RESUMO
Crowd counting is a challenging task due to large perspective, density, and scale variations. CNN-based crowd counting techniques have achieved significant performance in sparse to dense environments. However, crowd counting in high perspective-varying scenes (images) is getting harder due to different density levels occupied by the same number of pixels. In this way large variations for objects in the same spatial area make it difficult to count accurately. Further, existing CNN-based crowd counting methods are used to extract rich deep features; however, these features are used locally and disseminated while propagating through intermediate layers. This results in high counting errors, especially in dense and high perspective-variation scenes. Further, class-specific responses along channel dimensions are underestimated. To address these above mentioned issues, we therefore propose a CNN-based dense feature extraction network for accurate crowd counting. Our proposed model comprises three main modules: (1) backbone network, (2) dense feature extraction modules (DFEMs), and (3) channel attention module (CAM). The backbone network is used to obtain general features with strong transfer learning ability. The DFEM is composed of multiple sub-modules called dense stacked convolution modules (DSCMs), densely connected with each other. In this way features extracted from lower and middle-lower layers are propagated to higher layers through dense connections. In addition, combinations of task independent general features obtained by the former modules and task-specific features obtained by later ones are incorporated to obtain high counting accuracy in large perspective-varying scenes. Further, to exploit the class-specific response between background and foreground, CAM is incorporated at the end to obtain high-level features along channel dimensions for better counting accuracy. Moreover, we have evaluated the proposed method on three well known datasets: Shanghaitech (Part-A), Shanghaitech (Part-B), and Venice. The performance of the proposed technique justifies its relative effectiveness in terms of selected performance compared to state-of-the-art techniques.
RESUMO
BACKGROUND: Spinal cord injury (SCI) is a severe medical condition affecting the hand and locomotor function. New medical technologies, including various wearable devices, as well as rehabilitation treatments are being developed to enhance hand function in patients with SCI. As three-dimensional (3D) printing has the advantage of being able to produce low-cost personalized devices, there is a growing appeal to apply this technology to rehabilitation equipment in conjunction with scientific advances. In this study, we proposed a novel 3D-printed hand orthosis that is controlled by electromyography (EMG) signals. The orthosis was designed to aid the grasping function for patients with cervical SCI. We applied this hand exoskeleton system to individuals with tetraplegia due to SCI and validated its effectiveness. METHODS: The 3D architecture of the device was designed using computer-aided design software and printed with a polylactic acid filament. The dynamic hand orthosis enhanced the tenodesis grip to provide sufficient grasping function. The root mean square of the EMG signal was used as the input for controlling the device. Ten subjects with hand weakness due to chronic cervical SCI were enrolled in this study, and their hand function was assessed before and after wearing the orthosis. The Toronto Rehabilitation Institute Hand Function Test (TRI-HFT) was used as the primary outcome measure. Furthermore, improvements in functional independence in daily living and device usability were evaluated. RESULTS: The newly developed orthosis improved hand function of subjects, as determined using the TRI-HFT (p < 0.05). Furthermore, participants obtained immediate functionality on eating after wearing the orthosis. Moreover, most participants were satisfied with the device as determined by the usability test. There were no side effects associated with the experiment. CONCLUSIONS: The 3D-printed myoelectric hand orthosis was intuitive, easy to use, and showed positive effects in its ability to handle objects encountered in daily life. This study proved that combining simple EMG-based control strategies and 3D printing techniques was feasible and promising in rehabilitation engineering. TRIAL REGISTRATION: Clinical Research Information Service (CRiS), Republic of Korea. KCT0003995. Registered 2 May 2019 - Retrospectively registered.
Assuntos
Eletromiografia/instrumentação , Mãos , Aparelhos Ortopédicos , Impressão Tridimensional , Traumatismos da Medula Espinal/reabilitação , Idoso , Desenho Assistido por Computador , Eletromiografia/métodos , Feminino , Mãos/fisiopatologia , Humanos , Masculino , Pessoa de Meia-IdadeRESUMO
Surface electromyography (sEMG) signals comprise electrophysiological information related to muscle activity. As this signal is easy to record, it is utilized to control several myoelectric prostheses devices. Several studies have been conducted to process sEMG signals more efficiently. However, research on optimal algorithms and electrode placements for the processing of sEMG signals is still inconclusive. In addition, very few studies have focused on minimizing the number of electrodes. In this study, we investigated the most effective method for myoelectric signal classification with a small number of electrodes. A total of 23 subjects participated in the study, and the sEMG data of 14 different hand movements of the subjects were acquired from targeted muscles and untargeted muscles. Furthermore, the study compared the classification accuracy of the sEMG data using discriminative feature-oriented dictionary learning (DFDL) and other conventional classifiers. DFDL demonstrated the highest classification accuracy among the classifiers, and its higher quality performance became more apparent as the number of channels decreased. The targeted method was superior to the untargeted method, particularly when classifying sEMG signals with DFDL. Therefore, it was concluded that the combination of the targeted method and the DFDL algorithm could classify myoelectric signals more effectively with a minimal number of channels.
Assuntos
Eletromiografia/métodos , Movimento/fisiologia , Músculos/fisiologia , Adulto , Eletrodos , Feminino , Mãos/fisiologia , Humanos , Masculino , Processamento de Sinais Assistido por Computador , Máquina de Vetores de Suporte , Adulto JovemRESUMO
Different modalities such as structural MRI, FDG-PET, and CSF have complementary information, which is likely to be very useful for diagnosis of AD and MCI. Therefore, it is possible to develop a more effective and accurate AD/MCI automatic diagnosis method by integrating complementary information of different modalities. In this paper, we propose multi-modal sparse hierarchical extreme leaning machine (MSH-ELM). We used volume and mean intensity extracted from 93 regions of interest (ROIs) as features of MRI and FDG-PET, respectively, and used p-tau, t-tau, and Aß42 as CSF features. In detail, high-level representation was individually extracted from each of MRI, FDG-PET, and CSF using a stacked sparse extreme learning machine auto-encoder (sELM-AE). Then, another stacked sELM-AE was devised to acquire a joint hierarchical feature representation by fusing the high-level representations obtained from each modality. Finally, we classified joint hierarchical feature representation using a kernel-based extreme learning machine (KELM). The results of MSH-ELM were compared with those of conventional ELM, single kernel support vector machine (SK-SVM), multiple kernel support vector machine (MK-SVM) and stacked auto-encoder (SAE). Performance was evaluated through 10-fold cross-validation. In the classification of AD vs. HC and MCI vs. HC problem, the proposed MSH-ELM method showed mean balanced accuracies of 96.10% and 86.46%, respectively, which is much better than those of competing methods. In summary, the proposed algorithm exhibits consistently better performance than SK-SVM, ELM, MK-SVM and SAE in the two binary classification problems (AD vs. HC and MCI vs. HC).
Assuntos
Algoritmos , Doença de Alzheimer/diagnóstico , Disfunção Cognitiva/diagnóstico , Aprendizado Profundo , Imageamento por Ressonância Magnética , Tomografia por Emissão de Pósitrons , Idoso , Idoso de 80 Anos ou mais , Doença de Alzheimer/líquido cefalorraquidiano , Doença de Alzheimer/diagnóstico por imagem , Doença de Alzheimer/patologia , Disfunção Cognitiva/líquido cefalorraquidiano , Disfunção Cognitiva/diagnóstico por imagem , Disfunção Cognitiva/patologia , Bases de Dados Factuais , Diagnóstico Diferencial , Feminino , Humanos , Masculino , Pessoa de Meia-Idade , Máquina de Vetores de SuporteRESUMO
Sitting posture monitoring systems (SPMSs) help assess the posture of a seated person in real-time and improve sitting posture. To date, SPMS studies reported have required many sensors mounted on the backrest plate and seat plate of a chair. The present study, therefore, developed a system that measures a total of six sitting postures including the posture that applied a load to the backrest plate, with four load cells mounted only on the seat plate. Various machine learning algorithms were applied to the body weight ratio measured by the developed SPMS to identify the method that most accurately classified the actual sitting posture of the seated person. After classifying the sitting postures using several classifiers, average and maximum classification rates of 97.20% and 97.94%, respectively, were obtained from nine subjects with a support vector machine using the radial basis function kernel; the results obtained by this classifier showed a statistically significant difference from the results of multiple classifications using other classifiers. The proposed SPMS was able to classify six sitting postures including the posture with loading on the backrest and showed the possibility of classifying the sitting posture even though the number of sensors is reduced.
Assuntos
Postura , Computadores , Humanos , Aprendizado de MáquinaRESUMO
So far, many approaches have been developed for motion artifact (MA) reduction from photoplethysmogram (PPG). Specifically, single-input MA reduction methods are useful to apply wearable and mobile healthcare systems because of their low hardware costs and simplicity. However, most of them are insufficiently developed to be used in real-world situations, and they suffer from a phase distortion problem. In this study, we propose a novel single-input MA reduction algorithm based on time-variant forward-backward harmonic notch filtering. To verify the proposed method, we collected real PPG data corrupted by MA and compared it with existing single-input MA reduction methods. In conclusion, the proposed zero-phase line enhancer (ZLE) was found to be superior for MA reduction and exhibited zero phase response.
RESUMO
Virtual reality (VR) is a computer technique that creates an artificial environment composed of realistic images, sounds, and other sensations. Many researchers have used VR devices to generate various stimuli, and have utilized them to perform experiments or to provide treatment. In this study, the participants performed mental tasks using a VR device while physiological signals were measured: a photoplethysmogram (PPG), electrodermal activity (EDA), and skin temperature (SKT). In general, stress is an important factor that can influence the autonomic nervous system (ANS). Heart-rate variability (HRV) is known to be related to ANS activity, so we used an HRV derived from the PPG peak interval. In addition, the peak characteristics of the skin conductance (SC) from EDA and SKT variation can also reflect ANS activity; we utilized them as well. Then, we applied a kernel-based extreme-learning machine (K-ELM) to correctly classify the stress levels induced by the VR task to reflect five different levels of stress situations: baseline, mild stress, moderate stress, severe stress, and recovery. Twelve healthy subjects voluntarily participated in the study. Three physiological signals were measured in stress environment generated by VR device. As a result, the average classification accuracy was over 95% using K-ELM and the integrated feature (IT = HRV + SC + SKT). In addition, the proposed algorithm can embed a microcontroller chip since K-ELM algorithm have very short computation time. Therefore, a compact wearable device classifying stress levels using physiological signals can be developed.
RESUMO
BACKGROUNDS: The heartbeat is fundamental cardiac activity which is straightforwardly detected with a variety of measurement techniques for analyzing physiological signals. Unfortunately, unexpected noise or contaminated signals can distort or cut out electrocardiogram (ECG) signals in practice, misleading the heartbeat detectors to report a false heart rate or suspend itself for a considerable length of time in the worst case. To deal with the problem of unreliable heartbeat detection, PhysioNet/CinC suggests a challenge in 2014 for developing robust heart beat detectors using multimodal signals. METHODS: This article proposes a multimodal data association method that supplements ECG as a primary input signal with blood pressure (BP) and electroencephalogram (EEG) as complementary input signals when input signals are unreliable. If the current signal quality index (SQI) qualifies ECG as a reliable input signal, our method applies QRS detection to ECG and reports heartbeats. Otherwise, the current SQI selects the best supplementary input signal between BP and EEG after evaluating the current SQI of BP. When BP is chosen as a supplementary input signal, our association model between ECG and BP enables us to compute their regular intervals, detect characteristics BP signals, and estimate the locations of the heartbeat. When both ECG and BP are not qualified, our fusion method resorts to the association model between ECG and EEG that allows us to apply an adaptive filter to ECG and EEG, extract the QRS candidates, and report heartbeats. RESULTS: The proposed method achieved an overall score of 86.26 % for the test data when the input signals are unreliable. Our method outperformed the traditional method, which achieved 79.28 % using QRS detector and BP detector from PhysioNet. Our multimodal signal processing method outperforms the conventional unimodal method of taking ECG signals alone for both training and test data sets. CONCLUSIONS: To detect the heartbeat robustly, we have proposed a novel multimodal data association method of supplementing ECG with a variety of physiological signals and accounting for the patient-specific lag between different pulsatile signals and ECG. Multimodal signal detectors and data-fusion approaches such as those proposed in this article can reduce false alarms and improve patient monitoring.
Assuntos
Pressão Sanguínea , Eletroencefalografia , Coração/fisiologia , Modelos Estatísticos , Processamento de Sinais Assistido por Computador , Adulto , Algoritmos , Humanos , Fatores de TempoRESUMO
Fetal heart rate (FHR) is an important determinant of fetal health. Cardiotocography (CTG) is widely used for measuring the FHR in the clinical field. However, fetal movement and blood flow through the maternal blood vessels can critically influence Doppler ultrasound signals. Moreover, CTG is not suitable for long-term monitoring. Therefore, researchers have been developing algorithms to estimate the FHR using electrocardiograms (ECGs) from the abdomen of pregnant women. However, separating the weak fetal ECG signal from the abdominal ECG signal is a challenging problem. In this paper, we propose a method for estimating the FHR using sequential total variation denoising and compare its performance with that of other single-channel fetal ECG extraction methods via simulation using the Fetal ECG Synthetic Database (FECGSYNDB). Moreover, we used real data from PhysioNet fetal ECG databases for the evaluation of the algorithm performance. The R-peak detection rate is calculated to evaluate the performance of our algorithm. Our approach could not only separate the fetal ECG signals from the abdominal ECG signals but also accurately estimate the FHR.
Assuntos
Abdome/diagnóstico por imagem , Algoritmos , Eletrocardiografia , Feto/diagnóstico por imagem , Simulação por Computador , Bases de Dados como Assunto , Feminino , Humanos , Gravidez , Processamento de Sinais Assistido por ComputadorRESUMO
BACKGROUND: Monitoring of intracranial pressure (ICP) is highly important for detecting abnormal brain conditions such as intracranial hemorrhage, cerebral edema, or brain tumor. Until now, the monitoring of ICP requires an invasive method which has many disadvantages including the risk of infections, hemorrhage, or brain herniation. Therefore, many non-invasive methods have been proposed for estimating ICP. However, these methods are still insufficient to estimate sudden increases in ICP. METHODS: We proposed a simplified intracranial hemo- and hydro-dynamics model that consisted of two simple resistance circuits. From this proposed model, we designed an ICP estimation algorithm to trace ICP changes. First, we performed a simulation based on the original Ursino model with the real arterial blood pressure to investigate our proposed approach. We subsequently applied it to experimental data that were measured during the Valsalva maneuver (VM) and resting state, respectively. RESULTS: Simulation result revealed a small root mean square error (RMSE) between the estimated ICP by our approach and the reference ICP derived from the original Ursino model. Compared to the pulsatility index (PI) based approach and Kashif's model, our proposed method showed more statistically significant difference between VM and resting state. CONCLUSION: Our proposed method successfully tracked sudden ICP increases. Therefore, our method may serve as a suitable tool for non-invasive ICP monitoring.
Assuntos
Algoritmos , Determinação da Pressão Arterial/métodos , Simulação por Computador , Hemodinâmica , Hidrodinâmica , Hipertensão Intracraniana/diagnóstico , Pressão Intracraniana/fisiologia , Modelos Biológicos , Ultrassonografia Doppler Transcraniana/métodos , Manobra de Valsalva/fisiologia , Determinação da Pressão Arterial/instrumentação , Água Corporal , Sistemas Computacionais , Humanos , Hipertensão Intracraniana/fisiopatologia , Monitorização Fisiológica/métodos , Fluxo Pulsátil , DescansoRESUMO
BACKGROUND: Many researchers have attempted to acquire respiratory rate (RR) information from a photoplethysmogram (PPG) because respiration affects the waveform of the PPG. However, most of these methods were difficult to operate in real-time because of their complexity or computational requirements. From these needs, we attempted to develop a method to estimate RR from a PPG with a light computational burden. METHODS: To obtain RR information, we adopt a sequential filtering structure and frequency estimation technique, which extracts a dominant frequency from a given signal. In particular, we used an adaptive lattice notch filter (ALNF) to estimate RR from a PPG along with an additional heart rate that is utilized as an adaptation parameter of our method. Furthermore, we designed a sequential infinite impulse response (IIR) notch filtering system (i.e., harmonic IIR notch filter) to eliminate the cardiac component and its harmonics from the PPG. We compared the proposed method with Burg's AR modeling method, which is widely used to estimate RR from a PPG, using open-source data and measured data. RESULTS: By using a statistical test, it was determined that our adaptive lattice-type respiratory rate estimator (ALRE) was significantly more accurate than Burg's AR model method (p <0.0001). Furthermore, the ALRE's tracking performance was better than that of Burg's method, and the variances of its estimates were smaller than those of Burg's method. CONCLUSIONS: In short, our method showed a better performance than Burg's AR modeling method for real-time applications.
Assuntos
Oximetria/métodos , Fotopletismografia/métodos , Processamento de Sinais Assistido por Computador , Adulto , Algoritmos , Biofísica , Sistema Cardiovascular , Feminino , Frequência Cardíaca , Humanos , Masculino , Distribuição Normal , Oxigênio/química , Análise de Regressão , Respiração , Taxa Respiratória , SoftwareRESUMO
In this paper, we propose a system for inferring the pinch-to-zoom gesture using surface EMG (Electromyography) signals in real time. Pinch-to-zoom, which is a common gesture in smart devices such as an iPhone or an Android phone, is used to control the size of images or web pages according to the distance between the thumb and index finger. To infer the finger motion, we recorded EMG signals obtained from the first dorsal interosseous muscle, which is highly related to the pinch-to-zoom gesture, and used a support vector machine for classification between four finger motion distances. The powers which are estimated by Welch's method were used as feature vectors. In order to solve the multiclass classification problem, we applied a one-versus-one strategy, since a support vector machine is basically a binary classifier. As a result, our system yields 93.38% classification accuracy averaged over six subjects. The classification accuracy was estimated using 10-fold cross validation. Through our system, we expect to not only develop practical prosthetic devices but to also construct a novel user experience (UX) for smart devices.
Assuntos
Eletromiografia/métodos , Movimento (Física) , Interface Usuário-Computador , Adulto , Algoritmos , Feminino , Humanos , Masculino , Reconhecimento Automatizado de Padrão , Processamento de Sinais Assistido por Computador , Fatores de TempoRESUMO
A robust pattern recognition framework is required for ideal real-time human-machine interface (HMI) applications. Convolutional neural networks and recurrent neural networks have been widely used for the classification of gestures based on electromyography (EMG), but few studies have demonstrated the effectiveness of using a vision transformer for this purpose. Additionally, the accuracy achieved is influenced by the efficacy of the preprocessing pipeline. This study assessed ViT with and without an attention mechanism for precise motor intent decoding by investigating various input features and integrating convolutive blind source separation (BSS) preprocessing. All investigations were carried out with two open-access high-density surface EMG datasets of 34 and 21 hand gestures recorded from 20 and 5 healthy subjects respectively. Integration of centering and optimal extension factors resulted in better performance with raw input. However, spatial whitening increased the model's sensitivity to noise. The best-performing BSS-integrated convolution vision transformer model (BSS-CViT) model yielded an accuracy of 96.61% and 91.98% on test datasets one and two. This is a promising result for future studies in real-time HMI applications. The code implementation results reported in this study are available on GitHub. https://github.com/deremustapha/BSS-ViT.
RESUMO
Introduction: In this study, we classified electroencephalography (EEG) data of imagined speech using signal decomposition and multireceptive convolutional neural network. The imagined speech EEG with five vowels /a/, /e/, /i/, /o/, and /u/, and mute (rest) sounds were obtained from ten study participants. Materials and methods: First, two different signal decomposition methods were applied for comparison: noise-assisted multivariate empirical mode decomposition and wavelet packet decomposition. Six statistical features were calculated from the decomposed eight sub-frequency bands EEG. Next, all features obtained from each channel of the trial were vectorized and used as the input vector of classifiers. Lastly, EEG was classified using multireceptive field convolutional neural network and several other classifiers for comparison. Results: We achieved an average classification rate of 73.09 and up to 80.41% in a multiclass (six classes) setup (Chance: 16.67%). In comparison with various other classifiers, significant improvements for other classifiers were achieved (p-value < 0.05). From the frequency sub-band analysis, high-frequency band regions and the lowest-frequency band region contain more information about imagined vowel EEG data. The misclassification and classification rate of each vowel imaginary EEG was analyzed through a confusion matrix. Discussion: Imagined speech EEG can be classified successfully using the proposed signal decomposition method and a convolutional neural network. The proposed classification method for imagined speech EEG can contribute to developing a practical imagined speech-based brain-computer interfaces system.
RESUMO
Cervical cancer is a common and preventable disease that poses a significant threat to women's health and well-being. It is the fourth most prevalent cancer among women worldwide, with approximately 604,000 new cases and 342,000 deaths in 2020, according to the World Health Organization. Early detection and diagnosis of cervical cancer are crucial for reducing mortality and morbidity rates. The Papanicolaou smear test is a widely used screening method that involves the examination of cervical cells under a microscope to identify any abnormalities. However, this method is time-consuming, labor-intensive, subjective, and prone to human errors. Artificial intelligence techniques have emerged as a promising alternative to improve the accuracy and efficiency of Papanicolaou smear diagnosis. Artificial intelligence techniques can automatically analyze Papanicolaou smear images and classify them into normal or abnormal categories, as well as detect the severity and type of lesions. This paper provides a comprehensive review of the recent advances in artificial intelligence diagnostics of the Papanicolaou smear, focusing on the methods, datasets, performance metrics, and challenges. The paper also discusses the potential applications and future directions of artificial intelligence diagnostics of the Papanicolaou smear.
RESUMO
BACKGROUND AND OBJECTIVES: Cervical cancer affects around 0.5 million women per year, resulting in over 0.3 million fatalities. Therefore, repetitive screening for cervical cancer is of utmost importance. Computer-assisted diagnosis is key for scaling up cervical cancer screening. Current recognition algorithms, however, perform poorly on the whole-slide image (WSI) analysis, fail to generalize for different staining methods and on uneven distribution for subtype imaging, and provide sub-optimal clinical-level interpretations. Herein, we developed CervixFormer-an end-to-end, multi-scale swin transformer-based adversarial ensemble learning framework to assess pre-cancerous and cancer-specific cervical malignant lesions on WSIs. METHODS: The proposed framework consists of (1) a self-attention generative adversarial network (SAGAN) for generating synthetic images during patch-level training to address the class imbalanced problems; (2) a multi-scale transformer-based ensemble learning method for cell identification at various stages, including atypical squamous cells (ASC) and atypical squamous cells of undetermined significance (ASCUS), which have not been demonstrated in previous studies; and (3) a fusion model for concatenating ensemble-based results and producing final outcomes. RESULTS: In the evaluation, the proposed method is first evaluated on a private dataset of 717 annotated samples from six classes, obtaining a high recall and precision of 0.940 and 0.934, respectively, in roughly 1.2 minutes. To further examine the generalizability of CervixFormer, we evaluated it on four independent, publicly available datasets, namely, the CRIC cervix, Mendeley LBC, SIPaKMeD Pap Smear, and Cervix93 Extended Depth of Field image datasets. CervixFormer obtained a fairly better performance on two-, three-, four-, and six-class classification of smear- and cell-level datasets. For clinical interpretation, we used GradCAM to visualize a coarse localization map, highlighting important regions in the WSI. Notably, CervixFormer extracts feature mostly from the cell nucleus and partially from the cytoplasm. CONCLUSIONS: In comparison with the existing state-of-the-art benchmark methods, the CervixFormer outperforms them in terms of recall, accuracy, and computing time.
Assuntos
Teste de Papanicolaou , Neoplasias do Colo do Útero , Feminino , Humanos , Neoplasias do Colo do Útero/diagnóstico por imagem , Detecção Precoce de Câncer/métodos , Colo do Útero/diagnóstico por imagem , Colo do Útero/patologia , Diagnóstico por ComputadorRESUMO
Heart rate variability (HRV) is a known psychophysiological marker for diverse psychiatric symptoms. In this study, we aimed to explore the potential for clinical use of HRV by investigating the interrelationship between HRV indices and clinical measures mainly used to assess depressive and anxious symptoms. Participants who reported depressive and anxious symptoms were designated into the following groups: group 1, clinician-rated and self-rated depression; group 2, only self-rated depression; group 3, clinician-rated and self-rated anxiety; group 4, only self-rated anxiety. Statistical comparisons were performed between these groups to investigate the association between HRV and clinical measures. As a result, HRV variables showed significant correlations only with the clinician-rated assessments. Moreover, both time and frequency domain HRV indices were significantly different between groups 1 and 2, but groups 3 and 4 showed significant differences only in frequency domain HRV indices. Our study showed that HRV is an objective indicator for depressive or anxious symptoms. Additionally, it is considered a potential indicator for predicting the severity or state of depressive symptoms rather than of anxious symptoms. This study will contribute to increasing the diagnostic utility of discriminating those symptoms based on HRV in the future.
RESUMO
Over the years, considerable research has been conducted to investigate the mechanisms of speech perception and recognition. Electroencephalography (EEG) is a powerful tool for identifying brain activity; therefore, it has been widely used to determine the neural basis of speech recognition. In particular, for the classification of speech recognition, deep learning-based approaches are in the spotlight because they can automatically learn and extract representative features through end-to-end learning. This study aimed to identify particular components that are potentially related to phoneme representation in the rat brain and to discriminate brain activity for each vowel stimulus on a single-trial basis using a bidirectional long short-term memory (BiLSTM) network and classical machine learning methods. Nineteen male Sprague-Dawley rats subjected to microelectrode implantation surgery to record EEG signals from the bilateral anterior auditory fields were used. Five different vowel speech stimuli were chosen, /a/, /e/, /i/, /o/, and /u/, which have highly different formant frequencies. EEG recorded under randomly given vowel stimuli was minimally preprocessed and normalized by a z-score transformation to be used as input for the classification of speech recognition. The BiLSTM network showed the best performance among the classifiers by achieving an overall accuracy, f1-score, and Cohen's κ values of 75.18%, 0.75, and 0.68, respectively, using a 10-fold cross-validation approach. These results indicate that LSTM layers can effectively model sequential data, such as EEG; hence, informative features can be derived through BiLSTM trained with end-to-end learning without any additional hand-crafted feature extraction methods.
Assuntos
Percepção da Fala , Animais , Eletroencefalografia/métodos , Masculino , Memória de Curto Prazo , Redes Neurais de Computação , Ratos , Ratos Sprague-Dawley , FalaRESUMO
Abrupt and continuous nature of scale variation in a crowded scene is a challenging task to enhance crowd counting accuracy in an image. Existing crowd counting techniques generally used multi-column or single-column dilated convolution to tackle scale variation due to perspective distortion. However, due to multi-column nature, they obtain identical features, whereas, the standard dilated convolution (SDC) with expanded receptive field size has sparse pixel sampling rate. Due to sparse nature of SDC, it is highly challenging to obtain relevant contextual information. Further, features at multiple scale are not extracted despite some inception-based model is not used (which is cost effective). To mitigate theses drawbacks in SDC, we therefore, propose a hierarchical dense dilated deep pyramid feature extraction through convolution neural network (CNN) for single image crowd counting (HDPF). It comprises of three modules: general feature extraction module (GFEM), deep pyramid feature extraction module (PFEM) and fusion module (FM). The GFEM is responsible to obtain task independent general features. Whereas, PFEM plays a vital role to obtain the relevant contextual information due to dense pixel sampling rate caused by densely connected dense stacked dilated convolutional modules (DSDCs). Further, due to dense connections among DSDCs, the final feature map acquires multi-scale information with expanded receptive field as compared to SDC. Due to dense pyramid nature, it is very effective to propagate the extracted feature from lower dilated convolutional layers (DCLs) to middle and higher DCLs, which result in better estimation accuracy. The FM is used to fuse the incoming features extracted by other modules. The proposed technique is tested through simulations on three well known datasets: Shanghaitech (Part-A), Shanghaitech (Part-B) and Venice. Results justify its relative effectiveness in terms of selected performance.
Assuntos
Processamento de Imagem Assistida por Computador , Redes Neurais de Computação , Fusão Gênica , Processamento de Imagem Assistida por Computador/métodos , Manejo de EspécimesRESUMO
Electroencephalography (EEG) microstate analysis is a method wherein spontaneous EEG activity is segmented at sub-second levels to analyze quasi-stable states. In particular, four archetype microstates and their features are known to reflect changes in brain state in neuropsychiatric diseases. However, previous studies have only reported differences in each microstate feature and have not determined whether microstate features are suitable for schizophrenia classification. Therefore, it is necessary to validate microstate features for schizophrenia classification. Nineteen microstate features, including duration, occurrence, and coverage as well as thirty-one conventional EEG features, including statistical, frequency, and temporal characteristics were obtained from resting-state EEG recordings of 14 patients diagnosed with schizophrenia and from 14 healthy (control) subjects. Machine-learning based multivariate analysis was used to evaluate classification performance. EEG recordings of patients and controls showed different microstate features. More importantly, when differentiating among patients and controls, EEG microstate features outperformed conventional EEG ones. The performance of the microstate features exceeded that of conventional EEG, even after optimization using recursive feature elimination. EEG microstate features applied with conventional EEG features also showed better classification performance than conventional EEG features alone. The current study is the first to validate the use of microstate features to discriminate schizophrenia, suggesting that EEG microstate features are useful for schizophrenia classification.