A Spatial Pyramid Pooling-Based Deep Convolutional Neural Network for the Classification of Electrocardiogram Beats

The deep convolutional neural networks (CNN) have a vast amount of parameters, especially in the fully connected (FC) layers, which have become a bottleneck for real-time applications where processing latency is high due to computational cost. In this paper, we propose to optimize the FC layers in CNN via making it much slimmer. We make analysis of the statistical distribution of the weights in FC layer, and observe each column follows Gaussian distribution. Regression model analysis of the weights of FC layer based on Akaike information criteria and Bayesian information criterion demonstrates that they have Granger causality, which means the columns are correlated and they follow colored Gaussian distribution. Based on this distribution, we derive a CNN design and optimization theorem for FC layers from information theory point of view. The theorem provides two design criteria, rank and singular values. Further, we show that FC layer with weights of colored Gaussian is more efficient than that of white Gaussian. The optimization criteria is singular-values-based, so we apply singular value decomposition to find the maximal singular values and QR to identify the corresponding columns in FC layer. We evaluate our optimization approach to AlexNet and apply the slimmer CNN to ImageNet classification. Simulation results show our approach performs much better than random dropout. Specifically, with only around 28% of weights, the AlexNet could perform as well as the original AlexNet in terms of top one error and top five error.

Deep convolutional neural networks (CNNs) have seen significant growth in medical image classification applications due to their ability to automate feature extraction, leverage hierarchical learning, and deliver high classification accuracy. However, Deep CNNs require substantial computational power and memory, particularly for large datasets and complex architectures. Additionally, optimising the hyperparameters of deep CNNs, although critical for enhancing model performance, is challenging due to the high computational costs involved, making it difficult without access to high-performance computing resources. To address these limitations, this study presents a fast and efficient model that aims to achieve superior classification performance compared to popular Deep CNNs by developing lightweight CNNs combined with the Nonlinear Lévy chaotic moth flame optimiser (NLCMFO) for automatic hyperparameter optimisation. NLCMFO integrates the Lévy flight, chaotic parameters, and nonlinear control mechanisms to enhance the exploration capabilities of the Moth Flame Optimiser during the search phase while also leveraging the Lévy flight theorem to improve the exploitation phase. To assess the efficiency of the proposed model, empirical analyses were performed using a dataset of 2314 brain tumour detection images (1245 images of brain tumours and 1069 normal brain images). The evaluation results indicate that the CNN_NLCMFO outperformed a non-optimised CNN by 5% (92.40% accuracy) and surpassed established models such as DarkNet19 (96.41%), EfficientNetB0 (96.32%), Xception (96.41%), ResNet101 (92.15%), and InceptionResNetV2 (95.63%) by margins ranging from 1 to 5.25%. The findings demonstrate that the lightweight CNN combined with NLCMFO provides a computationally efficient yet highly accurate solution for medical image classification, addressing the challenges associated with traditional deep CNNs.

Medical images classification is a challenging research topic in the field of computer vision, especially when applied to diagnosis of breast cancer (BC). Nowadays, histopathological image is marked as the gold standard for diagnosing BC. However, such diagnosis is heavily dependent on the clinician's experience, which is extremely time consuming and is subjected to human error even for experienced doctors. To address those problems, this paper implements an automated method for distinguishing the benign from the malignant tumor based on a convolutional neural network (CNN). Traditional deep CNN and machine learning methods not only lead to poor performance, but also fail to make full use of the long-term dependence between some key features and image tags. To further meet the high accuracy requirement of diagnosis, according to the characteristics of histopathological images, we propose a novel CNN framework. Firstly, a normal image is augmented to solve the problem about having a limited database. Secondly, we introduce transfer learning to obtain more accurate weight parameters that were pre-trained on the ImageNet. Thirdly, we combine various features extracted by many individual models to obtain comprehensive features. Finally, random forest is introduced to enforce classification. The experimental results show that novel CNN frameworks have better performance compared with individual models, including DenseNet and ResNet. Experimental results are able to prove the effectiveness of our strategy.

Electrocardiography (ECG) is a technique for observing and recording the electrical activity of the human heart. The usage of an ECG signal is common among clinical professionals in the collection of time data for the examination of any rhythmic conditions associated with a subject. The investigation was carried out in order to computerize the assignment by exhibiting the issue using encoder-decoder techniques, creating the information that was simply typical of it, and utilising misfortune appropriation to anticipate standard or anomalous information. On a broad variety of applications such as voice recognition and prediction, the long short-term memory (LSTM) fully connected layer (FCL) and the two convolutional neural networks (CNNs) have shown superior performance over deep learning networks (DLNs). DNNs are suitable for making high points for a more divisible region and CNNs are suitable for reducing recurrence types, LSTMs are appropriate for temporary displays, in the same way as CNNs are appropriate for reducing recurrence types. The CNN, LSTM, and DNN algorithms are acceptable for viewing. The complementarity of DNNs, CNNs, and LSTMs was investigated in this research by bringing them all together under the single architectural company. The researchers got the ECG data from the MIT-BIH arrhythmia database as a result of the investigation. Our results demonstrate that the approach proposed may expressively describe ECG series and identify abnormalities via scores that outperform existing supervised and unsupervised methods in both the short term and long term. The LSTM network and FCL additionally demonstrated that the unbalanced datasets associated with the ECG beat detection problem could be consistently resolved and that they were not susceptible to the accuracy of ECG signals. It is recommended that cardiologists employ the unique technique to aid them in performing reliable and impartial interpretation of ECG data in telemedicine settings.

The convolutional layers of standard convolutional neural networks (CNNs) are equivariant to translation. However, the convolution and fully-connected layers are not equivariant or invariant to other affine geometric transformations. Recently, a new class of CNNs is proposed in which the conventional layers of CNNs are replaced with equivariant convolution, pooling, and batch-normalization layers. The final classification layer in equivariant neural networks is invariant to different affine geometric transformations such as rotation, reflection and translation, and the scalar value is obtained by either eliminating the spatial dimensions of filter responses using convolution and down-sampling throughout the network or average is taken over the filter responses. In this work, we propose to integrate the orthogonal moments which gives the high-order statistics of the function as an effective means for encoding global invariance with respect to rotation, reflection and translation in fully-connected layers. As a result, the intermediate layers of the network become equivariant while the classification layer becomes invariant. The most widely used Zernike, pseudo-Zernike and orthogonal Fourier-Mellin moments are considered for this purpose. The effectiveness of the proposed work is evaluated by integrating the invariant transition and fully-connected layer in the architecture of group-equivariant CNNs (G-CNNs) on rotated MNIST and CIFAR10 datasets.

Recent works about convolutional neural networks (CNN) show breakthrough performance on various tasks. However, most of them only use the features extracted from the topmost layer of CNN instead of leveraging the features extracted from different layers. As the first group which explicitly addresses utilizing the features from different layers of CNN, we propose cross-layer CNN features which consist of the features extracted from multiple layers of CNN. Our experimental results show that our proposed cross-layer CNN features outperform not only the state-of-the-art results but also the features commonly used in the traditional CNN framework on three tasks — artistic style, artist, and architectural style classification. As shown by the experimental results, our proposed cross-layer CNN features achieve the best known performance on the three tasks in different domains, which makes our proposed cross-layer CNN features promising solutions for generic tasks.

Cervical cancer is the fourth most prevalent disease among women. Prompt diagnosis and its management can significantly improve patients’ survival rates. Therefore, routine screening for cervical cancer is of paramount importance. Herein, we explore the potential of a deep learning model to automatically distinguish abnormal cells from normal cells. The ThinPrep cytologic test dataset was collected from the fourth central hospital of Baoding city, China. Based on the dataset, four classification models were developed. The first model was a 10-layer convolutional neural network (CNN). The second model was an advancement of the first model equipped with a spatial pyramid pooling (SPP) layer (CNN + SPP) to treat cell images based on their sizes. Based on the first model, the third model replaced the CNN layers with the inception module (CNN + Inception). However, the fourth model incorporated both the SPP layer and the inception module into the first model (CNN + inception + SPP). The performances of the four models are estimated and compared by using the same testing data and evaluation index. The testing results demonstrated that the fourth model yields the best performance. Moreover, the area under the curve (AUC) for module four was 0.997.

Fine-grained image classification is one of the key basic tasks of computer vision. The appearance of traditional deep convolutional neural network (DCNN) combined with attention mechanism can focus on partial and local features of fine-grained images, but it still lacks the consideration of the embedding mode of different attention modules in the network, leading to the unsatisfactory result of classification model. To solve the above problems, three different attention mechanisms are introduced into the DCNN network (like ResNet, VGGNet, etc.), including SE, CBAM and ECA modules, so that DCNN could better focus on the key local features of salient regions in the image. At the same time, we adopt three different embedding modes of attention modules, including serial, residual and parallel modes, to further improve the performance of the classification model. The experimental results show that the three attention modules combined with three different embedding modes can improve the performance of DCNN network effectively. Moreover, compared with SE and ECA, CBAM has stronger feature extraction capability. Among them, the parallelly embedded CBAM can make the local information paid attention to by DCNN richer and more accurate, and bring the optimal effect for DCNN, which is 1.98% and 1.57% higher than that of original VGG16 and Resnet34 in CUB-200-2011 dataset, respectively. The visualization analysis also indicates that the attention modules can be easily embedded into DCNN networks, especially in the parallel mode, with stronger generality and universality.

Convolutional Neural Networks (CNNs) have become a dominant solution for Electrocardiogram (ECG) beat classification, owing to their superior feature extraction capabilities. However, traditional CNN-based accelerators often encounter limitations such as large parameter counts, limited configurability, and inefficient hardware utilization. In this work, we propose a lightweight one-dimensional CNN (1D-CNN) inspired by the InceptionNet architecture, specifically designed for efficient and accurate ECG beat classification. The model integrates customized Inception blocks with parallel convolutional layers of varying kernel sizes to capture multi-scale temporal features. To reduce model complexity and enhance efficiency, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$1 \times 1$</tex> convolutions are used for dimensionality reduction, resulting in a compact network with only 7,740 parameters. Additionally, we present a hardware accelerator that exploits spatial data reuse across multiple processing units and employs dual-port RAM for concurrent data access, improving throughput and memory efficiency. When implemented on hardware with parallel and pipelined execution at 100 MHz, the system achieves a classification accuracy of 99.32%, sensitivity of 98.50%, specificity of 99.58%, and a positive predictive value (PPV) of 98.30% across five ECG classes with the MIT-BIH database, demonstrating its suitability for real-time deployment in resource-constrained environments.

The three main modules of autonomous vehicles, i.e., sensing, decision making, and motion controlling, have been studied separately in most existing works on autonomous driving, which overlook the correlations among these modules, leading to a result of unsatisfactory performance. In this paper, we propose a novel scheme that first tactfully processes the sensing data, then jointly learns and optimizes the decision-making and motion-controlling using reinforcement learning (RL). Specifically, the proposed scheme designs a novel state representation mechanism, where the sensing data goes through the attention layer and the convolutional neural network (CNN) layer sequentially. The attention layer focuses on extracting the most important local information and then CNN layer takes a broad view to comprehensively consider the global information for a better representation. Furthermore, the proposed scheme jointly learns decision-making and motion-controlling, therefore, the relevance of these two modules is implicitly considered, which helps to achieve a better autonomous driving policy. Extensive simulation results show that the proposed scheme is better than classic control methods and some RL methods in terms of safety, velocity, etc. We also demonstrate the respective functions of the attention layer and the CNN layer through ablation studies. Finally, we construct a traffic scene with a real autonomous vehicle, and verified the feasibility of the proposed scheme.

The occlusion in one of the coronary arteries of the heart leads to the cardiac ailment, myocardial infarction (MI). The localization of MI based on the investigation of the morphology of the multi-lead electrocardiogram (ECG) is the initial task for the diagnosis of this ailment. In this paper, the multiscale convolutional neural network is proposed for the automated localization of MI ailment from multi-lead electrocardiogram (ECG) beats. The Fourier-Bessel (FB) series expansion based empirical wavelet transform (EWT) with fixed order ranges is introduced for the multiscale analysis of multi-lead ECG beat. The FB spectrum of each lead ECG beat is segregated into contiguous segments using the fixed order ranges. Furthermore, the order ranges from these contiguous segments are used to design an empirical wavelet filter bank for the extraction of subband signals from each lead ECG beat. The convolutional neural network (CNN) is used for the classification of various categories of MI as anterior MI (AMI), anterio-lateral MI (ALMI), anterio-septal MI (ASMI), inferior MI (IMI), inferio-lateral MI (ILMI), inferio-posterio-lateral MI (IPLMI) and normal sinus rhythm (NSR). The experimental results reveal that the lower-order range subband signal coupled with CNN attains higher average accuracy values of 99.92%, 99.34%, 99.95%, 99.95%, 99.91%, and 99.86% respectively, for AMI, ALMI, ASMI, IMI, ILMI, and IPLMI classes. The subband signal of multi-lead ECG beats with order range of [1-26] is highly affected during various categories of MI heart disease, and this band signal has higher performance as compared to the existing MI localization approaches.

Onboard cable is important electrical equipment of an electric locomotive. In order to effectively evaluate the aging state of the cable, this article proposes a method to evaluate the aging state of the cable based on a new method of dielectric spectrum data processing and a new structure model of recognition algorithm. First, the log-min-max gray-scale (LMMGS) method is proposed, which transforms the dielectric parameters based on the frequency-domain dielectric spectrum (FDS) test into the gray matrix. The correlation among different dielectric parameters is retained, and the interference of their order of magnitude difference on feature extraction is reduced. Then, the new structure of spatial pyramid pooling (SPP) is improved by rectangular mesh dividing named Rectangular-SPP, which increases the number of divisible layers of the spatial pyramid, and it is used to optimize convolutional neural network (CNN), so as to remove the limitation of recognition algorithm on the size of input data and expand the scale of feature extraction. Finally, a set of methods for evaluating the aging state of onboard cable is proposed. The experimental results show that this method can effectively distinguish the insulation aging state of onboard cable. Meanwhile, the proposed Rectangular-SPP-CNN algorithm has a better performance than the conventional machine learning algorithms.

Millions of clinicians rely daily on automated preliminary electrocardiogram (ECG) interpretation. Critical comparisons of machine learning-based automated analysis against clinically accepted standards of care are lacking. To use readily available 12-lead ECG data to train and apply an explainability technique to a convolutional neural network (CNN) that achieves high performance against clinical standards of care. This cross-sectional study was conducted using data from January 1, 2003, to December 31, 2018. Data were obtained in a commonly available 12-lead ECG format from a single-center tertiary care institution. All patients aged 18 years or older who received ECGs at the University of California, San Francisco, were included, yielding a total of 365 009 patients. Data were analyzed from January 1, 2019, to March 2, 2021. A CNN was trained to predict the presence of 38 diagnostic classes in 5 categories from 12-lead ECG data. A CNN explainability technique called LIME (Linear Interpretable Model-Agnostic Explanations) was used to visualize ECG segments contributing to CNN diagnoses. Area under the receiver operating characteristic curve (AUC), sensitivity, and specificity were calculated for the CNN in the holdout test data set against cardiologist clinical diagnoses. For a second validation, 3 electrophysiologists provided consensus committee diagnoses against which the CNN, cardiologist clinical diagnosis, and MUSE (GE Healthcare) automated analysis performance was compared using the F1 score; AUC, sensitivity, and specificity were also calculated for the CNN against the consensus committee. A total of 992 748 ECGs from 365 009 adult patients (mean [SD] age, 56.2 [17.6] years; 183 600 women [50.3%]; and 175 277 White patients [48.0%]) were included in the analysis. In 91 440 test data set ECGs, the CNN demonstrated an AUC of at least 0.960 for 32 of 38 classes (84.2%). Against the consensus committee diagnoses, the CNN had higher frequency-weighted mean F1 scores than both cardiologists and MUSE in all 5 categories (CNN frequency-weighted F1 score for rhythm, 0.812; conduction, 0.729; chamber diagnosis, 0.598; infarct, 0.674; and other diagnosis, 0.875). For 32 of 38 classes (84.2%), the CNN had AUCs of at least 0.910 and demonstrated comparable F1 scores and higher sensitivity than cardiologists, except for atrial fibrillation (CNN F1 score, 0.847 vs cardiologist F1 score, 0.881), junctional rhythm (0.526 vs 0.727), premature ventricular complex (0.786 vs 0.800), and Wolff-Parkinson-White (0.800 vs 0.842). Compared with MUSE, the CNN had higher F1 scores for all classes except supraventricular tachycardia (CNN F1 score, 0.696 vs MUSE F1 score, 0.714). The LIME technique highlighted physiologically relevant ECG segments. The results of this cross-sectional study suggest that readily available ECG data can be used to train a CNN algorithm to achieve comparable performance to clinical cardiologists and exceed the performance of MUSE automated analysis for most diagnoses, with some exceptions. The LIME explainability technique applied to CNNs highlights physiologically relevant ECG segments that contribute to the CNN's diagnoses.

Age estimation is a fascinating field of study because its implementation can be utilized in a variety of contexts. Currently, the Convolutional Neural Network (CNN) technique is frequently utilized to solve the age estimation problem. However, this method has one disadvantage, namely a high computational cost that lengthens the training process. This paper proposes an alternative method for resolving age estimation issues. To overcome the shortcomings of the backpropagation method, the proposed method employs the Extreme Learning Machine (ELM) algorithm as a fully-connected CNN layer. ResNet50 and VGG16 are the CNN architectures employed in this study. Using the UTKFace dataset, the proposed method was trained and evaluated. Using ELM as a fully-connected layer on CNN provided significantly faster training time performance than the fully-connected layer trained with the backpropagation technique. In addition, the proposed combination of CNN and ELM can provide age classification performance that is competitive with state-of-the-art method for age estimation cases.

In deep convolutional neural networks (DCNNs), model size and computation complexity are two important factors governing throughput and energy efficiency when deployed to hardware for inference. Recent works on compact DCNNs as well as pruning methods are effective, yet with drawbacks. For instance, more than half the size of all MobileNet models lies in their last two layers, mainly because compact separable convolution (CONV) layers are not applicable to their last fully connected (FC) layers. Also, in pruning methods, the compression is gained at the expense of irregularity in the DCNN architecture, which necessitates additional indexing memory to address nonzero weights, thereby increasing memory footprint, decompression delays, and energy consumption. In this article, we propose cyclic sparsely connected (CSC) architectures, with memory/computation complexity of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mathcal {O}(N\log {}N)$ </tex-math></inline-formula> , where <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$N$ </tex-math></inline-formula> is the number of nodes/channels given a DCNN layer that, contrary to compact depthwise separable layers, can be used as an overlay for both FC and CONV layers of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mathcal {O}(N^{2})$ </tex-math></inline-formula> . Also, contrary to pruning methods, CSC architectures are structurally sparse and require no indexing due to their cyclic nature. We show that both standard convolution and depthwise convolution layers are special cases of the CSC layers, whose mathematical function, along with FC layers, can be unified into one single formulation and whose hardware implementation can be carried out under one arithmetic logic component. We examine the efficacy of the CSC architectures for compression of LeNet, AlexNet, and MobileNet DCNNs with precision ranging from 2 to 32 bits. More specifically, we surge upon the compact 8-bit quantized 0.5 MobileNet V1 and show that by compressing its last two layers with CSC architectures, the model is compressed by <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\sim 1.5\times $ </tex-math></inline-formula> with a size of only 873 kB and little accuracy loss. Finally, we design a configurable hardware that implements all types of DCNN layers including FC, CONV, depthwise, CSC-FC, and CSC-CONV indistinguishably within a unified pipeline. We implement the hardware on a tiny Xilinx field-programmable gate array (FPGA) for total on-chip processing of the compressed MobileNet that, compared to the related work, has the highest Inference/J while utilizing the smallest FPGA.