A medical image segmentation network (RPR-MLP) based on multi-scale convolution and parallel reverse attention is presented. In the encoder,Res2net modules and tokenized multi-layer perceptron modules are used as the backbone structure to extract multi-scale information and enhance the diversity of semantic features. Meanwhile,the accuracy of semantic information extraction in the decoder is improved through parallel partial decoder. Additionally,reverse attention module re-emphasizes the focus on important regions for further improving the accuracy of segmentation results. The proposed method achieves Dice scores of 0.8967 and 0.8762 on the Kvasir and ISIC 2018 public datasets,respectively,demonstrating its effectiveness and generalization ability in medical image segmentation. Furthermore,when applied to the lung tumor CT image dataset (LungCancer dataset) collected in the study,the proposed network has Dice score,IoU and F1 score of 0.7278,58.83％and 67.85％,respectively,outperforming baseline network (UNeXt) and common CNN (U-Net,AttU-Net,U-Net++and PraNet) by 0.0301-0.0578、3.16％-4.70％,and 6.72％-18.53％,respectively. The study confirms the effectiveness and generalization ability of RPR-MLP network on different datasets,providing important technical support for lung tumor image segmentation.

A medical image segmentation network (RPR-MLP) based on multi-scale convolution and parallel reverse attention is presented. In the encoder,Res2net modules and tokenized multi-layer perceptron modules are used as the backbone structure to extract multi-scale information and enhance the diversity of semantic features. Meanwhile,the accuracy of semantic information extraction in the decoder is improved through parallel partial decoder. Additionally,reverse attention module re-emphasizes the focus on important regions for further improving the accuracy of segmentation results. The proposed method achieves Dice scores of 0.8967 and 0.8762 on the Kvasir and ISIC 2018 public datasets,respectively,demonstrating its effectiveness and generalization ability in medical image segmentation. Furthermore,when applied to the lung tumor CT image dataset (LungCancer dataset) collected in the study,the proposed network has Dice score,IoU and F1 score of 0.7278,58.83％and 67.85％,respectively,outperforming baseline network (UNeXt) and common CNN (U-Net,AttU-Net,U-Net++and PraNet) by 0.0301-0.0578、3.16％-4.70％,and 6.72％-18.53％,respectively. The study confirms the effectiveness and generalization ability of RPR-MLP network on different datasets,providing important technical support for lung tumor image segmentation.

Considering the classification problems of electroencephalogram(EEG)signals and their nonlinear,non-stationary characteristics,multivariate variational mode decomposition(MVMD)is introduced to process steady-state visual evoked potential(SSVEP)signals.Herein a novel classification algorithm for SSVEP called MVMDMS-CCA which combines a new approach for mode selection with canonical correlation analysis(CCA)algorithm is presented.MVMDMS-CCA method uses the signal-to-noise ratio to determine the key parameter K in MVMD,and then performs MVMD decomposition.Mode selection is carried out by setting a threshold using the maximal information coefficient(MIC)method,and the modes not meeting the threshold are adaptively denoised using wavelet denoising.A new combination of modes is constructed and input into the CCA algorithm to achieve SSVEP signal classification.The proposed method is validated on a self-collected EEG dataset,and it achieves an average classification accuracy of 93.23%under a 3 s window,showing 5.78%higher than standard CCA and 1.51%higher than the improved filter bank CCA.MVMDMS-CCA can effectively extract SSVEP components from EEG signals while suppressing noises,providing a new perspective for the research of SSVEP decoding algorithms.

Considering the classification problems of electroencephalogram(EEG)signals and their nonlinear,non-stationary characteristics,multivariate variational mode decomposition(MVMD)is introduced to process steady-state visual evoked potential(SSVEP)signals.Herein a novel classification algorithm for SSVEP called MVMDMS-CCA which combines a new approach for mode selection with canonical correlation analysis(CCA)algorithm is presented.MVMDMS-CCA method uses the signal-to-noise ratio to determine the key parameter K in MVMD,and then performs MVMD decomposition.Mode selection is carried out by setting a threshold using the maximal information coefficient(MIC)method,and the modes not meeting the threshold are adaptively denoised using wavelet denoising.A new combination of modes is constructed and input into the CCA algorithm to achieve SSVEP signal classification.The proposed method is validated on a self-collected EEG dataset,and it achieves an average classification accuracy of 93.23%under a 3 s window,showing 5.78%higher than standard CCA and 1.51%higher than the improved filter bank CCA.MVMDMS-CCA can effectively extract SSVEP components from EEG signals while suppressing noises,providing a new perspective for the research of SSVEP decoding algorithms.

To address the inability of the existing machine learning methods to simultaneously consider both the temporal and spatial domain features of electroencephalogram(EEG)signals in classifying EEG features,a feature fusion based Attention-EEGNet-BiGRU(AEBGNet)is presented for classifying motor imagery(MI)EEG signals.AEBGNet is capable of fusing the temporal domain features extracted by convolutional neural network with attention mechanism and spatial domain features extracted by a bidirectional gated recurrent unit to obtain more distinctive spatiotemporal features.The constructed AEBGNet classification model achieves an average accuracy of 80.37%on the BCI competition IV 2b dataset,and there is an improvement of 6.09%over the standard EEGNet method.The results demonstrate the effectiveness of the proposed method in enhancing the classification accuracy of MI EEG signals,providing a new idea for MI EEG signal classification.