In recent years, interest has grown in using computer-aided diagnosis (CAD) for Alzheimer's disease (AD) and its prodromal stage, mild cognitive impairment (MCI). However, existing CAD technologies often overfit data and have poor generalizability. In this study, we proposed a sparse-response deep belief network (SR-DBN) model based on rate distortion (RD) theory and an extreme learning machine (ELM) model to distinguish AD, MCI, and normal controls (NC). We used [18F]-AV45 positron emission computed tomography (PET) and magnetic resonance imaging (MRI) images from 340 subjects enrolled in the ADNI database, including 116 AD, 82 MCI, and 142 NC subjects. The model was evaluated using five-fold cross-validation. In the whole model, fast principal component analysis (PCA) served as a dimension reduction algorithm. An SR-DBN extracted features from the images, and an ELM obtained the classification. Furthermore, to evaluate the effectiveness of our method, we performed comparative trials. In contrast experiment 1, the ELM was replaced by a support vector machine (SVM). Contrast experiment 2 adopted DBN without sparsity. Contrast experiment 3 consisted of fast PCA and an ELM. Contrast experiment 4 used a classic convolutional neural network (CNN) to classify AD. Accuracy, sensitivity, specificity, and area under the curve (AUC) were examined to validate the results. Our model achieved 91.68% accuracy, 95.47% sensitivity, 86.68% specificity, and an AUC of 0.87 separating between AD and NC groups; 87.25% accuracy, 79.74% sensitivity, 91.58% specificity, and an AUC of 0.79 separating MCI and NC groups; and 80.35% accuracy, 85.65% sensitivity, 72.98% specificity, and an AUC of 0.71 separating AD and MCI groups, which gave better classification than other models assessed.
We proposed a modified concrete bridge crack detector based on a deep learning-assisted image processing approach. Data augmentation technology is introduced to extend the limited dataset. In our proposed method, the bounding box for the crack is detected by YOLOv5. Then, the image covered by the bounding box is processed by the image processing techniques. Compared with the conventional image processing-based crack detection method, the deep learning-assisted image processing approach leads to higher detection accuracy and lower computation cost. More precisely, the mask filter is employed to remove handwritten marks, and the ratio filter is adopted to eliminate speckle linear noises. When a single crack is detected by several bounding boxes, we proposed a novel fusion method to merge these bounding boxes. Furthermore, we proposed a connected component search approach based on the crack trend of the area to improve the connection accuracy. With the crack detector, the cracks that are wider than 0.15 mm can be correctly detected, quantified, and visualized. The detection absolute error of the crack width is less than 0.05 mm. Thus, we realized fast and precise detection and quantification of bridge crack based on the practical engineering dataset.
We propose a deep convolutional spiking neural network (DCSNN) with direct training to classify concrete bridge damage in a real engineering environment. The leaky-integrate-and-fire (LIF) neuron model is employed in our DCSNN that is similar to VGG. Poisson encoding and convolution encoding strategies are considered. The gradient surrogate method is introduced to realize the supervised training for the DCSNN. In addition, we have examined the effect of observation time step on the network performance. The testing performance for two different spike encoding strategies are compared. The results show that the DCSNN using gradient surrogate method can achieve a performance of 97.83%, which is comparable to traditional CNN. We also present a comparison with STDP-based unsupervised learning and a converted algorithm, and the proposed DCSNN is proved to have the best performance. To demonstrate the generalization performance of the model, we also use a public dataset for comparison. This work paves the way for the practical engineering applications of the deep SNNs.
Spiking neural networks (SNNs) offer powerful computation capability due to its event-driven nature and temporal processing. However, it is still limited to shallow structure and simple tasks due to the training difficulty. In this work, we propose a deep convolutional residual spiking neural network (DCRSNN) for text classification tasks. In the DCRSNN, the feature extraction is achieved via a convolution SNN with residual connection, using the surrogate gradient direct training technique. Classification is performed by a fully-connected network. We also suggest a hybrid photonic DCRSNN, in which photonic SNNs are used for classification with a converted training method. The accuracy of hard and soft reset methods, as well as three different surrogate functions, were evaluated and compared across four different datasets. Results indicated a maximum accuracy of 76.36% for MR, 91.03% for AG News, 88.06% for IMDB and 93.99% for Yelp review polarity. Soft reset methods used in the deep convolutional SNN yielded slightly better accuracy than their hard reset counterparts. We also considered the effects of different pooling methods and observation time windows and found that the convergence accuracy achieved by convolutional SNNs was comparable to that of convolutional neural networks under the same conditions. Moreover, the hybrid photonic DCRSNN also shows comparable testing accuracy. This work provides new insights into extending the SNN applications in the field of text classification and natural language processing, which is interesting for the resources-restrained scenarios.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.