Performance Of Convolutional Neural Networks Research Articles

Application of light-weighted CNN for diagnosis of internal concrete defects using hammering sound

ABSTRACT In the past, diagnoses of the internal defects of ageing concrete members by sound tests have relied on the ability and experience of the inspector. However, with the development of deep neural networks, building diagnosis tools without the subjective judgements of humans is feasible. In this study, sound samples are collected from a tapping sound test on concrete slab specimens with internally embedded artificial defects. Spectrograms are generated from the sound samples and used to train a light-weighted convolutional neural network (CNN). Thereafter, the performance of the proposed CNN models at diagnosing the defects of the concrete slabs are examined and compared with convolutional autoencoder (CAE) and support vector regression (SVR). All the models performed well at distinguishing between healthy and defective concrete. Whereas noise can impair the performance of CAE and SVR, CNN is less affected, particularly when classifying sound samples with distinctive features. CNN also has other advantages: the physical area of the artificial defect can be output directly from CNN, and CNN performance is also less affected by the depth of the artificial defect and the rebar embedded in the concrete. At increased depths, the CNN is still able to identify the defect area.

Anatomical sites identification in both ordinary and capsule gastroduodenoscopy via deep learning

Anatomical sites recognition is a basic requirement for gastroenterologists. But there is not a unified framework for anatomical sites identification in both ordinary and capsule endoscopes. Deep learning (DL), especially vision transformer (ViT), is promising in the field of medical imaging, however, the performance of them is not comprehensively compared. The retrospective cohort study included 322 patients who visited Friendship hospital for capsule endoscopy and 556 patients who visited Minhang hospital for ordinary endoscopy. Convolutional neural network (CNN) and two types of ViT (B/16 and L/32) that was trained to identify qualified and low-quality images (the first model), then the qualified images were used to train the second model for distinguishing different anatomical sites. 62,850 images from capsule endoscopy and 17,434 images from ordinary endoscopy were included in developing models. In internal cross-validation, CNN achieved average area under receiver operating characteristic curve (AUROC) of 0.9844 (95 % confidence interval [CI] 0.9640–0.9960) in distinguishing qualified and low-quality images, and 0.9251 average accuracy (95 % CI 0.9133–0.9369) in distinguishing different anatomical sites. Besides, the performance of ViT did not surpass the performance of CNN. 18,636 images from 355 patients who received capsule endoscopy and 15,949 images of 501 patients who received ordinary endoscopy were prospectively collected. The AUROC of CNN reached 0.8715 (95 % CI 0.8674–0.8754) in the first model, and 0.8376 accuracy (95 % CI 0.8336–0.8414) for the second model, respectively. The performance of CNN is better than ViT with the same hyperparameter setting for sifting out the unqualified images and distinguishing anatomical sites effectively.

Convolutional Neural Network Activation Function Performance on Image Recognition of The Batak Script

Deep Learning is a sub-set of Machine learning, Deep Learning is widely used to solve problems in various fields. One of the popular deep learning architectures is The Convolutional Neural Network (CNN), CNN has a layer that transforms feature extraction automatically so it is widely used in image recognition. However, CNN's performance using the tanh function is still relatively low, therefore it is necessary to select the right activation function to improve accuracy performance. This study analyzes the use of the activation function in image recognition of the Batak script. The result of this study is that the CNN model using the ReLU and eLU functions produces the highest accuracy compared to the CNN model using the tanh function. The CNN model using eLU produces the best accuracy performance in the training process, which is 99.71% with an error value of 0.0108. Meanwhile, in the testing process, the highest accuracy value is generated by the CNN Model using the ReLU function with an accuracy of 94.11%, an error value of 0.3282, a precision value of 0.9411, a recall of 0.9411, and an f1-score of 0.9416.

Wavelet-Based Dual-Task Network.

In image processing, wavelet transform (WT) offers multiscale image decomposition, generating a blend of low-resolution approximation images and high-resolution detail components. Drawing parallels to this concept, we view feature maps in convolutional neural networks (CNNs) as a similar mix, but uniquely within the channel domain. Inspired by multitask learning (MTL) principles, we propose a wavelet-based dual-task (WDT) framework. This novel framework employs WT in the channel domain to split a single task into two parallel tasks, thereby reforming traditional single-task CNNs into dynamic dual-task networks. Our WDT framework integrates seamlessly with various popular network architectures, enhancing their versatility and efficiency. It offers a more rational approach to resource allocation in CNNs, balancing between low-frequency and high-frequency information. Rigorous experiments on Cifar10, ImageNet, HMDB51, and UCF101 validate our approach's effectiveness. Results reveal significant improvements in the performance of traditional CNNs on classification tasks, and notably, these enhancements are achieved with fewer parameters and computations. In summary, our work presents a pioneering step toward redefining the performance and efficiency of CNN-based tasks through WT.

Local and Global Feature-Aware Dual-Branch Networks for Plant Disease Recognition.

Accurate identification of plant diseases is important for ensuring the safety of agricultural production. Convolutional neural networks (CNNs) and visual transformers (VTs) can extract effective representations of images and have been widely used for the intelligent recognition of plant disease images. However, CNNs have excellent local perception with poor global perception, and VTs have excellent global perception with poor local perception. This makes it difficult to further improve the performance of both CNNs and VTs on plant disease recognition tasks. In this paper, we propose a local and global feature-aware dual-branch network, named LGNet, for the identification of plant diseases. More specifically, we first design a dual-branch structure based on CNNs and VTs to extract the local and global features. Then, an adaptive feature fusion (AFF) module is designed to fuse the local and global features, thus driving the model to dynamically perceive the weights of different features. Finally, we design a hierarchical mixed-scale unit-guided feature fusion (HMUFF) module to mine the key information in the features at different levels and fuse the differentiated information among them, thereby enhancing the model's multiscale perception capability. Subsequently, extensive experiments were conducted on the AI Challenger 2018 dataset and the self-collected corn disease (SCD) dataset. The experimental results demonstrate that our proposed LGNet achieves state-of-the-art recognition performance on both the AI Challenger 2018 dataset and the SCD dataset, with accuracies of 88.74% and 99.08%, respectively.

Data augmentation for 3D seismic fault interpretation using deep learning

Manual seismic interpretation of faults is a tedious and complicated process, which is prone to human error and bias. A semi-automatic approach for interpreting faults on seismic data is to use attributes that highlight discontinuities. These methods need to be optimized by the interpreter, thus constantly rely on the interpreter's knowledge. Recently, Machine Learning (ML) techniques in general and Convolutional Neural Networks (CNN) as part of Deep Neural Networks (DNN), in particular, have been used to detect and image faults on seismic data in order to make the process more automated and accurate. Supervised CNN learns and evolves from manually annotated or labeled fault interpretations.In this study, we have applied supervised 2D CNN to image faults on seismic data through binary segmentation. The study was performed on 3D marine seismic surveys in off-shore Norway collected from three separate locations along the Norwegian Continental Shelf, utilizing Efficient UNET CNN architecture. We applied data augmentation (geometric transformations) and hyperparameter tuning to improve the learning process and performance of the deep learning algorithm.We used noise content, acquisition type, imaging type, and fault scale as the main criteria to select seismic surveys. The application of data augmentation to the training and testing data generally led to improvement in the performance of 2D CNN on fault predictions, although the amount of improvement varied with respect to different surveys. The initial 2D CNN fault prediction improvement mainly relied on the quality, and size of faults present in the 3D seismic volumes. Further, improvement was achieved by the adjustment of certain hyperparameters affecting the training and testing process of the 2D CNN. However, we found little to no improvement on one seismic volume containing high levels of noise.

An Adaptive Spatio-temporal Global Sampling for Presentation Attack Detection

Without developing dedicated countermeasures, facial biometric systems can be spoofed with printed photos, replay attacks, silicone masks, or even a 3D mask of a targeted person. Thus, the threat of presentation attacks needs to be addressed to strengthen the security of the biometric systems. Since a 2D convolutional neural network (CNN) captures static features from video frames, the camera motion might hinders the performance of modern CNNs for video-based presentation attack detection (PAD). Inspired by the egomotion theory, we introduce an adaptive spatiotemporal global sampling (ASGS) technique to compensate the camera motion and use the resulting estimation to encode the appearance and dynamics of the video sequences into a single RGB image. This is achieved by adaptively splitting the video into small segments and capturing their global motion within each segment. The proposed global motion is estimated based on four key steps: dense sampling, FREAK feature extraction and matching, similarity transformation, and aggregation function. This allows using deep models pre-trained on images for video-based PAD detection. Moreover, the interpretation of ASGS reveals that the most important parts for supporting the decision on PAD are consistent with motion cues associated with the artifacts, i.e., hand movement, material reflection, and expression changes. Extensive experiments on four standard face PAD databases demonstrate its effectiveness and encourage further study in this domain.

Face Detection and Counting: Recent Advances and Future Research Directions

In recent years, face detection and counting have undergone a remarkable transfor-mation due to the emergence of deep learning techniques, particularly Convolutional Neural Networks (CNNs). These methods have surpassed traditional computer vision approaches, benefiting from vast datasets and robust computational resources. CNN’s, the cornerstone of contemporary face recognition systems, excel in accurately identifying faces across diverse conditions and environments. The ascendancy of CNNs in face detection stems from their capacity to autonomously learn hierarchical features from raw pixel data, obviating the need for manual feature engineering. Consequently, they yield resilient systems capable of accommodating real-world variations like lighting, pose, expression, and occlusion. Their scalability enables the efficient processing of extensive image databases, rendering them invaluable for applications requiring face counting amidst crowded scenes. Nonetheless, challenges persist, particularly in deploying these systems on resources-constrained devices. The computational complexity and storage demand of deep CNNs necessitate ongoing exploration of lightweight network architectures that balances accuracy with reduced computational footprint. The historical Violajones algorithm, while foundational, has been eclipsed by the superior performance of CNNs. Deep learning’s prowess lies in its adaptability to various tasks and its capacity to continuously refine itself with more data. By learning features directly from data, CNNs excel in capturing intricate patterns crucial for precise face detection and counting. Deep learning, especially through CNNs, has revolutionized face detection and counting by delivering unparal-leled accuracy and robustness. However, ongoing efforts are crucial to address challenges in the efficiency and accessibility of these systems for broader deployment across various applications and devices.

A Hybrid DL Architecture for Improved Generalizability with Self-Adaptive Jaya Optimizer for Diabetic Retinopathy

Anomalies such as microaneurysms, exudates, and hemorrhages are diagnosed by retinal vessel segmentation to identify different phases of Diabetic Retinopathy (DR); several typical ways to detect Hard Exudates (HE) in retinal images have been used to determine the severity of diabetes. This work proposes a hybrid approach to better segment such abnormal cells. It optimizes the learning parameters of a Convolutional Neural Network (CNN) with a nature-inspired optimizer, the Self-Adaptive Jaya Optimization Algorithm (SAJOA). The SAJOA helps in avoiding trapping in local minima. It uses the SAJOA to steer the search for obtaining near-optimal parameter values for the CNN training. We choose the learning parameters such as batch size, dropout rate, learning rate, etc. The batch size affects the batch normalization performance, and the dropout rate involves the regularization property of the model. The near-optimal values of these two parameters help improve the generalization ability of the deep learning architecture. Thus, this hybrid approach improves the CNN performance and generalizability, leading to precise segmentation of unseen data, which aids in diagnosing diabetic retinopathy. The results of many trials were evaluated against the DRIVE dataset; the proposed method achieved accuracy and F1 score on average, higher by 7 to 8% than the state-of-the-art methods.

The integration of machine learning and IoT for the early detection of tomato leaf disease in real-time

The impact of climate change, pests, and inadequate agricultural practices on crop health is becoming a growing concern. It is estimated that 20-40% of global crop yield is adversely affected by pests and diseases. This has a direct negative impact on food security and nutritional well-being, as staple cereal crops (such as rice, maize, and wheat) and tuber crops (like potato, onion, and tomato) are affected. Advancements in Artificial Intelligence (AI), Computer Vision (CV), and IoT have a significant influence on reducing crop losses by detecting crop diseases at early stages. The primary focus of this research is to develop a real-time system that can detect tomato crop diseases at an early stage by integrating Machine Learning (ML) algorithms and IoT. The most common tomato leaf diseases, such as Tomato Mosaic Virus (TMV), Tomato Bacterial Leaf Spot (TBLS), Tomato Early Blight (TEB), and Tomato Late Blight (TLB), are considered in this work. The hybrid discriminative feature space is derived from the integration of low- and high-level features via Convolution Neural Network (CNN) Layers. The 3-stage Stacked Deep Convolutional Autoencoder is used to optimize the CNN performance by reducing computation complexity. The proposed model is implemented on the Plantvillage benchmark dataset and achieves the highest recognition accuracy of 95.6% for the 5-class problem using 5-fold cross-validation.

Automatic liver segmentation and assessment of liver fibrosis using deep learning with MR T1-weighted images in rats

ObjectivesTo validate the performance of nnU-Net in segmentation and CNN in classification for liver fibrosis using T1-weighted images. Materials and methodsIn this prospective study, animal models of liver fibrosis were induced by injecting subcutaneously a mixture of Carbon tetrachloride and olive oil. A total of 99 male Wistar rats were successfully induced and underwent MR scanning with no contrast agent to get T1-weighted images. The regions of interest (ROIs) of the whole liver were delineated layer by layer along the liver edge by 3D Slicer. For segmentation task, all T1-weighted images were randomly divided into training and test cohorts in a ratio of 7:3. For classification, images containing the hepatic maximum diameter of every rat were selected and 80% images of no liver fibrosis (NLF), early liver fibrosis (ELF) and progressive liver fibrosis (PLF) stages were randomly selected for training, while the rest were used for testing. Liver segmentation was performed by the nnU-Net model. The convolutional neural network (CNN) was used for classification task of liver fibrosis stages. The Dice similarity coefficient was used to evaluate the segmentation performance of nnU-Net. Confusion matrix, ROC curve and accuracy were used to show the classification performance of CNN. ResultsA total of 2628 images were obtained from 99 Wistar rats by MR scanning. For liver segmentation by nnU-Net, the Dice similarity coefficient in the test set was 0.8477. The accuracies of CNN in staging NLF, ELF and PLF were 0.73, 0.89 and 0.84, respectively. The AUCs were 0.76, 0.88 and 0.79, respectively. ConclusionThe nnU-Net architecture is of high accuracy for liver segmentation and CNN for assessment of liver fibrosis with T1-weighted images.

Comparison of state-of-the-art deep learning architectures for detection of freezing of gait in Parkinson's disease.

Freezing of gait (FOG) is one of the most debilitating motor symptoms experienced by patients with Parkinson's disease (PD). FOG detection is possible using acceleration data from wearable sensors, and a convolutional neural network (CNN) is often used to determine the presence of FOG epochs. We compared the performance of a standard CNN for the detection of FOG with two more complex networks, which are well suited for time series data, the MiniRocket and the InceptionTime. We combined acceleration data of people with PD across four studies. The final data set was split into a training (80%) and hold-out test (20%) set. A fifth study was included as an unseen test set. The data were windowed (2 s) and five-fold cross-validation was applied. The CNN, MiniRocket, and InceptionTime models were evaluated using a receiver operating characteristic (ROC) curve and its area under the curve (AUC). Multiple sensor configurations were evaluated for the best model. The geometric mean was subsequently calculated to select the optimal threshold. The selected model and threshold were evaluated on the hold-out and unseen test set. A total of 70 participants (23.7 h, 9% FOG) were included in this study for training and testing, and in addition, 10 participants provided an unseen test set (2.4 h, 11% FOG). The CNN performed best (AUC = 0.86) in comparison to the InceptionTime (AUC = 0.82) and MiniRocket (AUC = 0.76) models. For the CNN, we found a similar performance for a seven-sensor configuration (lumbar, upper and lower legs and feet; AUC = 0.86), six-sensor configuration (upper and lower legs and feet; AUC = 0.87), and two-sensor configuration (lower legs; AUC = 0.86). The optimal threshold of 0.45 resulted in a sensitivity of 77% and a specificity of 58% for the hold-out set (AUC = 0.72), and a sensitivity of 85% and a specificity of 68% for the unseen test set (AUC = 0.90). We confirmed that deep learning can be used to detect FOG in a large, heterogeneous dataset. The CNN model outperformed more complex networks. This model could be employed in future personalized interventions, with the ultimate goal of using automated FOG detection to trigger real-time cues to alleviate FOG in daily life.

Image Splicing Forgery Detection using Error Level Analysis and CNN

The issue of image forgery through splicing has become increasingly relevant in the current digital era. Splicing involves the manipulation of images by combining parts of two or more different images to create a deceptive composite image. This technique can be employed for various purposes, including the dissemination of false information, damaging someone's reputation, or even creating confusion in specific contexts. Several techniques used to detect splicing involve statistical analysis, color analysis, and texture analysis. Additionally, artificial intelligence developments, such as deep learning, have been applied to enhance detection capabilities. In this study, we employed a Convolutional Neural Network (CNN) model to identify image deviations caused by splicing. Optimization was performed on the convolutional layers of the model to improve CNN performance. The integration of Error Level Analysis (ELA) was also implemented to aid in identifying splicing forgeries, where portions of one image are combined with parts of another. Areas that have undergone splicing may exhibit noticeable differences in error levels. The dataset utilized for this research was sourced from DVMM and CUISDE. The validation accuracy results for our CNN model before incorporating ELA were 61% for DVMM and 74% for CUISDE. After adding ELA, the CNN model demonstrated improved detection accuracy, achieving validation rates of 72% for DVMM and 71% for CUISDE.

Degradation model and attention guided distillation approach for low resolution face recognition

Deep convolution neural networks (CNN) have shown their efficacy in face recognition tasks due to their ability to extract highly discriminant face representations from face images. On high-resolution benchmark datasets, outstanding identification and verification results have been achieved. However, the performance of these algorithms is significantly degraded when tested on low-resolution (LR) images such as those captured from surveillance cameras. A straightforward solution to this problem is to use both high-resolution (HR) images and corresponding down-sampled LR images during training. Although this strategy improves the performance for LR images, it has some limitations. First, there is a significant difference between down-sampled LR images and LR images from surveillance cameras, leading to performance saturation at an earlier stage. Another limitation is the deterioration in the performance of HR images. In this work, solutions to both these limitations are proposed. A degradation model is proposed that synthesizes LR images from corresponding HR, emulating the real-world degradation effects in synthetic data, thus enabling the face recognition system to tolerate various blurry and noisy effects. To address the deterioration in the performance of HR images, an attention-guided distillation is proposed, which utilizes attention maps from convolutional layers in combination with deep features to transfer informative HR features from teacher to student network. The attention maps from the teacher network guide the student network to a better optimum and produce resolution robust face representations. The results of the proposed approach on the popular LR datasets like SCface, Coxface, and PaSC show that it outperforms the recent state-of-the-art (SOTA) techniques by a significant margin demonstrating its effectiveness for different cross-resolution scenarios.

Image Classification Using Deep Learning

- Image classification is an important topic of study in the field of image processing nowadays and is a popular area of research. By providing the computer with data to learn from, image categorization was created to close the gap between computer vision and human vision. In this paper, the methods for categorising images using traditional machine learning and deep learning are compared and investigated. This study employs a tensor flow framework and convolutional neural networks to classify images. This paper implements CNN in binary classification and multi-class classification for object identification and analyses the performance of well-known convolutional neural networks (CNNs). We built five unique image datasets on our own for multiclass classification, using the dog vs. cat dataset for binary classification. For our investigation, we classified the photos using a separate machine learning model and then classified them again using CNN because evaluating CNN's performance on a single data set hides its actual potential and limits. Additionally, trained CNNs perform very differently across various categories of objects, and we will thus talk about some potential causes. Key Words: Image classification, python, Deep Learning, Tensor flow, Convolutional Neural Network, Open CV, NLP

Enhanced negation handling for sentiment analysis on Twitter using deep neural networks

<span>Sentiment analysis is a tool to identify and measure the emotion in a piece of text. Negation handling is an important aspect of natural language processing (NLP) for Twitter data. It is a process of correctly interpreting the sentences containing negation words, such as, "never", "no", "neither" and so on. Negation words are used in machine learning to express negative sentiment or indicate the absence of something. In this article, a negation handling technique using deep learning models. Artificial neural networks (ANNs) and convolutional neural networks (CNNs) for classification is proposed. The system is evaluated on SemEval-2017 dataset. The classification performance is improved by using ANN and CNN on the negative tweets. The study aims to improve the classification accuracy by considering negation words in the text. The paper compares the performance of ANNs and CNNs in handling negation words and evaluates them on the tweets data. This study provides insights into the effectiveness of using deep learning techniques for negation handling in sentiment analysis and highlights the importance of considering negation words in the text for improved sentiment analysis performance. The proposed negation strategy attains a superior performance accuracy over machine learning models by preventing misclassified tweets.</span>

Deriving the Terrestrial Water Storage Anomaly From GRACE Spherical Harmonic Coefficients Using a Convolutional Neural Network

AbstractTerrestrial water storage anomaly (TWSA), derived from Gravity Recovery and Climate Experiment (GRACE) satellites, has been widely used in hydrology studies. The inversion is commonly achieved by truncating and filtering spherical harmonic coefficients (SHC), whereby the result is characterized by leakage error and low resolution. It remains unclear whether machine learning methods can help resolve this challenging issue. In this study, we present a convolutional neural network (CNN) approach to correct TWSA from GRACE SHC by leveraging the knowledge of the leakage effect determined from global hydrological models (GHMs) and land surface models (LSMs). The CNN approach is implemented in three representative regions in China, that is, the human‐impacted Haihe River Basin, the nature‐impacted Yangtze River Basin and the model‐limited Tibetan Plateau. The results show the following: (a) The recovery performance of CNN at the basin scale is better than that at the grid scale, and the grid‐scale recovery is significantly influenced by the spatial heterogeneity of TWSA and the input GHM/LSMs; (b) The more accurate the GHM/LSMs used for training, the better the recovery performance of CNN; and (c) The trained model retains comparable performance in deriving the TWSA time series from GRACE SHC when compared to that derived from other methods (i.e., scaling factor and mass concentration solutions) with average r = 0.90 and RMSE = 21.50 mm. This study highlights the potential of machine learning to supplement conventional correction methods when deriving the TWSA from GRACE SHC by utilizing signal restoration knowledge learned from multiple accurate GHMs/LSMs.

Time Reversal vs. Integration of Time Reversal with Convolution Neural Network in Diagnosing Partial Discharge in Power Transformer

Partial discharge (PD) is a common issue in power transformers that can lead to catastrophic failures if left undetected. Time reversal (TR) is a well-known technique in signal processing that can reconstruct signals by reversing the direction of time. The paper investigates the use of time reversal and the integration of time reversal with convolution neural networks (CNNs) for diagnosing PD in power transformers. We compare the performance of these techniques on a dataset of PD signals collected from power transformers. We propose a novel method of using time reversal as a pre-processing step to improve the accuracy of CNNs on noisy or distorted signals. Our experimental results demonstrate that this approach can significantly enhance the performance of CNNs on various datasets, including speech, audio, and image datasets. This paper provides a novel approach to signal processing and demonstrates the potential of time reversal as a pre-processing step in CNNs.

DM-CNN: Dynamic Multi-scale Convolutional Neural Network with uncertainty quantification for medical image classification

Convolutional neural network (CNN) has promoted the development of diagnosis technology of medical images. However, the performance of CNN is limited by insufficient feature information and inaccurate attention weight. Previous works have improved the accuracy and speed of CNN but ignored the uncertainty of the prediction, that is to say, uncertainty of CNN has not received enough attention. Therefore, it is still a great challenge for extracting effective features and uncertainty quantification of medical deep learning models In order to solve the above problems, this paper proposes a novel convolutional neural network model named DM-CNN, which mainly contains the four proposed sub-modules : dynamic multi-scale feature fusion module (DMFF), hierarchical dynamic uncertainty quantifies attention (HDUQ-Attention) and multi-scale fusion pooling method (MF Pooling) and multi-objective loss (MO loss). DMFF select different convolution kernels according to the feature maps at different levels, extract different-scale feature information, and make the feature information of each layer have stronger representation ability for information fusion HDUQ-Attention includes a tuning block that adjust the attention weight according to the different information of each layer, and a Monte-Carlo (MC) dropout structure for quantifying uncertainty MF Pooling is a pooling method designed for multi-scale models, which can speed up the calculation and prevent overfitting while retaining the main important information Because the number of parameters in the backbone part of DM-CNN is different from other modules, MO loss is proposed, which has a fast optimization speed and good classification effect DM-CNN conducts experiments on publicly available datasets in four areas of medicine (Dermatology, Histopathology, Respirology, Ophthalmology), achieving state-of-the-art classification performance on all datasets. DM-CNN can not only maintain excellent performance, but also solve the problem of quantification of uncertainty, which is a very important task for the medical field. The code is available: https://github.com/QIANXIN22/DM-CNN.

EEG-Based Seizure Detection Using Variable-Frequency Complex Demodulation and Convolutional Neural Networks

Epilepsy is a complex neurological disorder characterized by recurrent and unpredictable seizures that affect millions of people around the world. Early and accurate epilepsy detection is critical for timely medical intervention and improved patient outcomes. Several methods and classifiers for automated epilepsy detection have been developed in previous research. However, the existing research landscape requires innovative approaches that can further improve the accuracy of diagnosing and managing patients. This study investigates the application of variable-frequency complex demodulation (VFCDM) and convolutional neural networks (CNN) to discriminate between healthy, interictal, and ictal states using electroencephalogram (EEG) data. For testing this approach, the EEG signals were collected from the publicly available Bonn dataset. A high-resolution time–frequency spectrum (TFS) of each EEG signal was obtained using the VFCDM. The TFS images were fed to the CNN classifier for the classification of the signals. The performance of CNN was evaluated using leave-one-subject-out cross-validation (LOSO CV). The TFS shows variations in its frequency for different states that correspond to variation in the neural activity. The LOSO CV approach yields a consistently high performance, ranging from 90% to 99% between different combinations of healthy and epilepsy states (interictal and ictal). The extensive LOSO CV validation approach ensures the reliability and robustness of the proposed method. As a result, the research contributes to advancing the field of epilepsy detection and brings us one step closer to developing practical, reliable, and efficient diagnostic tools for clinical applications.