In this study, we proposed a network structure based on the shifted attention mechanism, namely U-Shiftformer, to overcome the limitation of existing convolution neural networks (CNNs) in brain tumor segmentation that lacks multimodal information interaction. The U-Shiftformer takes the U-shape encoder-decoder structure as the backbone and embeds the proposed Shiftformer module to exchange the information between modalities in the downsampling process. The Shiftformer module contains one standard attention, and three proposed shifted attention modules, in which the shifted attention module considers the information exchange by constructing the relationship between adjacent modalities. In the experiments, we compare the proposed U-Shiftformer with SOTA networks in the dice, precision, and Hausdorff metrics. Its average accuracies of these metrics surpass all the comparison networks and achieve 0.8424, 0.8675, 0.9244, and 1.2961 in dice, precision, sensitivity, and Hausdorff metrics, respectively.

In this work, we propose a new framework, called Geospatial-temporal Convolutional Neural Network (GT-CNN), and construct the video-based geospatial-temporal precipitation dataset from the surveillance cameras of the eight weather stations (sampling points) to recognize the precipitation intensity. GT-CNN has three key modules: (1) Geospatial module, (2) Temporal module, (3) Fusion module. In the geospatial module, we extract the precipitation information from each sampling point simultaneously, and that is used to construct the geospatial relationships using LSTM between various sampling points. In the temporal module, we take 3D convolution to grab the precipitation features with time information, considering a series of precipitation images for each sampling point. Finally, we generate the fusion module to fuse the geospatial and temporal features. We evaluate our framework with three metrics and compare GT-CNN with the state-of-the-art methods using the self-collected dataset. Experimental results demonstrated that our approach surpasses state-of-the-art methods concerning various metrics.