancer can be greatly aided by knowledge of the cancer genes. The Convolution Neural Network (CNN) and neural pattern recognition (NPR) approaches are used in this study paper to detect and predict the type of cancer. Different Convolution Neural Networks (CNNs) have been proposed by various researchers up to this point. Each model concentrated on a certain set of parameters to simulate the expression of genes. We have developed a novel CNN-NPR architecture that predicts cancer type while accounting for the tissue of origin using high-dimensional gene expression inputs. The 5000-person sample of the 1-D CNN integrated with NPR is trained and tested on the gene profile, mapping with various cancer kinds. The proposed model's accuracy of 94% suggests that the suggested combination may be useful for long-term cancer diagnosis and detection. Fewer parameters are required for the suggested model to be efficiently trained before prediction.

Classification of intelligence quotient via brainwave sub-band power ratio features and artificial neural network

This paper elaborates on the novel intelligence assessment method using the brainwave sub-band power ratio features. The study focuses only on the left hemisphere brainwave in its relaxed state. Distinct intelligence quotient groups have been established earlier from the score of the Raven Progressive Matrices. Sub-band power ratios are calculated from energy spectral density of theta, alpha and beta frequency bands. Synthetic data have been generated to increase dataset from 50 to 120. The features are used as input to the artificial neural network. Subsequently, the brain behaviour model has been developed using an artificial neural network that is trained with optimized learning rate, momentum constant and hidden nodes. Findings indicate that the distinct intelligence quotient groups can be classified from the brainwave sub-band power ratios with 100% training and 88.89% testing accuracies.

Early diagnosis of complex diseases by molecular biomarkers, network biomarkers, and dynamical network biomarkers

Many studies have been carried out for early diagnosis of complex diseases by finding accurate and robust biomarkers specific to respective diseases. In particular, recent rapid advance of high-throughput technologies provides unprecedented rich information to characterize various disease genotypes and phenotypes in a global and also dynamical manner, which significantly accelerates the study of biomarkers from both theoretical and clinical perspectives. Traditionally, molecular biomarkers that distinguish disease samples from normal samples are widely adopted in clinical practices due to their ease of data measurement. However, many of them suffer from low coverage and high false-positive rates or high false-negative rates, which seriously limit their further clinical applications. To overcome those difficulties, network biomarkers (or module biomarkers) attract much attention and also achieve better performance because a network (or subnetwork) is considered to be a more robust form to characterize diseases than individual molecules. But, both molecular biomarkers and network biomarkers mainly distinguish disease samples from normal samples, and they generally cannot ensure to identify predisease samples due to their static nature, thereby lacking ability to early diagnosis. Based on nonlinear dynamical theory and complex network theory, a new concept of dynamical network biomarkers (DNBs, or a dynamical network of biomarkers) has been developed, which is different from traditional static approaches, and the DNB is able to distinguish a predisease state from normal and disease states by even a small number of samples, and therefore has great potential to achieve "real" early diagnosis of complex diseases. In this paper, we comprehensively review the recent advances and developments on molecular biomarkers, network biomarkers, and DNBs in particular, focusing on the biomarkers for early diagnosis of complex diseases considering a small number of samples and high-throughput data (or big data). Detailed comparisons of various types of biomarkers as well as their applications are also discussed.

[How to biopsy only men with high grade prostate cancer]

INTRODUCTION

Fuzzy logic and Artificial Neural Networks (ANN) are complementary technologies that together generate neuro-fuzzy system. The aim of our study is to compare 2 models for predicting the presence of high-grade prostate cancer (Gleason score 7 or more).

METHODS

We evaluated data from 1000 men with PSA less than 50 ng/mL, who underwent prostate biopsy. A prostate cancer was found in 313 (31%), and in 172 (17.2%) we detected high-grade prostate cancer. With those data, we developed 2 Co-Active Neuro-Fuzzy Inference Systems to predict the presence of high-grade prostate cancer. The first model had four input neurons (PSA, free PSA percentage [%freePSA], PSA density, and age) and the second model had three input neurons (PSA, %freePSA, and age).

RESULTS

The model with four input neurons (PSA, %freePSA, PSA density, and age) showed better performances than the one with three input neurons (PSA, %freePSA, and age). In fact the average testing error was 0.42 for the model with four input neurons and 0.44 for the other model.

CONCLUSIONS

The addition of PSA density to the model has allowed to obtain better results for the diagnosis of high grade prostate cancer.

Hyperspectral imaging based method for fast characterization of kidney stone types

The formation of kidney stones is a common and highly studied disease, which causes intense pain and presents a high recidivism. In order to find the causes of this problem, the characterization of the main compounds is of great importance. In this sense, the analysis of the composition and structure of the stone can give key information about the urine parameters during the crystal growth. But the usual methods employed are slow, analyst dependent and the information obtained is poor. In the present work, the near infrared (NIR)-hyperspectral imaging technique was used for the analysis of 215 samples of kidney stones, including the main types usually found and their mixtures. The NIR reflectance spectra of the analyzed stones showed significant differences that were used for their classification. To do so, a method was created by the use of artificial neural networks, which showed a probability higher than 90% for right classification of the stones. The promising results, robust methodology, and the fast analytical process, without the need of an expert assistance, lead to an easy implementation at the clinical laboratories, offering the urologist a rapid diagnosis that shall contribute to minimize urolithiasis recidivism.

CLASSIFICATION OF BINARY SPATIAL TEXTURES USING STOCHASTIC GEOMETRY, NONLINEAR DETERMINISTIC ANALYSIS AND ARTIFICIAL NEURAL NETWORKS

Stereology and stochastic geometry can be used as auxiliary tools for diagnostic purposes in tumour pathology. The role of first-order parameters and stochastic–geometric functions for the classification of the texture of biological tissues has been investigated recently. The volume fraction and surface area per unit volume, the pair correlation function and the centred quadratic contact density function of epithelium were estimated in three case series of benign and malignant lesions of glandular tissues. This approach was further extended by applying the Laslett test, i.e. a point process statistic computed after transformation of the convex tangent points of sectioned random sets from planar images. This method has not yet been applied to histological images so far. Also the nonlinear deterministic approach to tissue texture was applied by estimating the correlation dimension as a function of embedding dimension. We used the stochastic–geometric functions, the first-order parameters and the correlation dimensions for the classification of cases using various algorithms. Learning vector quantization was applied as neural paradigm. Applications included distinction between mastopathy and mammary cancer, between benign prostatic hyperplasia and prostatic cancer, and between chronic pancreatitis and pancreatic cancer. The same data sets were also classified with discriminant analysis and support vector machines. The stereological estimates provided high accuracy in the classification of individual cases. The question: which category of estimator is the most informative, cannot be answered globally, but must be explored empirically for each specific data set. The results obtained by the three algorithms were similar.

Non-invasive diagnosis of risk in dengue patients using bioelectrical impedance analysis and artificial neural network

This paper presents a new approach to diagnose and classify early risk in dengue patients using bioelectrical impedance analysis (BIA) and artificial neural network (ANN). A total of 223 healthy subjects and 207 hospitalized dengue patients were prospectively studied. The dengue risk severity criteria was determined and grouped based on three blood investigations, namely, platelet (PLT) count (less than or equal to 30,000 cells per mm(3)), hematocrit (HCT) (increase by more than or equal to 20%), and either aspartate aminotransferase (AST) level (raised by fivefold the normal upper limit) or alanine aminotransferase (ALT) level (raised by fivefold the normal upper limit). The dengue patients were classified according to their risk groups and the corresponding BIA parameters were subsequently obtained and quantified. Four parameters were used for training and testing the ANN which are day of fever, reactance, gender, and risk group's quantification. Day of fever was defined as the day of fever subsided, i.e., when the body temperature fell below 37.5°C. The blood investigation and the BIA data were taken for 5 days. The ANN was trained via the steepest descent back propagation with momentum algorithm using the log-sigmoid transfer function while the sum-squared error was used as the network's performance indicator. The best ANN architecture of 3-6-1 (3 inputs, 6 neurons in the hidden layer, and 1 output), learning rate of 0.1, momentum constant of 0.2, and iteration rate of 20,000 was pruned using a weight-eliminating method. Eliminating a weight of 0.05 enhances the dengue's prediction risk classification accuracy of 95.88% for high risk and 96.83% for low risk groups. As a result, the system is able to classify and diagnose the risk in the dengue patients with an overall prediction accuracy of 96.27%.

Developing computational model-based diagnostics to analyse clinical chemistry data

This article provides methodological and technical considerations to researchers starting to develop computational model-based diagnostics using clinical chemistry data. These models are of increasing importance, since novel metabolomics and proteomics measuring technologies are able to produce large amounts of data that are difficult to interpret at first sight, but have high diagnostic potential. Computational models aid interpretation and make the data accessible for clinical diagnosis. We discuss the issues that a modeller has to take into account during the design, construction and evaluation phases of model development. We use the example of Particle Profiler development, a model-based diagnostic tool for lipoprotein disorders, as a case study, to illustrate our considerations. The case study also offers techniques for efficient model formulation, model calculation, workflow structuring and quality control.

The value of an artificial neural network in the decision-making for prostate biopsies

PURPOSE

In majority of patients who are subjected to prostate biopsies, no prostate cancer (PCa) is found. It is important to prevent unnecessary biopsies since serious complications may occur. An artificial neural network (ANN) may be able to predict the risk of the presence of PCa.

METHODS

Included were all patients, who underwent transrectal ultrasound-guided prostate biopsies between June 2006 and June 2007 with a total PSA (tPSA) level between 2 and 20 microg/l. The patients were divided into two groups according to their tPSA level (2-10 microg/l and 10-20 microg/l). The ANN Prostataclass of the Universitätsklinikum Charité in Berlin was used. The predictions of the ANN were compared to the pathology results of the biopsies.

RESULTS

Overall 165 patients were included. PCa was diagnosed in 53 patients, whereas the ANN predicted "no risk" in 19 of these patients (36%). The ANN output receiver operator characteristic (ROC) plots for the range of tPSA 2-10 microg/l and tPSA 10-20 microg/l showed an area under the curve (AUC) of 63 and 88% for the initial biopsy group, versus 69 and 57%, respectively, for the repeat biopsy group.

CONCLUSIONS

The ANN resulted in a false negative rate of 36%, missing PCa in 19 patients. For use in an outpatient-clinical setting, this ANN is insufficient to predict the risk of presence of PCa reliably.

Analysis of meal patterns with the use of supervised data mining techniques--artificial neural networks and decision trees

BACKGROUND

At present, the analysis of dietary patterns is based on the intake of individual foods. This article demonstrates how a coding system at the meal level might be analyzed by using data mining techniques.

OBJECTIVE

The objective was to evaluate the usability of supervised data mining methods to predict an aspect of dietary quality based on dietary intake with a food-based coding system and a novel meal-based coding system.

DESIGN

Food consumption databases from the North-South Ireland Food Consumption Survey 1997-1999 were used. This was a randomized cross-sectional study of 7-d recorded food and nutrient intakes of a representative sample of 1379 Irish adults. Meal definitions were recorded by the respondent. A healthy eating index (HEI) score was developed. Artificial neural networks (ANNs) and decision trees were used to predict quintiles of the HEI based on combinations of foods consumed at breakfast and main meals.

RESULTS

This study applied both data mining techniques to the food and meal-based coding systems. The ANN had a slightly higher accuracy than did the decision tree in relation to its ability to predict HEI quintiles 1 and 5 based on the food coding system (78.7% compared with 76.9% and 71.9% compared with 70.1%, respectively). However, the decision tree had higher accuracies than did the ANN on the basis of the meal coding system (67.5% compared with 54.6% and 75.1% compared with 72.4%, respectively).

CONCLUSIONS

ANNs and decision trees were successfully used to predict an aspect of dietary quality. However, further exploration of the use of ANNs and decision trees in dietary pattern analysis is warranted.

Clinical proteomics: discovery of cancer biomarkers using mass spectrometry and bioinformatics approaches--a prostate cancer perspective

Prostate cancer (PCa) is an intractable disease, where diagnosis and clinical prediction of the disease course and response to treatment is compromised by the lack of objective and robust biomarker assays. In late stage metastatic disease, treatment options are limited, although it is recognized that some patients may benefit from immunotherapy and in particular vaccine therapy. However, research into biomarkers that correlate with the clinical outcome of immunotherapy has lagged behind vaccine development. Thus, proteomic tools are increasingly being utilized for the discovery of biomarkers which will allow us to make clinical decisions about patient treatment at an earlier stage and should aid in shortening the development time for vaccines. In this review we will summarize the various proteomic platforms used to investigate new biomarkers in PCa for better patient diagnosis, prognosis, patient stratification, treatment monitoring and clinical surrogate endpoints. We will discuss method limitations and highlight the key areas of research required for understanding the etiology of PCa.

Prognostic accuracy of an artificial neural network in patients undergoing radical cystectomy for bladder cancer: a comparison with logistic regression analysis

OBJECTIVE

To compare the prognostic performance of an artificial neural network (ANN) with that of standard logistic regression (LR), in patients undergoing radical cystectomy for bladder cancer.

PATIENTS AND METHODS

From February 1982 to February 1994, 369 evaluable patients with non-metastatic bladder cancer had pelvic lymph node dissection and radical cystectomy for either stage Ta-T1 (any grade) tumour not responding to intravesical therapy, with or with no carcinoma in situ, or stage T2-T4 tumour. LR analysis based on 12 variables was used to identify predictors of overall 5-year survival, and the ANN model was developed to predict the same outcome. The LR analysis, based on statistically significant predictors, and the ANN model were the compared for their accuracy in predicting survival.

RESULTS

The median age of the patients was 63 years, and overall 201 of them died. The tumour stage and nodal involvement (both P<0.001) were the only statistically independent predictors of overall 5-year survival on LR analysis. Based on these variables, LR had a sensitivity and specificity for predicting survival of 68.4% and 82.8%, respectively; corresponding values for the ANN were 62.7% and 86.1%. For LR and ANN, the positive predictive values were 78.6% and 76.2%, and the negative predictive values were 73.9% and 76.5%, respectively. The index of diagnostic accuracy was 75.9% for LR and 76.4% for ANN.

CONCLUSIONS

The ANN accurately predicted the survival of patients undergoing radical cystectomy for bladder cancer and had a prognostic performance comparable with that of LR. As ANNs are based on easy-to-use software that can identify nonlinear interactions between variables, they might become the preferred tool for predicting outcome.

PSA and new biomarkers within multivariate models to improve early detection of prostate cancer

This review gives an overview of the use of prostate-specific antigen (PSA) and percent free-PSA (%fPSA)-based artificial neural networks (ANNs) and logistic regression models (LR) to reduce unnecessary prostate biopsies. There is a clear advantage in including clinical data such as age, digital rectal examination and transrectal ultrasound (TRUS) variables like prostate volume and PSA density as additional factors to tPSA and %fPSA within ANNs and LR models. There is also positive impact of tPSA and fPSA assays on the outcome of ANNs. New markers provide additional value within ANNs but to prove their clinical usefulness further testing is necessary.

Prognosis in critical care

Prognostic risk prediction models have been employed in the intensive care unit (ICU) setting since the 1980s and provide health care providers with important information to help inform decisions related to treatment and prognosis, as well as to compare outcomes across institutions. Prognostic models for critical care are among the most widely utilized and tested predictive models in healthcare. In this article, we review and compare mortality prediction models, including the APACHE (1981), SAPS (1984), APACHE-II (1985), MPM (1987), APACHE-III (1991), SAPS-II (1993), and MPM-II (1993). We emphasize the importance of model calibration in this domain. In addition, we present a brief review of the statistical methodology, multiple logistic regression, which underlies most of the models currently used in critical care.

Neuro-fuzzy system for prostate cancer diagnosis

OBJECTIVES

To develop a neuro-fuzzy system to predict the presence of prostate cancer. Neuro-fuzzy systems harness the power of two paradigms: fuzzy logic and artificial neural networks. We compared the predictive accuracy of our neuro-fuzzy system with that obtained by total prostate-specific antigen (tPSA) and percent free PSA (%fPSA).

METHODS

The data from 1030 men (both outpatients and hospitalized patients) were used. All men had a tPSA level of less than 20 ng/mL. Of the 1030 men, 195 (18.9%) had prostate cancer. A neuro-fuzzy system was developed using the coactive neuro-fuzzy inference system model.

RESULTS

The mean area under the receiver operating characteristic curve for the neuro-fuzzy system output was 0.799 +/- 0.029 (95% confidence interval 0.760 to 0.835), for tPSA, it was 0.724 +/- 0.032 (95% confidence interval 0.681 to 0.765), and for %fPSA, 0.766 +/- 0.024 (95% confidence interval 0.725 to 0.804). Furthermore, pairwise comparison of the area under the curves evidenced differences among %fPSA, tPSA, and neuro-fuzzy system's output (tPSA versus neuro-fuzzy system's output, P = 0.008; %fPSA versus neuro-fuzzy system's output, P = 0.032). The comparison at 95% sensitivity showed that the neuro-fuzzy system had the best specificity (31.9%).

CONCLUSIONS

This study presented a neuro-fuzzy system based on both serum data (tPSA and %fPSA) and clinical data (age) to enhance the performance of tPSA to discriminate prostate cancer. The predictive accuracy of the neuro-fuzzy system was superior to that of tPSA and %fPSA.

Role of transrectal ultrasonography in the prediction of prostate cancer: artificial neural network analysis

OBJECTIVE

The purpose of this study was to evaluate the diagnostic performance of an artificial neural network (ANN) model with and without transrectal ultrasonographic (TRUS) data.

METHODS

Six hundred eighty-four consecutive patients who had undergone TRUS-guided prostate biopsy from May 2003 to January 2005 were enrolled. We constructed 2 ANN models. One (ANN_1) incorporated patient age, digital rectal examination findings, prostate-specific antigen (PSA) level, PSA density, transitional zone volume, and PSA density in the transitional zone as input data, whereas the other (ANN_2) was constructed with the above and TRUS findings as input data. The performances of these 2 ANN models according to PSA levels (group A, 0-4 ng/mL; group B, 4-10 ng/mL; and group C, >10 ng/mL) were evaluated using receiver operating characteristic analysis.

RESULTS

Of the 684 patients who underwent prostate biopsy, 214 (31.3%) were confirmed to have prostate cancer; of 137 patients with positive digital rectal examination results, 60 (43.8%) were confirmed to have prostate cancer; and of 131 patients with positive TRUS findings, 93 (71%) were confirmed to have prostate cancer. In groups A, B, and C, the AUCs for ANN_1 were 0.738, 0.753, and 0.774, respectively; the AUCs for ANN_2 were 0.859, 0.797, and 0.894. In all groups, ANN_2 showed better accuracy than ANN_1 (P < .05).

CONCLUSIONS

According to receiver operating characteristic analysis, ANN with TRUS findings was found to be more accurate than ANN without. We conclude that TRUS findings should be included as an input data component in ANN models used to diagnose prostate cancer.

Computer technology in detection and staging of prostate carcinoma: A review

After two decades of increasing interest and research activity, computer-assisted diagnostic approaches are reaching the stage where more routine deployment in clinical practice is becoming a possibility [Kruppinski, E.A., 2004. Computer-aided detection in clinical environment: Benefits and challenges for radiologists. Radiology 231, 7-9]. This is particularly the case in the analysis of mammographic images [Helvie, M.A., Hadjiiski, L., Makariou, E., Chan, H.P., Petrick, N., Sahiner, B., Lo, S.C., Freedman, M., Adler, D., Bailey, J., Blane, C., Hoff, D., Hunt, K., Joynt, L., Klein, K., Paramagul, C., Patterson, S.K., Roubidoux, M.A., 2004. Sensitivity of noncommercial computer-aided detection system for mammographic breast cancer detection: pilot clinical trial. Radiology 231, 208-214] and in the detection of pulmonary nodules [Reeves, A.P., Kostis, W.J., 2000. Computer-aided diagnosis for lung cancer. Radiol. Clin. North Am. 38, 497-509]. However, similar approaches can be applied more widely with the promise of increasing clinical utility in other areas. We review how computer-aided approaches may be applied in the diagnosis and staging of prostatic cancer. The current status of computer technology is reviewed, covering artificial neural networks for detection and staging, computerised biopsy simulation and computer-assisted analysis of ultrasound and magnetic resonance images.

New approaches to identification of bacterial pathogens by surface enhanced laser desorption/ionization time of flight mass spectrometry in concert with artificial neural networks, with special reference to Neisseria gonorrhoeae

Surface enhanced laser desorption/ionization-time of flight mass spectrometry (SELDI-TOF MS) has been applied in large numbers of oncological studies but the microbiological field has not been extensively explored to date. This paper describes the application of SELDI-TOF MS in concert with a multi-layer perceptron artificial neural network (ANN) with a back propagation algorithm for the identification of Neisseria gonorrhoeae. N. gonorrhoeae, the aetiological agent of gonorrhoea, is the second most common sexually transmitted disease in the UK and USA. Analysis of over 350 strains of N. gonorrhoeae and closely related species by SELDI-TOF MS facilitated the design of an ANN model and revealed 20 ion peak descriptors of positive, negative and secondary nature that were paramount for the identification of the pathogen. The model performed with over 96 % efficiency when based on these 20 ion peak descriptors and exhibited a sensitivity of 95.7 % and a specificity of 97.1 %, with an area under the curve value of 0.996. The technology has the potential to link several ANN models for a comprehensive rapid identification platform for clinically important pathogens.

Classification of bacterial species from proteomic data using combinatorial approaches incorporating artificial neural networks, cluster analysis and principal components analysis

MOTIVATION

Robust computer algorithms are required to interpret the vast amounts of proteomic data currently being produced and to generate generalized models which are applicable to 'real world' scenarios. One such scenario is the classification of bacterial species. These vary immensely, some remaining remarkably stable whereas others are extremely labile showing rapid mutation and change. Such variation makes clinical diagnosis difficult and pathogens may be easily misidentified.

RESULTS

We applied artificial neural networks (Neuroshell 2) in parallel with cluster analysis and principal components analysis to surface enhanced laser desorption/ionization (SELDI)-TOF mass spectrometry data with the aim of accurately identifying the bacterium Neisseria meningitidis from species within this genus and other closely related taxa. A subset of ions were identified that allowed for the consistent identification of species, classifying >97% of a separate validation subset of samples into their respective groups.

AVAILABILITY

Neuroshell 2 is commercially available from Ward Systems.

Classification of prostatic carcinoma with artificial neural networks using comparative genomic hybridization and quantitative stereological data

Staging of prostate cancer is a mainstay of treatment decisions and prognostication. In the present study, 50 pT2N0 and 28 pT3N0 prostatic adenocarcinomas were characterized by Gleason grading, comparative genomic hybridization (CGH), and histological texture analysis based on principles of stereology and stochastic geometry. The cases were classified by learning vector quantization and support vector machines. The quality of classification was tested by cross-validation. Correct prediction of stage from primary tumor data was possible with an accuracy of 74-80% from different data sets. The accuracy of prediction was similar when the Gleason score was used as input variable, when stereological data were used, or when a combination of CGH data and stereological data was used. The results of classification by learning vector quantization were slightly better than those by support vector machines. A method is briefly sketched by which training of neural networks can be adapted to unequal sample sizes per class. Progression from pT2 to pT3 prostate cancer is correlated with complex changes of the epithelial cells in terms of volume fraction, of surface area, and of second-order stereological properties. Genetically, this progression is accompanied by a significant global increase in losses and gains of DNA, and specifically by increased numerical aberrations on chromosome arms 1q, 7p, and 8p.

Incidental carcinoma of the prostate: clinicopathological, stereological and immunohistochemical findings studied with logistic regression and self-organizing feature maps

OBJECTIVE

To identify significant predictive factors determining category T1a and T1b in incidental prostatic carcinoma with classical and neural multivariate data analysis methods.

MATERIALS AND METHODS

Incidental prostatic carcinomas diagnosed in our department during 1990-99 (66 cases) were re-examined. Besides acquiring routine clinical and pathological data the tumours were assessed by scoring immunohistochemistry for proliferative activity and p53-overexpression. Tumour vascularization (angiogenesis) and epithelial texture variables were investigated by quantitative stereology. The data were evaluated by classical statistical methods (t-test, correlation analysis, logistic regression). Moreover, self-organizing feature maps (SOMs) were applied as an exploratory approach to unsupervised data analysis by artificial neural networks.

RESULTS

The proliferative fraction, p53 overexpression of tumour cell nuclei, preoperative prostate-specific antigen value and density of capillary vascularization correlated with the Gleason score in incidental prostatic carcinoma. In a stepwise logistic regression analysis with the tumour categories T1a and T1b as dependent variables, the Gleason score and the volume fraction of epithelial cells were significant independent predictors of the tumour category. The cases could be grouped into clusters of different degrees of malignancy using SOMs.

CONCLUSIONS

Texture variables of tumour cells are of central importance for the extent of propagation in the prostate in incidental prostatic adenocarcinomas. Gleason score and quantitative stereological estimates of the volume fraction of tumour cells are significant predictors of T1a and T1b categories of incidental prostatic carcinoma. Unsupervised clustering of T1 prostate carcinoma cases by SOMs correlates well with the dichotomous classification into T1a and T1b according to the UICC.

Artificial neural networks for decision-making in urologic oncology

The authors are presenting a thorough introduction in Artificial Neural Networks (ANNs) and their contribution to modern Urologic Oncology. The article covers a description of Artificial Neural Network methodology and points out the differences of Artificial Intelligence to traditional statistic models in terms of serving patients and clinicians, in a different way than current statistical analysis. Since Artificial Intelligence is not yet fully understood by many practicing clinicians, the authors have reviewed a careful selection of articles in order to explore the clinical benefit of Artificial Intelligence applications in modern Urology questions and decision-making. The data are from real patients and reflect attempts to achieve more accurate diagnosis and prognosis, especially in prostate cancer that stands as a good example of difficult decision-making in everyday practice. Experience from current use of Artificial Intelligence is also being discussed, and the authors address future developments as well as potential problems such as medical record quality, precautions in using ANNs or resistance to system use, in an attempt to point out future demands and the need for common standards. The authors conclude that both methods should continue to be used in a complementary manner. ANNs still do not prove always better as to replace standard statistical analysis as the method of choice in interpreting medical data.

Artificial neural networks in predicting optimum renal stone fragmentation by extracorporeal shock wave lithotripsy: a preliminary study

OBJECTIVE

To assess the ability of artificial neural networks (ANNs) to predict optimum renal stone fragmentation in patients being managed by extracorporeal shock wave lithotripsy (ESWL).

PATIENTS AND METHODS

The study included 82 patients with renal stones who were being treated by ESWL. Data (input and output values) from 60 patients in whom there was optimum fragmentation of stones by ESWL were used for training the ANN. These data mainly included the 24-h urinary variables, the radiological features of the stone disease and the ESWL settings used. The predictability of the trained ANN was tested on 22 subsequent patients, by supplying the input variables of the 22 patients into the trained ANN and recording the output values (predicted values). After subjecting these patients to ESWL, the actual results (observed values) were recorded. The predicted and the observed values were then compared.

RESULTS

In the 22 patients in whom predictability was tested, the trained ANN predicted optimum fragmentation at < or = 13 000 shocks/stone (as per study protocol) in 17 and optimum fragmentation at> 13 000 shocks/stone in the other five. In the 17 patients (test set) where the trained ANN had predicted optimum fragmentation at < or = 13 000 shocks/stone, the optimum fragmentation was at that value, although the predicted and observed values were not identical. The overall correlation between the predicted and the observed values was 75.5% (correlation coefficient 0.7547) in these 17 patients. Of the other five patients, none had optimum fragmentation at < 13 000 shocks/stone, as predicted by the trained ANN, giving complete accuracy for this factor.

CONCLUSION

This was a pilot study, i.e. an initial attempt to use an ANN in this regard, and although there were few patients, such that it is not possible to make final recommendations, the overall predictability was approximately 75%. An encouraging outcome of the study was that the trained ANN identified patients unlikely to benefit from ESWL. Using a larger dataset and identifying more significant variables, while eliminating inputs with a negative effect, the efficiency and utility of this ANN can probably be enhanced and in future it might be possible to predict stone fragmentation with reasonable accuracy.

Early prostate cancer: is there a need for new treatment options?

Improvements in diagnostic techniques have led to prostate cancer being diagnosed in younger patients and at an earlier stage of disease. The question therefore arises as to what is the best treatment for early prostate cancer. The main issues to be considered are whether the cancer is likely to progress quicker if these patients do not receive early treatment and what the quality of life implications are for patients receiving early treatment. As yet, due to the lack of valid comparisons of treatments, there is no clear "best treatment" for early prostate cancer. A number of clinical trials, comparing current treatments or investigating potential new treatment options for early prostate cancer, are in progress. The results of these should clarify the relative benefits of currently available treatments. This article reviews the latest information on the incidence, prognosis and current treatments for early prostate cancer and discusses the need for new treatments. Potential clinical benefits and cost implications of new treatments for early prostate cancer, such as improved surgical and radiotherapy techniques and adjuvant medical therapy, are also evaluated.

Postgenomics: Proteomics and Bioinformatics in Cancer Research

Now that the human genome is completed, the characterization of the proteins encoded by the sequence remains a challenging task. The study of the complete protein complement of the genome, the “proteome,” referred to as proteomics, will be essential if new therapeutic drugs and new disease biomarkers for early diagnosis are to be developed. Research efforts are already underway to develop the technology necessary to compare the specific protein profiles of diseased versus nondiseased states. These technologies provide a wealth of information and rapidly generate large quantities of data. Processing the large amounts of data will lead to useful predictive mathematical descriptions of biological systems which will permit rapid identification of novel therapeutic targets and identification of metabolic disorders. Here, we present an overview of the current status and future research approaches in defining the cancer cell's proteome in combination with different bioinformatics and computational biology tools toward a better understanding of health and disease.

Role of advanced 2 and 3-dimensional ultrasound for detecting prostate cancer

PURPOSE

We explored the clinical usefulness of spectrum analysis and neural networks for classifying prostate tissue and identifying prostate cancer in patients undergoing transrectal ultrasound for diagnostic or therapeutic reasons.

MATERIALS AND METHODS

Data on a cohort of 215 patients who underwent transrectal ultrasound guided prostate biopsies at Memorial-Sloan Kettering Cancer Center, New York, New York were included in this study. Radio frequency data necessary for 2 and 3-dimensional (D) computer reconstruction of the prostate were digitally recorded at transrectal ultrasound and prostate biopsy. The data were spectrally processed and 2-D tissue typing images were generated based on a pre-trained neural network classification. We used manually masked 2-D tissue images as building blocks for generating 3-D tissue images and the images were tissue type color coded using custom software. Radio frequency data on the study cohort were analyzed for cancer probability using the data set pre-trained by neural network methods and compared with conventional B-mode imaging. ROC curves were generated for the 2 methods using biopsy results as the gold standard.

RESULTS

The mean area under the ROC curve plus or minus SEM for detecting prostate cancer for the conventional B-mode and neural network methods was 0.66 +/- 0.03 and 0.80 +/- 0.05, respectively. Sensitivity and specificity for detecting prostate cancer by the neural network method were significantly increased compared with conventional B-mode imaging. In addition, the 2 and 3-D prostate images provided excellent visual identification of areas with a higher likelihood of cancer.

CONCLUSIONS

Spectrum analysis could significantly improve the detection and evaluation of prostate cancer. Routine real-time application of spectrum analysis may significantly decrease the number of false-negative biopsies and improve the detection of prostate cancer at transrectal ultrasound guided prostate biopsy. It may also provide improved identification of prostate cancer foci during therapeutic intervention, such as brachytherapy, external beam radiotherapy or cryotherapy. In addition, 2 and 3-D images with prostate cancer foci specifically identified can help surgical planning and may in the distant future be an additional reliable noninvasive method of selecting patients for prostate biopsy.

Comparison of Logistic Regression and Neural Net Modeling for Prediction of Prostate Cancer Pathologic Stage

Background: Prostate cancer (PCa) pathologic staging remains a challenge for the physician using individual pretreatment variables. We have previously reported that UroScoreTM, a logistic regression (LR)-derived algorithm, can correctly predict organ-confined (OC) disease state with &gt;90% accuracy. This study compares statistical and neural network (NN) approaches to predict PCa stage.

Methods: A subset (756 of 817) of radical prostatectomy patients was assessed: 434 with OC disease, 173 with capsular penetration (NOC-CP), and 149 with metastases (NOC-AD) in the training sample. Additionally, an OC + NOC-CP (n = 607) vs NOC-AD (n = 149) two-outcome model was prepared. Validation sets included 120 or 397 cases not used for modeling. Input variables included clinical and several quantitative biopsy pathology variables. The classification accuracies achieved with a NN with an error back-propagation architecture were compared with those of LR statistical modeling.

Results: We demonstrated &gt;95% detection of OC PCa in three-outcome models, using both computational approaches. For training patient samples that were equally distributed for the three-outcome models, NNs gave a significantly higher overall classification accuracy than the LR approach (40% vs 96%, respectively). In the two-outcome models using either unequal or equal case distribution, the NNs had only a marginal advantage in classification accuracy over LR.

Conclusions: The strength of a mathematics-based disease-outcome model depends on the quality of the input variables, quantity of cases, case sample input distribution, and computational methods of data processing of inputs and outputs. We identified specific advantages for NNs, especially in the prediction of multiple-outcome models, related to the ability to pre- and postprocess inputs and outputs.

Recent innovations in intensive care unit risk-prediction models

During the past 20 years, ICU risk-prediction models have undergone significant development, validation, and refinement. Among the general ICU severity of illness scoring systems, the Acute Physiology and Chronic Health Evaluation (APACHE), Mortality Prediction Model (MPM), and the Simplified Acute Physiology Score (SAPS) have become the most accepted and used. To risk-adjust patients with longer, more severe illnesses like sepsis and acute respiratory distress syndrome, several models of organ dysfunction or failure have become available, including the Multiple Organ Dysfunction Score (MODS), the Sequential Organ Failure Assessment (SOFA), and the Logistic Organ Dysfunction Score (LODS). Recent innovations in risk adjustment include automatic physiology and diagnostic variable retrieval and the use of artificial intelligence. These innovations have the potential of extending the uses of case-mix and severity-of-illness adjustment in the areas of clinical research, patient care, and administration. The challenges facing intensivists in the next few years are to further develop these models so that they can be used throughout the IUC stay to assess quality of care and to extend them to more specific patient groups such as the elderly and patients with chronic ICU courses.

Multicenter Evaluation of an Artificial Neural Network to Increase the Prostate Cancer Detection Rate and Reduce Unnecessary Biopsies

Background: The percentage of free prostate-specific antigen (%fPSA) has been shown to improve specificity for the diagnosis of prostate cancer (PCa) over total PSA (tPSA). A multicenter study was performed to evaluate the diagnostic value of a %fPSA-based artificial neural network (ANN) in men with tPSA concentrations between 2 and 20 μg/L for detecting patients with increased risk of a positive prostate biopsy for cancer.

Methods: We enrolled 1188 men from six different hospitals with PCa or benign prostates between 1996 and 2001. We used a newly developed ANN with input data of tPSA, %fPSA, patient age, prostate volume, and digital rectal examination (DRE) status to calculate the risk for the presence of PCa within different tPSA ranges (2–4, 4.1–10, 2–10, 10.1–20, and 2–20 μg/L) at the 90% and 95% specificity or sensitivity cutoffs, depending on the tPSA concentration. ROC analysis and cutoff calculations were used to estimate the diagnostic improvement of the ANN compared with %fPSA alone.

Results: In the low tPSA range (2–4 μg/L), the ANN detected 72% and 65% of cancers at specificities of 90% or 95%, respectively. At 4–10 μg/L tPSA, the ANN detected 90% and 95% of cancers with specificities of 62% and 41%, respectively. Use of the ANN with 2–10 μg/L tPSA enhanced the specificity of %fPSA by 20–22%, thus reducing the number of unnecessary biopsies.

Conclusions: Enhanced accuracy of PCa detection over that obtained using %fPSA alone can be achieved with a %fPSA-based ANN that also includes clinical information from DRE and prostate volume measurements.

An artificial neural network considerably improves the diagnostic power of percent free prostate-specific antigen in prostate cancer diagnosis: results of a 5-year investigation

Our study was performed to evaluate the diagnostic usefulness of %fPSA alone and combined with an ANN at different PSA concentration ranges, including the low range 2-4 ng/ml, to improve the risk assessment of prostate cancer. A total of 928 men with prostate cancer and BPH without any pretreatment of the prostate in the PSA range 2-20 ng/ml were enrolled in the study between 1996 and 2001. An ANN with input data of PSA, %fPSA, patient's age, prostate volume and DRE status was developed to calculate the individual's risk before performing a prostate biopsy within the different PSA ranges 2-4, 4.1-10 and 10.1-20 ng/ml. ROC analysis and cut-off calculations were used to estimate the diagnostic improvement of %fPSA and ANN in comparison to PSA. At the 90% sensitivity level, %fPSA and ANN performed better than PSA in all ranges, enhancing the specificity by 15-28% and 32-44%, respectively. For the low PSA range 2-4 ng/mL, we recommend a first-time biopsy at an ANN specificity level of 90%. For PSA 4-10 ng/mL, we recommend a first-time biopsy based on the ANN at the 90% sensitivity level. Use of an ANN enhances the %fPSA performance to further reduce the number of unnecessary biopsies within the PSA range 2-10 ng/ml.

Computer-assisted diagnostics: application to prostate cancer

Artificial neural networks (ANNs) have only recently been applied to solve problems in the diagnosis, staging, and prediction of treatment outcome in prostate cancer. A literature search provided information on 10 published journal articles that were selected for review and analysis. In all but one of the studies that compared the ANN output with logistic regression modeling, the ANN performed better. Specific training issues for neural networks are discussed and examples provided. We conclude that the continued development and refinement of computer-assisted diagnostic methodology are warranted to enhance conventional statistical approaches to the classification and pattern recognition found in data sets from men either suspected of having or known to have prostate cancer.

Predictive modeling techniques in prostate cancer

A number of new predictive modeling techniques have emerged in the past several years. These methods can be used independently or in combination with traditional modeling techniques to produce useful tools for the management of prostate cancer. Investigators should be aware of these techniques and avail themselves of their potentially useful properties. This review outlines selected predictive methods that can be used to develop models that may be useful to patients and clinicians for prostate cancer management.

Impact of different variables on the outcome of patients with clinically confined prostate carcinoma: prediction of pathologic stage and biochemical failure using an artificial neural network

BACKGROUND

The advent of advanced computing techniques has provided the opportunity to analyze clinical data using artificial intelligence techniques. This study was designed to determine whether a neural network could be developed using preoperative prognostic indicators to predict the pathologic stage and time of biochemical failure for patients who undergo radical prostatectomy.

METHODS

The preoperative information included TNM stage, prostate size, prostate specific antigen (PSA) level, biopsy results (Gleason score and percentage of positive biopsy), as well as patient age. All 309 patients underwent radical prostatectomy at the University of Colorado Health Sciences Center. The data from all patients were used to train a multilayer perceptron artificial neural network. The failure rate was defined as a rise in the PSA level > 0.2 ng/mL. The biochemical failure rate in the data base used was 14.2%. Univariate and multivariate analyses were performed to validate the results.

RESULTS

The neural network statistics for the validation set showed a sensitivity and specificity of 79% and 81%, respectively, for the prediction of pathologic stage with an overall accuracy of 80% compared with an overall accuracy of 67% using the multivariate regression analysis. The sensitivity and specificity for the prediction of failure were 67% and 85%, respectively, demonstrating a high confidence in predicting failure. The overall accuracy rates for the artificial neural network and the multivariate analysis were similar.

CONCLUSIONS

Neural networks can offer a convenient vehicle for clinicians to assess the preoperative risk of disease progression for patients who are about to undergo radical prostatectomy. Continued investigation of this approach with larger data sets seems warranted.

The relationship of the International Prostate Symptom Score and objective parameters for diagnosing bladder outlet obstruction. Part II: the potential usefulness of artificial neural networks

PURPOSE

The International Prostate Symptom Score (I-PSS) is used exclusively for evaluating patients with a prostate condition and following various treatment modalities. As previously demonstrated, there is poor or no correlation of bladder outlet obstruction diagnosed by pressure flow study with the symptoms projected by the I-PSS. Thus, we applied an artificial neural network model to assess patients with lower urinary tract symptoms.

MATERIALS AND METHODS

Data on 460 patients enrolled in part 1 of our study were entered into a multilayer feed forward, back propagation network.

RESULTS

In the training set of 305 patients the model predicted obstruction in 94% with 94% sensitivity and 68% specificity. In the test set of 155 patients it predicted obstruction in 87% with 87% sensitivity and 44% specificity.

CONCLUSIONS

The accuracy of the model for diagnosing obstruction based on the I-PSS is acceptable, considering that statistical models failed to demonstrate a correlation of symptoms with objective obstruction.

Predicting the outcome of prostate biopsy in screen-positive men by a multilayer perceptron network

OBJECTIVES

To assess whether an artificial neural network (multilayer perceptron, MLP) and logistic regression (LR) could eliminate more false-positive prostate-specific antigen (PSA) results than the proportion of free PSA in a prostate cancer screening.

METHODS

MLP and LR models were constructed on the basis of data on total PSA, the proportion of free PSA, digital rectal examination (DRE), and prostate volume from 656 consecutive men (aged 55 to 67 years) with total serum PSA concentrations of 4 to 10 ng/mL in the randomized population-based prostate cancer screening study in Finland. The MLP and LR models were validated using the "leave-one-out" method.

RESULTS

Of the 656 men, 23% had prostate cancer and 77% had either normal prostatic histology or a benign disease. At a 95% sensitivity level, 19% of the false-positive PSA results could be eliminated by using the proportion of free PSA versus 24% with the LR model and 33% with the MLP model (P < 0.001). At 80% to 99% sensitivity levels, the accuracy of the MLP and LR models was significantly higher than that of the proportion of free PSA. At 89% to 99% sensitivities, the accuracy of the MLP was higher than that of LR (P </= 0.001).

CONCLUSIONS

At clinically relevant sensitivity levels, the MLP and LR models based on total PSA, the proportion of free PSA, DRE, and prostate volume could reduce the number of unnecessary biopsies significantly better than the proportion of free PSA alone in men with total PSA levels in the range 4 to 10 ng/mL.