Pub Date : 2022-11-05DOI: 10.1109/CISP-BMEI56279.2022.9979903
Xiaoning Liu, Peiyao Guo, Jinhong Liu, Dongcheng Tuo, Shiyu Lei, Yuejin Wang
Facial attribute editing aims to change the facial attributes, which can be regarded as an image translation problem. Facial attribute editing is usually realized by combining encoder-decoder and Generative Adversarial Networks, but the generated image is not realistic enough, and the model has weak ability to control the fine granularity of face attributes of generated images. In this work, we propose a Generative Adversarial Network ISTSA-GAN based on Independent Selective Transfer Unit (ISTU) and Self-attention Mechanism. On the basis of STGAN, we use ISTU instead of Selective Transfer Unit (STU) to combine with encoder-decoder to selectively transfer the features of encoder. In addition, a self-attention mechanism is introduced into the transposed convolution layer of the decoder to establish long-distance dependence of the model across image regions. Finally, attribute interpolation loss and source domain adversarial loss are added to constrain the training of the model. Experimental results show that this method can improve the ability of editing attributes and saving much details, and enhance the ability of fine-grained control of editing attributes. It is superior to classical methods in attribute editing accuracy and image quality.
人脸属性编辑的目的是改变人脸属性，这可以看作是一个图像翻译问题。人脸属性编辑通常采用编码器-解码器和生成对抗网络相结合的方式来实现，但生成的图像不够逼真，模型对生成图像人脸属性细粒度的控制能力较弱。在这项工作中，我们提出了一个基于独立选择转移单元(ISTU)和自注意机制的生成式对抗网络ISTSA-GAN。在STGAN的基础上，我们用ISTU代替选择性传输单元(Selective Transfer Unit, STU)与编解码器结合，选择性地传输编码器的特征。此外，在解码器的转置卷积层中引入自关注机制，建立模型跨图像区域的远距离依赖关系。最后，加入属性插值损失和源域对抗损失来约束模型的训练。实验结果表明，该方法提高了编辑属性和保存大量细节的能力，增强了编辑属性的细粒度控制能力。该方法在属性编辑精度和图像质量方面优于经典方法。
{"title":"Facial Attribute Editing based on Independent Selective Transfer Unit and Self-attention Mechanism","authors":"Xiaoning Liu, Peiyao Guo, Jinhong Liu, Dongcheng Tuo, Shiyu Lei, Yuejin Wang","doi":"10.1109/CISP-BMEI56279.2022.9979903","DOIUrl":"https://doi.org/10.1109/CISP-BMEI56279.2022.9979903","url":null,"abstract":"Facial attribute editing aims to change the facial attributes, which can be regarded as an image translation problem. Facial attribute editing is usually realized by combining encoder-decoder and Generative Adversarial Networks, but the generated image is not realistic enough, and the model has weak ability to control the fine granularity of face attributes of generated images. In this work, we propose a Generative Adversarial Network ISTSA-GAN based on Independent Selective Transfer Unit (ISTU) and Self-attention Mechanism. On the basis of STGAN, we use ISTU instead of Selective Transfer Unit (STU) to combine with encoder-decoder to selectively transfer the features of encoder. In addition, a self-attention mechanism is introduced into the transposed convolution layer of the decoder to establish long-distance dependence of the model across image regions. Finally, attribute interpolation loss and source domain adversarial loss are added to constrain the training of the model. Experimental results show that this method can improve the ability of editing attributes and saving much details, and enhance the ability of fine-grained control of editing attributes. It is superior to classical methods in attribute editing accuracy and image quality.","PeriodicalId":198522,"journal":{"name":"2022 15th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"102 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115609885","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-05DOI: 10.1109/CISP-BMEI56279.2022.9979931
Zhaoxin Hao, J. Sun, D. Gu
Airborne terahertz synthetic aperture radar (THz-SAR) is sensitive to the tiny vibration of the platform because of the short wavelength. Therefore, the phase errors caused by high-frequency vibration of the platform needs to be considered in the motion compensation (MOCO) for THz-SAR imaging. There have been many MOCO methods to compensate the phase errors caused by high-frequency vibration. However, in some cases, the low-frequency motion errors also need to be considered. Different from these methods, this paper proposes a novel MOCO method which compensates both the high-frequency vibration and the low-frequency motion errors. Firstly, the instantaneous chirp rate (ICR) and the instantaneous frequency are both estimated using chirplet decomposition. After filtering out the low-frequency component of the ICR, we obtain the estimate of high-frequency component by using the least squares (LS) sequential estimators. Then, the high-frequency component in the instantaneous frequency is removed, and the parameters of the low-frequency motion are estimated using LS estimator. Finally, the errors are compensated according to the estimated parameters, and the residual phase errors can be compensated by the phase gradient autofocus (PGA) algorithm. The simulation results validate the effectivity of the proposed method.
{"title":"A Novel Motion Compensation Method for High Resolution Terahertz SAR Imaging","authors":"Zhaoxin Hao, J. Sun, D. Gu","doi":"10.1109/CISP-BMEI56279.2022.9979931","DOIUrl":"https://doi.org/10.1109/CISP-BMEI56279.2022.9979931","url":null,"abstract":"Airborne terahertz synthetic aperture radar (THz-SAR) is sensitive to the tiny vibration of the platform because of the short wavelength. Therefore, the phase errors caused by high-frequency vibration of the platform needs to be considered in the motion compensation (MOCO) for THz-SAR imaging. There have been many MOCO methods to compensate the phase errors caused by high-frequency vibration. However, in some cases, the low-frequency motion errors also need to be considered. Different from these methods, this paper proposes a novel MOCO method which compensates both the high-frequency vibration and the low-frequency motion errors. Firstly, the instantaneous chirp rate (ICR) and the instantaneous frequency are both estimated using chirplet decomposition. After filtering out the low-frequency component of the ICR, we obtain the estimate of high-frequency component by using the least squares (LS) sequential estimators. Then, the high-frequency component in the instantaneous frequency is removed, and the parameters of the low-frequency motion are estimated using LS estimator. Finally, the errors are compensated according to the estimated parameters, and the residual phase errors can be compensated by the phase gradient autofocus (PGA) algorithm. The simulation results validate the effectivity of the proposed method.","PeriodicalId":198522,"journal":{"name":"2022 15th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"20 4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128299036","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-05DOI: 10.1109/CISP-BMEI56279.2022.9979870
Yunlong Zhu, Wenlong Zhang, Biao Yan, Rongqian Yang
Intramuscular (IM) injection is mainly performed manually at present. Large-scale COVID-19 vaccination has exposed various problems of manual IM injection. In addition, the clinical success rate of manual IM injection is also unsatisfactory. Using robotic intramuscular injection system (RIMIS) is expected to realize automated vaccination and improve the success rate of IM injection. The existing robotic needle insertion system based on image guidance is not a practical option for IM injection because of the time-consuming medical imaging process. In this paper, an optical guidance method for RIMIS is proposed, which uses near-infrared optical tracking system and retro-reflective patch to achieve rapid acquisition of surface normal vector. A closed loop formed by six coordinate systems is used to realize the accurate control of the injection angle and depth. Experimental results show that the RIMIS based on the proposed method can complete the simulated IM injection operation without image guidance and possess accurate control of the injection angle and depth.
{"title":"An Optical Guidance Method for Robotic Intramuscular Injection System","authors":"Yunlong Zhu, Wenlong Zhang, Biao Yan, Rongqian Yang","doi":"10.1109/CISP-BMEI56279.2022.9979870","DOIUrl":"https://doi.org/10.1109/CISP-BMEI56279.2022.9979870","url":null,"abstract":"Intramuscular (IM) injection is mainly performed manually at present. Large-scale COVID-19 vaccination has exposed various problems of manual IM injection. In addition, the clinical success rate of manual IM injection is also unsatisfactory. Using robotic intramuscular injection system (RIMIS) is expected to realize automated vaccination and improve the success rate of IM injection. The existing robotic needle insertion system based on image guidance is not a practical option for IM injection because of the time-consuming medical imaging process. In this paper, an optical guidance method for RIMIS is proposed, which uses near-infrared optical tracking system and retro-reflective patch to achieve rapid acquisition of surface normal vector. A closed loop formed by six coordinate systems is used to realize the accurate control of the injection angle and depth. Experimental results show that the RIMIS based on the proposed method can complete the simulated IM injection operation without image guidance and possess accurate control of the injection angle and depth.","PeriodicalId":198522,"journal":{"name":"2022 15th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129360349","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-05DOI: 10.1109/CISP-BMEI56279.2022.9980092
Yupeng Wang, Shuqing He, Xiaowei Wei, Samuel Akolade George
Most of the human action recognition systems based on 3-Dimensional Convolutional Neural Network (3D CNN) architecture recognize human actions frame by frame in video streams, which need to be deployed on high-performance platforms such as cloud servers. Through the targeted optimization of the processing method of each frame of the video in the process of human action recognition, the computing power requirements and the total processing time of human action recognition are reduced. The optimization of human action recognition is tested and verified by the Kinetics-700 dataset, and the accuracy of action recognition is similar to that before optimization, and the total recognition time is only 14.1 % of the total time before optimization. It effectively reduces the performance requirements of the deployment platform, improves the real-time performance of action recognition, and increases the practicability of human action recognition based on deep learning in the application of low computing power platforms.
{"title":"Research on an Effective Human Action Recognition Model Based on 3D CNN","authors":"Yupeng Wang, Shuqing He, Xiaowei Wei, Samuel Akolade George","doi":"10.1109/CISP-BMEI56279.2022.9980092","DOIUrl":"https://doi.org/10.1109/CISP-BMEI56279.2022.9980092","url":null,"abstract":"Most of the human action recognition systems based on 3-Dimensional Convolutional Neural Network (3D CNN) architecture recognize human actions frame by frame in video streams, which need to be deployed on high-performance platforms such as cloud servers. Through the targeted optimization of the processing method of each frame of the video in the process of human action recognition, the computing power requirements and the total processing time of human action recognition are reduced. The optimization of human action recognition is tested and verified by the Kinetics-700 dataset, and the accuracy of action recognition is similar to that before optimization, and the total recognition time is only 14.1 % of the total time before optimization. It effectively reduces the performance requirements of the deployment platform, improves the real-time performance of action recognition, and increases the practicability of human action recognition based on deep learning in the application of low computing power platforms.","PeriodicalId":198522,"journal":{"name":"2022 15th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"61 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131209518","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-05DOI: 10.1109/CISP-BMEI56279.2022.9980099
Guang Li, Chengwei Sun, Zeyu Sun
At the beginning of 2020, coronavirus disease 2019(COVID-19) infection spread in Wuhan, China and all over the world. Until April, it had affected millions of people. The computed tomography (CT) imaging is confirmed as one of the assessment method for COVID-19 patients. However distinguish the COVID-19 from those CT images is extremely challenging as it is very time-consuming, and lack of the experienced radiologists. So deep learning based approaches are proposed to triage the COVID-19 images from the normal or other pneumonia images. Here, we proposed a novel global average pooling (GAP) method for the deep neural network to improve the performance of the COVID-19 classification. The novel GAP method is using lung mask region as weighting factor for GAP, which reduce the influence of background region and highlight the classification features of interesting tissue region. The result of our method achieved the triage of COVID-19 with sensitivity 96.4 % and specificity 93.3 % on the independence validation dataset with 2062 CT scans.
{"title":"A Deep Learning Based Method For COVID-19 Classification Using Chest CT Images","authors":"Guang Li, Chengwei Sun, Zeyu Sun","doi":"10.1109/CISP-BMEI56279.2022.9980099","DOIUrl":"https://doi.org/10.1109/CISP-BMEI56279.2022.9980099","url":null,"abstract":"At the beginning of 2020, coronavirus disease 2019(COVID-19) infection spread in Wuhan, China and all over the world. Until April, it had affected millions of people. The computed tomography (CT) imaging is confirmed as one of the assessment method for COVID-19 patients. However distinguish the COVID-19 from those CT images is extremely challenging as it is very time-consuming, and lack of the experienced radiologists. So deep learning based approaches are proposed to triage the COVID-19 images from the normal or other pneumonia images. Here, we proposed a novel global average pooling (GAP) method for the deep neural network to improve the performance of the COVID-19 classification. The novel GAP method is using lung mask region as weighting factor for GAP, which reduce the influence of background region and highlight the classification features of interesting tissue region. The result of our method achieved the triage of COVID-19 with sensitivity 96.4 % and specificity 93.3 % on the independence validation dataset with 2062 CT scans.","PeriodicalId":198522,"journal":{"name":"2022 15th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130969111","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-05DOI: 10.1109/CISP-BMEI56279.2022.9980292
Xuewei Chen, Zhihua Huang
P300 brain-computer interface (BCI) is an important field of brain science exploration, but the calibration of P300 affects its application. To solve this problem, we propose an algorithm that combines transfer learning and reinforcement learning. In the reinforcement learning algorithm, we refer to P300 linear upper confidence bound(PLUCB). Due to the particularity of the PLUCB algorithm, we modify it and integrate the idea of online transfer learning. The new algorithm is applied to the calibration-free classification of P300 BCI, using the classifier matrices of the subjects in the source domain, without collecting additional session data of the target subjects for calibration. We test the performance of the classifier at different stages of the algorithm. For each subject, the agent constantly updates on the first part of the data and the second part of the data is used for testing. The results show that our designed algorithm P300 Homogeneous Online Transfer Learning (PHomOTL) has better performance than PLUCB, transfer PLUCB (TPLUCB) and Stepwise Linear Discriminant Analysis (SWLDA). When 10000 trials are used for training and the remaining 5120 trials are used for testing, the average P300 classification accuracy of PHomOTL is 73.15% and the average character classification accuracy of PHomOTL is 79.46%.
{"title":"A P300 BCI calibration-free algorithm based on intersubject transfer and reinforcement learning","authors":"Xuewei Chen, Zhihua Huang","doi":"10.1109/CISP-BMEI56279.2022.9980292","DOIUrl":"https://doi.org/10.1109/CISP-BMEI56279.2022.9980292","url":null,"abstract":"P300 brain-computer interface (BCI) is an important field of brain science exploration, but the calibration of P300 affects its application. To solve this problem, we propose an algorithm that combines transfer learning and reinforcement learning. In the reinforcement learning algorithm, we refer to P300 linear upper confidence bound(PLUCB). Due to the particularity of the PLUCB algorithm, we modify it and integrate the idea of online transfer learning. The new algorithm is applied to the calibration-free classification of P300 BCI, using the classifier matrices of the subjects in the source domain, without collecting additional session data of the target subjects for calibration. We test the performance of the classifier at different stages of the algorithm. For each subject, the agent constantly updates on the first part of the data and the second part of the data is used for testing. The results show that our designed algorithm P300 Homogeneous Online Transfer Learning (PHomOTL) has better performance than PLUCB, transfer PLUCB (TPLUCB) and Stepwise Linear Discriminant Analysis (SWLDA). When 10000 trials are used for training and the remaining 5120 trials are used for testing, the average P300 classification accuracy of PHomOTL is 73.15% and the average character classification accuracy of PHomOTL is 79.46%.","PeriodicalId":198522,"journal":{"name":"2022 15th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129523828","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-05DOI: 10.1109/CISP-BMEI56279.2022.9980236
Tianhao Wang, Haiqing Jiang
The analysis of intra-pulse characteristics of radar signal is an important part of radar reconnaissance, real-time analysis of intra-pulse features based on instantaneous frequency can efficiently recognize signals of various modulation types and extract parameters. This method has a high recognition rate under certain signal-to-noise ratio, and the algorithm is simple. It can be implemented at high speed on radar reconnaissance digital receiver.
{"title":"Real-time analysis of Intra-pulse characteristics based on instantaneous frequency","authors":"Tianhao Wang, Haiqing Jiang","doi":"10.1109/CISP-BMEI56279.2022.9980236","DOIUrl":"https://doi.org/10.1109/CISP-BMEI56279.2022.9980236","url":null,"abstract":"The analysis of intra-pulse characteristics of radar signal is an important part of radar reconnaissance, real-time analysis of intra-pulse features based on instantaneous frequency can efficiently recognize signals of various modulation types and extract parameters. This method has a high recognition rate under certain signal-to-noise ratio, and the algorithm is simple. It can be implemented at high speed on radar reconnaissance digital receiver.","PeriodicalId":198522,"journal":{"name":"2022 15th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"69 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127773437","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-05DOI: 10.1109/CISP-BMEI56279.2022.9980111
Qing Li
In face recognition, we may encounter face images with shadow and illumination, which will affect the recognition. In this scenario, the low-rank matrix and a sparse matrix can be obtained by low-rank matrix decomposition of the collected original face image, where the low-rank matrix is the face image without shadow and illumination. In order to obtain the low-rank matrix, the Sub-gradient method and AIRLS method are used in this paper, and their effects are compared in the experimental verification of Yale face database.
{"title":"Face Recognition with Robust Matrix Factorization","authors":"Qing Li","doi":"10.1109/CISP-BMEI56279.2022.9980111","DOIUrl":"https://doi.org/10.1109/CISP-BMEI56279.2022.9980111","url":null,"abstract":"In face recognition, we may encounter face images with shadow and illumination, which will affect the recognition. In this scenario, the low-rank matrix and a sparse matrix can be obtained by low-rank matrix decomposition of the collected original face image, where the low-rank matrix is the face image without shadow and illumination. In order to obtain the low-rank matrix, the Sub-gradient method and AIRLS method are used in this paper, and their effects are compared in the experimental verification of Yale face database.","PeriodicalId":198522,"journal":{"name":"2022 15th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125811628","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-11-05DOI: 10.1109/CISP-BMEI56279.2022.9979952
Yashpal Singh, Seba Susan
Cancer subtyping from gene expression data is trending research in the field of bioinformatics. Classification of gene expression data is a challenging task due to the small number of samples and large number of features involved. The problem is further complicated due to the strong class imbalance issue prevalent in gene expression datasets. The challenge here is to find an end-to-end machine learning solution to classify cancer subtypes from small sample, high-dimensional, imbalanced gene expression datasets. In this study, we propose a SMOTE-LASSO-DeepNet framework for the identification of cancer subtypes from gene expression data. The proposed framework balances the training set using SMOTE, and then finds the most informative genes using LASSO. The balanced and pruned training set is then applied as input to a deep neural network (DeepNet) with four hidden layers having 512, 256, 128 and 64 neurons respectively. We tested our framework on four different cancer gene expression datasets: Leukemia, Lung cancer, Brain cancer and Breast cancer. It is observed from the results that our proposed SMOTE-LASSO-DeepNet framework performs consistently best as compared to the existing methods.
{"title":"SMOTE-LASSO-DeepNet Framework for Cancer Subtyping from Gene Expression Data","authors":"Yashpal Singh, Seba Susan","doi":"10.1109/CISP-BMEI56279.2022.9979952","DOIUrl":"https://doi.org/10.1109/CISP-BMEI56279.2022.9979952","url":null,"abstract":"Cancer subtyping from gene expression data is trending research in the field of bioinformatics. Classification of gene expression data is a challenging task due to the small number of samples and large number of features involved. The problem is further complicated due to the strong class imbalance issue prevalent in gene expression datasets. The challenge here is to find an end-to-end machine learning solution to classify cancer subtypes from small sample, high-dimensional, imbalanced gene expression datasets. In this study, we propose a SMOTE-LASSO-DeepNet framework for the identification of cancer subtypes from gene expression data. The proposed framework balances the training set using SMOTE, and then finds the most informative genes using LASSO. The balanced and pruned training set is then applied as input to a deep neural network (DeepNet) with four hidden layers having 512, 256, 128 and 64 neurons respectively. We tested our framework on four different cancer gene expression datasets: Leukemia, Lung cancer, Brain cancer and Breast cancer. It is observed from the results that our proposed SMOTE-LASSO-DeepNet framework performs consistently best as compared to the existing methods.","PeriodicalId":198522,"journal":{"name":"2022 15th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114439697","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
In this paper, the Hurst index calculation method based on multipoint fractional Brown bridge was used to analyze the electroencephalogram(EEG) of schizophrenia patients and healthy people under the same sound paradigm experiment. We used this method to analyze the short-term EEG signals of the healthy group and the patient group around the time point 100ms after stimulation and found that the method can effectively analyze the Hurst index of short-time series, in the frontal lobe and central area. There were significant differences in passage, and the Hurst index was lower in healthy people than in patients. The results show that in this experiment, the long-term correlation of EEG signals after stimulation in patients with schizophrenia is higher, and the complexity of EEG signals is lower, which can help clinical diagnosis of schizophrenia better. At the same time, this paper compares the Hurst exponent calculation method based on the multi-point fractional Brown bridge with the traditional rescaled range analysis method. The Hurst index calculation of the sequence can analyze the difference between the healthy group and the patient group on a smaller scale.
{"title":"Hurst Exponent Analysis Of Schizophrenia Electroencephalogram Based On Multi-point Fractional Brownian Bridge","authors":"Congzhou Zhong, Wenpo Yao, Wanyi Yi, Jui-Pin Wang, Dengxuan Bai, Qiong Wang","doi":"10.1109/CISP-BMEI56279.2022.9980315","DOIUrl":"https://doi.org/10.1109/CISP-BMEI56279.2022.9980315","url":null,"abstract":"In this paper, the Hurst index calculation method based on multipoint fractional Brown bridge was used to analyze the electroencephalogram(EEG) of schizophrenia patients and healthy people under the same sound paradigm experiment. We used this method to analyze the short-term EEG signals of the healthy group and the patient group around the time point 100ms after stimulation and found that the method can effectively analyze the Hurst index of short-time series, in the frontal lobe and central area. There were significant differences in passage, and the Hurst index was lower in healthy people than in patients. The results show that in this experiment, the long-term correlation of EEG signals after stimulation in patients with schizophrenia is higher, and the complexity of EEG signals is lower, which can help clinical diagnosis of schizophrenia better. At the same time, this paper compares the Hurst exponent calculation method based on the multi-point fractional Brown bridge with the traditional rescaled range analysis method. The Hurst index calculation of the sequence can analyze the difference between the healthy group and the patient group on a smaller scale.","PeriodicalId":198522,"journal":{"name":"2022 15th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI)","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122787693","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：