Classification of normal and abnormal electrogastrograms using multilayer feedforward neural networks

A neural network approach is proposed for the automated classification of the normal and abnormal EGG. Two learning algorithms, the quasi-Newton and the scaled conjugate gradient method for the multilayer feedforward neural networks (MFNN), are introduced and compared with the error backpropagation algorithm. The configurations of the MFNN are determined by experiment. The raw EGG data, its power spectral data, and its autoregressive moving average (ARMA) modelling parameters are used as the input to the MFNN and compared with each other. Three indexes (the percent correct, sumsquared error and complexity per iteration) are used to evaluate the performance of each learning algorithm. The results show that the scaled conjugate gradient algorithm performs best, in that it is robust and provides a super-linear convergence rate. The power spectral representation and the ARMA modelling parameters of the EGG are found to be better types of the input to the network for this specific application, both yielding a percent correctness of 95% on the test set. Although the results are focused on the classification of the EGG, this paper should provide useful information for the classification of other biomedical signals.     

An ischemia detection method based on artificial neural networks

An automated technique was developed for the detection of ischemic episodes in long duration electrocardiographic (ECG) recordings that employs an artificial neural network. In order to train the network for beat classification, a cardiac beat dataset was constructed based on recordings from the European Society of Cardiology (ESC) ST-T database. The network was trained using a Bayesian regularisation method. The raw ECG signal containing the ST segment and the T wave of each beat were the inputs to the beat classification system and the output was the classification of the beat. The input to the network was produced through a principal component analysis (PCA) to achieve dimensionality reduction. The network performance in beat classification was tested on the cardiac beat database providing 90% sensitivity (Se) and 90% specificity (Sp). The neural beat classifier is integrated in a four-stage procedure for ischemic episode detection. The whole system was evaluated on the ESC ST-T database. When aggregate gross statistics was used the Se was 90% and the positive predictive accuracy (PPA) 89%. When aggregate average statistics was used the Se became 86% and the PPA 87%. These results are better than other reported.     

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.     

Utilization of artificial neural networks in the diagnosis of optic nerve diseases

This research is concentrated on the diagnosis of optic nerve disease through the analysis of pattern electroretinography (PERG) signals with the help of artificial neural network (ANN). Multilayer feed forward ANN trained with a Levenberg Marquart (LM) backpropagation algorithm was implemented. The designed classification structure has about 96.4% sensitivity, 90.4% specifity and positive prediction is calculated to be 94.2%. The end results are classified as healthy and diseased. Testing results were found to be compliant with the expected results that are derived from the physician's direct diagnosis. The end benefit would be to assist the physician to make the final decision without hesitation.     

Magnesium degradation as determined by artificial neural networks

Magnesium degradation under physiological conditions is a highly complex process in which temperature, the use of cell culture growth medium and the presence of CO₂, O₂ and proteins can influence the corrosion rate and the composition of the resulting corrosion layer. Due to the complexity of this process it is almost impossible to predict the parameters that are most important and whether some parameters have a synergistic effect on the corrosion rate. Artificial neural networks are a mathematical tool that can be used to approximate and analyse non-linear problems with multiple inputs. In this work we present the first analysis of corrosion data obtained using this method, which reveals that CO₂ and the composition of the buffer system play a crucial role in the corrosion of magnesium, whereas O₂, proteins and temperature play a less prominent role.     

Image feature analysis and computer-aided diagnosis in digital radiography: classification of normal and abnormal lungs with interstitial disease in chest images

In order to detect and characterize interstitial disease in the lungs, we are developing an automated method for the determination of physical texture measures, which assess the magnitude and coarseness (or fineness) of lung texture in digital chest radiographs. This method is based on an analysis of the power spectrum of lung texture. We now describe an automated classification method for distinction between normal and abnormal lungs with interstitial disease, in which we employ these texture measures and their data base. This computerized method includes three independent tests, one for a definitely abnormal focal pattern, one for a relatively localized abnormal pattern, and one for a diffuse abnormal pattern. The performance of this computerized classification scheme is compared with that of radiologists by means of receiver operating characteristic (ROC) analysis. Our results indicate that this computerized method can be a valuable aid to radiologists in their assessment of interstitial infiltrates.     

Swallow segmentation with artificial neural networks and multi-sensor fusion

Swallow segmentation is a critical precursory step to the analysis of swallowing signal characteristics. In an effort to automatically segment swallows, we investigated artificial neural networks (ANN) with information from cervical dual-axis accelerometry, submental MMG, and nasal airflow. Our objectives were (1) to investigate the relationship between segmentation performance and the number of signal sources and (2) to identify the signals or signal combinations most useful for swallow segmentation. Signals were acquired from 17 healthy adults in both discrete and continuous swallowing tasks using five stimuli. Training and test feature vectors were constructed with variances from single or multiple signals, estimated within 200 ms moving windows with 50% overlap. Corresponding binary target labels (swallow or non-swallow) were derived by manual segmentation. A separate 3-layer ANN was trained for each participant–signal combination, and all possible signal combinations were investigated. As more signal sources were included, segmentation performance improved in terms of sensitivity, specificity, accuracy, and adjusted accuracy. The combination of all four signal sources achieved the highest mean accuracy and adjusted accuracy of 88.5% and 89.6%, respectively. A–P accelerometry proved to be the most discriminatory source, while the inclusion of MMG or nasal airflow resulted in the least performance improvement. These findings suggest that an ANN, multi-sensor fusion approach to segmentation is worthy of further investigation in swallowing studies.     

Stability problems with artificial neural networks and the ensemble solution

Artificial neural networks (ANNs) are very popular as classification or regression mechanisms in medical decision support systems despite the fact that they are unstable predictors. This instability means that small changes in the training data used to build the model (i.e. train the ANN) may result in very different models. A central implication of this is that different sets of training data may produce models with very different generalisation accuracies. In this paper, we show in detail how this can happen in a prediction system for use in in-vitro fertilisation. We argue that claims for the generalisation performance of ANNs used in such a scenario should only be based on k-fold cross-validation tests. We also show how the accuracy of such a predictor can be improved by aggregating the output of several predictors.     

Prediction of vascular tissue engineering results with artificial neural networks

Tissue engineers are often confused on finding the most successful strategy for specific patient. In this study, we used artificial neural networks to predict the outcomes of different vascular tissue engineering strategies, thus providing advisory information for experimental designers. Over 30 variables were used as features of the tissue engineering strategies. Different architectures of artificial neural networks with back propagation algorithm were tested to obtain the best model configuration for the prediction of the tissue engineering strategies. In the computational experiments, the artificial neural networks with one and two hidden layers could, respectively, detect unsuccessful strategies with the highest predictive accuracy of 91.45 and 94.24%. In conclusion, artificial intelligence has great potential in tissue engineering decision support. It can provide accurate advisory information for tissue engineers, thus reducing failures and improving therapeutic effects.     

A topographical method for the development of neural networks for artificial brain evolution

Developmental neural networks, which are constructed according to developmental rules (i.e., genes), have the potential to be differentiated into heteromorphic neural structures capable of performing various kinds of activities. The fact that the biological neural architectures are found to be highly repetitive, layered, and topographically organized has important consequences for neural development methods. The purpose of this article is to propose a neural development method that can construct topographical neural connections, that is, a topographical development method, to facilitate fast and efficient development. This is achieved by arborizing neural connections on a developmental tree that rarely produces dead connections. Modular gene expression and corresponding modular networks have an important role in a gradual evolutionary process. Gene expression for modular networks is also proposed here as a way to reduce the probability of fatal mutants created through gene alteration. The corresponding evolutionary experiment shows that various neural structures--layered, repetitive, modular, and complex ones like those in the biological brain--can be constructed and easily observed. It also demonstrates that due to the efficiency of the proposed method, large neural networks can be easily managed, thereby making it suitable for long duration evolutionary experiments.     

Generating compensation designs for tangential breast irradiation with artificial neural networks

In this paper we discuss a study comparing an algorithm implemented clinically to design intensity-modulated fields with two artificial neural networks (ANNs) trained to design the same fields. The purpose of the algorithm is to produce compensation for tangential breast radiotherapy in order to improve dose homogeneity. This was achieved by creating intensity-modulated fields to supplement standard wedged fields. Portal image data were used to create thickness maps of the medial and lateral fields, which in turn were used to design the wedged and intensity-modulated fields. The ANNs were developed to design the intensity-modulated fields from the portal image data and corresponding fluence map alone. One used localized groups of portal image pixels related to the fluence map (method 2), and the other used a one-to-one mapping between spatially corresponding pixels (method 3). A dosimetric comparison of the methods was performed by calculating the overall dose distribution. The volume of tissue outside the dose range 95-105% was used to assess dose homogeneity. The average volume outside 95-105%, averaged over 80 cases, was shown to be 2.3% for the algorithm, whilst average values of 9.9% and 13.5% were obtained for methods 2 and 3, respectively. The results of this study demonstrate the ability of an ANN to learn the general shape of compensation required and explore the use of image-based ANNs in the design of intensity-modulated fields.     

Classification of reflectance spectra from pigmented skin lesions, a comparison of multivariate discriminant analysis and artificial neural networks

Successful treatment of skin cancer, especially melanoma, depends on early detection, but diagnostic accuracy, even by experts, can be as low as 56% so there is an urgent need for a simple, accurate, non-invasive diagnostic tool. In this paper we have compared the performance of an artificial neural network (ANN) and multivariate discriminant analysis (MDA) for the classification of optical reflectance spectra (320 to 1100 nm) from malignant melanoma and benign naevi. The ANN was significantly better than MDA, especially when a larger data set was used, where the classification accuracy was 86.7% for ANN and 72.0% for MDA (p < 0.001). ANN was better at learning new cases than MDA for this particular classification task. This study has confirmed that the convenience of ANNs could lead to the medical community and patients benefiting from the improved diagnostic performance which can be achieved by objective measurement of pigmented skin lesions using spectrophotometry.     

Reduction of false positives in computerized detection of lung nodules in chest radiographs using artificial neural networks,discriminant analysis,and a rule-based scheme

A computer-aided diagnosis (CAD) scheme is being developed to identify image regions considered suspicious for lung nodules in chest radiographs to assist radiologists in making correct diagnoses. Automated classifiers—an artificial neural network, discriminant analysis, and a rule-based scheme—are used to reduce the number of false-positive detections of the CAD scheme. The CAD scheme first detects nodule candidates from chest radiographs based on a difference image technique. Nine image features characterizing nodules are extracted automatically for each of the nodule candidates. The extracted image features are then used as input data to the classifiers for distinguishing actual nodules from the false-positive detections. The performances of the classifiers are evaluated by receiver-operating characteristic analysis. On the basis of the database of 30 normal and 30 abnormal chest images, the neural network achieves an A_Z value (area under the receiver-operating-characteristic curve) of 0.79 in detecting lung nodules, as tested by the round-robin method. The neural network, after being trained with a training database, is able to eliminate more than 83% of the false-positive detections reported by the CAD scheme. Moreover, the combination of the trained neural network and a rule-based scheme eliminates 96% of the false-positive detections of the CAD scheme.     

Application of artificial neural networks for quantitative analysis of image data in chest radiographs for detection of interstitial lung disease

The authors have developed an automated computeraided diagnostic (CAD) scheme by using artificial neural networks (ANNs) on quantitative analysis of image data. Three separate ANNs were applied for detection of interstitial disease on digitized chest images. The first ANN was trained with horizontal profiles in regions of interest (ROIs) selected from normal and abnormal chest radiographs for distinguishing between normal and abnormal patterns. For training and testing of the second ANN, the vertical output patterns obtained from the 1st ANN were used for each ROI. The output value of the second ANN was used to distinguish between normal and abnormal ROIs with interstitial infiltrates. If the ratio of the number of abnormal ROIs to the total number of all ROIs in a chest image was greater than a specified threshold level, the image was classified as abnormal. In addition, the third ANN was applied to distinguish between normal and abnormal chest images. The combination of the rule-based method and the third ANN also was applied to the classification between normal and abnormal chest images. The performance of the ANNs was evaluated by means of receiver operating characteristic (ROC) analysis. The average A_z value (area under the ROC curve) for distinguishing between normal and abnormal cases was 0.976±0.012 for 100 chest radiographs that were not used in training of ANNs. The results indicate that the ANN trained with image data can learn some statistical properties associated with interstitial infiltrates in chest radiographs.     

Classification of acute myocardial ischemia by artificial neural network using echocardiographic strain waveforms

Echocardiographic strain waveforms are highly variable, so their interpretation is experience-dependent and subjective. We tested whether an artificial neural network (ANN) can distinguish between strain waveforms obtained at baseline and during experimentally induced acute ischemia. An open-chest model of coronary occlusion and acute ischemia was used in 14 adult pigs. Strain waveforms were obtained using a GE Vivid 7 ultrasound system. An ANN design was implemented in MATLAB, and backpropagation and "leave-one-out" processes were used to train and test it. Specificity of 86% and sensitivity of 87% suggest that ANNs could aid in diagnostic prescreening of echocardiographic strain waveforms.     

Realistic animation of human figures using artificial neural networks

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

Feature-based classification of myoelectric signals using artificial neural networks

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

Classification of lung nodules in CT scans using three-dimensional deep convolutional neural networks with a checkpoint ensemble method

Background

Accurately detecting and examining lung nodules early is key in diagnosing lung cancers and thus one of the best ways to prevent lung cancer deaths. Radiologists spend countless hours detecting small spherical-shaped nodules in computed tomography (CT) images. In addition, even after detecting nodule candidates, a considerable amount of effort and time is required for them to determine whether they are real nodules. The aim of this paper is to introduce a high performance nodule classification method that uses three dimensional deep convolutional neural networks (DCNNs) and an ensemble method to distinguish nodules between non-nodules.

Methods

In this paper, we use a three dimensional deep convolutional neural network (3D DCNN) with shortcut connections and a 3D DCNN with dense connections for lung nodule classification. The shortcut connections and dense connections successfully alleviate the gradient vanishing problem by allowing the gradient to pass quickly and directly. Connections help deep structured networks to obtain general as well as distinctive features of lung nodules. Moreover, we increased the dimension of DCNNs from two to three to capture 3D features. Compared with shallow 3D CNNs used in previous studies, deep 3D CNNs more effectively capture the features of spherical-shaped nodules. In addition, we use an alternative ensemble method called the checkpoint ensemble method to boost performance.

Results

The performance of our nodule classification method is compared with that of the state-of-the-art methods which were used in the LUng Nodule Analysis 2016 Challenge. Our method achieves higher competition performance metric (CPM) scores than the state-of-the-art methods using deep learning. In the experimental setup ESB-ALL, the 3D DCNN with shortcut connections and the 3D DCNN with dense connections using the checkpoint ensemble method achieved the highest CPM score of 0.910.

Conclusion

The result demonstrates that our method of using a 3D DCNN with shortcut connections, a 3D DCNN with dense connections, and the checkpoint ensemble method is effective for capturing 3D features of nodules and distinguishing nodules between non-nodules.

     

Improvement and automation of artificial neural networks to estimate medical outcomes

The lengthy process of manually optimizing a feedforward backpropagation artificial neural network (ANN) provided the incentive to develop an automated system that could fine-tune the network parameters without user supervision. A new stopping criterion was introduced--the logarithmic-sensitivity index--that manages a good balance between sensitivity and specificity of the output classification. The automated network automatically monitored the classification performance to determine when was the best time to stop training-after no improvement in the performance measure (either highest correct classification rate, lowest mean squared error or highest log-sensitivity index value) occurred in the subsequent 500 epochs. Experiments were performed on three medical databases: an adult intensive care unit, a neonatal intensive care unit and a coronary surgery patient database. The optimal network parameter settings found by the automated system were similar to those found manually. The results showed that the automated networks performed equally well or better than the manually optimized ANNs, and the best classification performance was achieved using the log-sensitivity index as a stopping criterion.