Composing Diverse Ensembles of Convolutional Neural Networks by Penalization

Background and Purpose:Accurate automated infarct segmentation is needed for acute ischemic stroke studies relying on infarct volumes as an imaging phenotype or biomarker that require large numbers of subjects. This study investigates whether an ensemble of convolutional neural networks (CNN) trained on multiparametric DWI maps outperforms single networks trained on solo DWI parametric maps.Materials and Methods:CNNs were trained on combinations of DWI, ADC, and low b-value-weighted images from 116 subjects. The performances of the networks (measured by Dice score, sensitivity and precision) were compared to one another and to ensembles of 5 networks. To assess the generalizability of the approach, the best performing model was applied to an independent evaluation cohort of 151 subjects. Agreement between manual and automated segmentations for identifying patients with large lesions volumes was calculated across multiple thresholds (21 cm3, 31 cm3, 51 cm3, and 70 cm3).ResultsAn ensemble of CNNs trained on DWI, ADC and low b-value-weighted images produced the most accurate acute infarct segmentation over individual networks (p<0.0001). Automated volumes correlated with manually measured volumes (Spearman’s ρ=0.91, p<0.0001) for the independent cohort. For the task of identifying patients with large lesion volumes, agreement between manual outlines and automated outlines was high (Cohen’s κ 0.86 to 0.90, p<0.0001).ConclusionAcute infarcts are more accurately segmented using ensembles of CNNs trained with multi-parametric maps than using a single model trained with a solo map. Automated lesion segmentation can perform with high agreement with manual techniques for identifying patients with large lesion volumes.

This work presents a system based on an ensemble of Convolutional Neural Networks (CNNs) and descriptors for bioimage classification that has been validated on different datasets of color images. The proposed system represents a very simple yet effective way of boosting the performance of trained CNNs by composing multiple CNNs into an ensemble and combining scores by sum rule. Several types of ensembles are considered, with different CNN topologies along with different learning parameter sets. The proposed system not only exhibits strong discriminative power but also generalizes well over multiple datasets thanks to the combination of multiple descriptors based on different feature types, both learned and handcrafted. Separate classifiers are trained for each descriptor, and the entire set of classifiers is combined by sum rule. Results show that the proposed system obtains state-of-the-art performance across four different bioimage and medical datasets. The MATLAB code of the descriptors will be available at https://github.com/LorisNanni.

Training an ensemble of convolutional neural networks requires much computational resources for a large set of high-resolution medical 3D scans because deep representation requires many parameters and layers. In this study, 100 3D late gadolinium-enhanced (LGE)-MRIs with a spatial resolution of 0.625 mm × 0.625 mm × 0.625 mm from patients with atrial fibrillation were utilized. To contain cost of the training, down-sampling of images, transfer learning and ensemble of network’s past weights were deployed. This paper proposes an image processing stage using down-sampling and contrast limited adaptive histogram equalization, a network training stage using a cyclical learning rate schedule, and a testing stage using an ensemble. While this method achieves reasonable segmentation accuracy with the median of the Dice coefficients at 0.87, this method can be used on a computer with a GPU that has a Kepler architecture and at least 3 GB memory.

Imbalanced classification using deep learning has attracted much attention in intelligent fault diagnosis of machinery. However, the existing methods use individual deep neural network to extract features and recognize the health conditions under imbalanced dataset, which may easily over-fit the mechanical data and affect the diagnosis accuracy. To deal with this problem, this paper takes the advantages of ensemble learning and proposes an ensemble convolutional neural network (EnCNN) for the intelligent fault diagnosis for machines under imbalanced data. In the proposed method, a convolutional neural network with the input of multi-sensor signals is used as the base classifier. Firstly, the mechanical imbalance dataset is first split into balanced training subsets through under-sampling strategy, and each subset is used to train a base classifier. Then the weight coefficients of each trained base classifier are calculated by G-mean score and anomalous base classifiers are screened using classifier selection. Finally, the base classifiers are integrated into EnCNN through weighted voting strategy. The proposed EnCNN is validated by the imbalanced dataset collected from a machinery fault test bench. By comparing with the related methods, the superiority of EnCNN is verified in intelligent fault diagnosis of machines under imbalanced data.

Classification method for nailfold capillary images using an optimized sugeno fuzzy ensemble of convolutional neural networks.

This letter proposes a novel way of using an ensemble of multiple deep convolutional neural networks (DCNNs) for Alzheimer's disease classification, based on magnetic resonance imaging (MRI) images. To create this ensemble of DCNNs, we propose to combine the use of multiple MRI projections (as input) with that of different DCNN architectures to increase the deep ensemble diversity. In particular, to find the optimal fusion weights of the DCNN members, we designed a novel deep ensemble generalization loss, which accounts for interaction and cooperation during the optimal weight search. The optimization framework, equipped with our ensemble generalization loss, was formulated and solved using the sequential quadratic programming. Through this method, we achieved optimal DCNN fusion weights (i.e., a high generalization performance). The experimental results showed that our proposed DCNN ensemble outperforms current deep learning-based methods: it is able to produce state-of-the-art results on the Alzheimer's disease neuroimaging initiative (ADNI) dataset.

Alzheimer's disease (AD) is one of the most common progressive neurodegenerative diseases. Structural magnetic resonance imaging (MRI) would provide abundant information on the anatomical structure of human organs. Fluorodeoxy‐glucose positron emission tomography (PET) obtains the metabolic activity of the brain. Previous studies have demonstrated that multi‐modality images could contribute to improve diagnosis of AD. However, these methods need to extract the handcrafted features that demand domain specific knowledge and image processing stage is time consuming. In order to tackle these problems, in this study, the authors propose a novel framework that ensembles three state‐of‐the‐art deep convolutional neural networks (DCNNs) with multi‐modality images for AD classification. In detail, they extract some slices from each subject of each modality, and every DCNN generates a probabilistic score for the input slices. Furthermore, a ‘dropout’ mechanism is introduced to discard low discrimination slices of the category probabilities. Then average reserved slices of each subject are acquired as a new feature. Finally, they train the Adaboost ensemble classifier based on single decision tree classifier with the MRI and PET probabilistic scores of each DCNN. Evaluations on Alzheimer's Disease Neuroimaging Initiative database show that the proposed algorithm has better performance compared to existing method, the algorithm proposed in this study significantly improved the classification accuracy.

Model optimization is a problem of great concern and challenge for developing an image classification model. In image classification, selecting the appropriate hyperparameters can substantially boost the model’s ability to learn intricate patterns and features from complex image data. Hyperparameter optimization helps to prevent overfitting by finding the right balance between complexity and generalization of a model. The ensemble genetic algorithm and convolutional neural network (EGACNN) are proposed to enhance image classification by fine-tuning hyperparameters. The convolutional neural network (CNN) model is combined with a genetic algorithm GA) using stacking based on the Modified National Institute of Standards and Technology (MNIST) dataset to enhance efficiency and prediction rate on image classification. The GA optimizes the number of layers, kernel size, learning rates, dropout rates, and batch sizes of the CNN model to improve the accuracy and performance of the model. The objective of this research is to improve the CNN-based image classification system by utilizing the advantages of ensemble learning and GA. The highest accuracy is obtained using the proposed EGACNN model which is 99.91% and the ensemble CNN and spiking neural network (CSNN) model shows an accuracy of 99.68%. The ensemble approaches like EGACNN and CSNN tends to be more effective as compared to CNN, RNN, AlexNet, ResNet, and VGG models. The hyperparameter optimization of deep learning classification models reduces human efforts and produces better prediction results. Performance comparison with existing approaches also shows the superior performance of the proposed model.

Crop pest classification with a genetic algorithm-based weighted ensemble of deep convolutional neural networks

Pulmonary tuberculosis is one of the top 10 causes of death in the world in which Latin America represents 3% of the worldwide incidence with a mortality rate of 7.3%. According to goals of World Health Organization to end the global tuberculosis epidemy by 2030, the development of novel diagnosis strategies is crucial. Taking this context into account, the present work shows results on the use of ensembles of Convolutional Neural Networks to aid Computer Vision diagnosis of tuberculosis from patients' chest X-ray. Our solution comprises an ensemble of three different deep architectures of such networks having accuracy higher than 93% on the proposed task. The solution was obtained and evaluated in real-world data from public datasets, favouring reproducibility, surpasses human experts performance reported by literature in these very same datasets, uses canonical architectures of convolutional neural networks, with and without Transfer Learning, and requires minimal effort on data preparation and no previous feature extraction.

Multi-model ensemble is an important fundamental technique of practical value for many artificial intelligence applications. However, the usage for multi-model ensemble has been limited when it is combined with deep neural networks to construct ensemble of deep neural networks. Due to the big time and computing resources required to train and to integrate multiple deep neural networks for the achievement of multi-model ensemble, the engineering application field where developing time and computing resources are usually restricted, has not yet widespreadly benefited from ensemble of deep neural networks. To alleviate this situation, we present a new multi-model ensemble approach entitled feature transferring based multi-model ensemble (FTBME), for ensemble of deep neural networks. Primarily, we propose a feature transferring based multi-model training strategy to more affordably find multiple extra models based on a given previously optimized deep neural network model. Sequentially, to develop better ensemble solutions, we design a more effective random greedy based ensemble selection strategy to filter out models non-positive to ensemble generalization. Finally, inspired by the idea of averaging parameter points, we propose to fuse the obtained models in weight space which eventually reduces the expense of ensemble at the testing stage to a single deep neural network model while retaining the generalization. These three advances constitute the resulting technique FTBME. We conducted extensive experiments using deep neural networks, from light weight to complex, on ImageNet, CIFAR-10 and CIFAR-100. Results show that, given a deep neural network model which has been well-optimized and reaching its limit, FTBME can obtain better generalization with minor extra training expense while maintaining the expense to a single model at ensemble testing. This promising property of FTBME make us believe that it could be leveraged to broaden the usage for ensemble of deep neural networks, alleviating the situation that the engineering application field has not yet widespreadly benefited from ensemble of deep neural networks.

The early and accurate detection of brain tumors is a critical challenge in diagnostics and healthcare due to the severe consequences of delayed diagnosis. This paper addresses this issue by employing an ensemble of Convolutional Neural Network (CNN) models to enhance the identification of brain tumors using MRI images. The methodology integrates pre-processing techniques such as image augmentation, Gaussian blurring, and Sobel edge detection to improve image quality. Various CNN architectures, including Scratch CNN, InceptionV3, Xception, EfficientNetB0, ResNet50, and VGG19, were evaluated alongside machine learning classifiers such as AdaBoost, Random Forest, SVM, KNN, and SoftMax. Among these, EfficientNetB0, Xception, and InceptionV3 demonstrated superior performance, achieving the highest classification accuracy of 98.67% and an average accuracy of 96.90%. This research underscores the significance of selecting appropriate models and classifiers for medical image classification and highlights the potential for further advancements in clinical applications.

Computerized classification of gastrointestinal polyps using stacking ensemble of convolutional neural network

The World Health Organization (WHO) identified Gliomas as the most widespread brain tumour category. Toxicity levels combined with growth rates serve as the basis for dividing glioma malignancies into four distinct categories. The accurate grading of tumors determines which treatment method could be best suited including chemotherapy, surgery or radiation therapy. The use of magnetic resonance imaging (MRI) images to separate and grade tumours works well yet demands significant time and money and includes intricate management variations. The objective of this study is to meet the need for precise and reliable tumour classification by proposing a computational method for glioma grading that leverages an ensemble of Convolutional Neural Networks (CNNs). The proposed method introduces two ensemble models named Mean-Ensemble-CNNs & NN-Ensemble-CNNs that merge the pretrained CNN models ResNet-101, ResNeXt-50, Inception-v3 and DenseNet-161. Neural Network Ensembles combine deep features from multiple algorithms through a neural network to identify gliomas while Mean Ensembles calculate the average predictions from various algorithms to detect gliomas. The following sections provide a detailed analysis of Mean-Ensemble-CNNs. The system utilizes advanced data preparation and augmentation methods to fully exploit available informational resources. The technique reduces dependence on invasive biopsies while enabling flexible clinical treatment planning. Automatic segmentation and grading of tumours makes this process possible. The proposed NN-Ensemble CNNs achieve the top score with 96.21% accuracy which surpasses all individual models.

We present a deep learning method to downscale low-resolution geophysical fields by merging them with high-resolution data. The downscaling was performed using an ensemble of convolutional neural networks (CNNs), whose prediction values are the average values of the outputs of 20 CNNs. Academic experiments were conducted on simulated ocean data in the Gulf Stream region, given by the outputs of the NATL60 model. The CNNs forced with low-resolution (120 × 120 km) sea surface high (SSH) data and mesoscale resolution (12 × 12 km) sea surface temperature (SST) data allowed us to obtain mesoscale resolution sea surface currents with good accuracy. Sensitivity experiments have shown that taking SST into account significantly increases the accuracy of the high-resolution velocity retrieval, even when noise is added to the SSH data. The velocity information embedded in the transport equation modeling the SST advection is taken into account by the CNN, which greatly increases the resolution of ocean currents provided by SSH. In the present work, we only consider spatial downscaling by assuming that SSH and SST are daily observations. The method we developed is generic and can be used to improve the resolution of a wide variety of large-scale fields by merging them with high-resolution fields.