Fusion Block Research Articles

A retinal vessel segmentation network with multiple-dimension attention and adaptive feature fusion

The incidence of blinding eye diseases is highly correlated with changes in retinal morphology, and is clinically detected by segmenting retinal structures in fundus images. However, some existing methods have limitations in accurately segmenting thin vessels. In recent years, deep learning has made a splash in the medical image segmentation, but the lack of edge information representation due to repetitive convolution and pooling, limits the final segmentation accuracy. To this end, this paper proposes a pixel-level retinal vessel segmentation network with multiple-dimension attention and adaptive feature fusion. Here, a multiple dimension attention enhancement (MDAE) block is proposed to acquire more local edge information. Meanwhile, a deep guidance fusion (DGF) block and a cross-pooling semantic enhancement (CPSE) block are proposed simultaneously to acquire more global contexts. Further, the predictions of different decoding stages are learned and aggregated by an adaptive weight learner (AWL) unit to obtain the best weights for effective feature fusion. The experimental results on three public fundus image datasets show that proposed network could effectively enhance the segmentation performance on retinal blood vessels. In particular, the proposed method achieves AUC of 98.30%, 98.75%, and 98.71% on the DRIVE, CHASE_DB1, and STARE datasets, respectively, while the F1 score on all three datasets exceeded 83%. The source code of the proposed model is available at https://github.com/gegao310/VesselSeg-Pytorch-master.

EfficientPolypSeg: Efficient Polyp Segmentation in colonoscopy images using EfficientNet-B5 with dilated blocks and attention mechanisms

Background:Colorectal cancer ranks as the second leading cause of cancer-related death worldwide, prompting the need for effective strategies to combat this disease. The urgent need for effective strategies to combat this disease has led to the exploration of innovative approaches in medical imaging and diagnosis. Automatic segmentation and detection of colorectal polyps in colonoscopy images have emerged as a promising approach, enabling timely identification and diagnosis of colorectal diseases. Methods:This paper introduces EfficientPolypSeg, an advanced segmentation architecture that leverages the capabilities of EfficientNet-B5 as the encoder, enhanced by two dilated blocks, a Channel Attention Block (CAB), a Spatial Attention Block (SAB), and a Feature Fusion Block (FFB). Notably, our model exhibits superior performance compared to state-of-the-art methods, achieving exceptional accuracy and segmentation quality. Results:To validate the effectiveness of EfficientPolypSeg, extensive experiments were conducted on four publicly available datasets. The results demonstrate the remarkable performance of our model, showcasing its ability to outperform existing approaches in terms of accuracy and segmentation quality. Our model achieved an impressive mean Dice score of 0.92 and a mean Intersection over Union (IoU) of 0.86 on these datasets. These findings highlight the potential impact of EfficientPolypSeg in improving the detection and diagnosis of colorectal polyps. Conclusion:EfficientPolypSeg, the segmentation architecture proposed in this study, exhibits remarkable capabilities in the automatic detection and segmentation of colorectal polyps. The model’s superior performance, as evidenced by extensive experimentation on diverse datasets, emphasizes its potential impact on improving the detection and diagnosis of colorectal diseases.

A Multi-Stage Feature Aggregation and Structure Awareness Network for Concrete Bridge Crack Detection.

One of the most significant problems affecting a concrete bridge's safety is cracks. However, detecting concrete bridge cracks is still challenging due to their slender nature, low contrast, and background interference. The existing convolutional methods with square kernels struggle to capture crack features effectively, fail to perceive the long-range dependencies between crack regions, and have weak suppression ability for background noises, leading to low detection precision of bridge cracks. To address this problem, a multi-stage feature aggregation and structure awareness network (MFSA-Net) for pixel-level concrete bridge crack detection is proposed in this paper. Specifically, in the coding stage, a structure-aware convolution block is proposed by combining square convolution with strip convolution to perceive the linear structure of concrete bridge cracks. Square convolution is used to capture detailed local information. In contrast, strip convolution is employed to interact with the local features to establish the long-range dependence relationship between discrete crack regions. Unlike the self-attention mechanism, strip convolution also suppresses background interference near crack regions. Meanwhile, the feature attention fusion block is presented for fusing features from the encoder and decoder at the same stage, which can sharpen the edges of concrete bridge cracks. In order to fully utilize the shallow detail features and deep semantic features, the features from different stages are aggregated to obtain fine-grained segmentation results. The proposed MFSA-Net was trained and evaluated on the publicly available concrete bridge crack dataset and achieved average results of 73.74%, 77.04%, 75.30%, and 60.48% for precision, recall, F1 score, and IoU, respectively, on three typical sub-datasets, thus showing optimal performance in comparison with other existing methods. MFSA-Net also gained optimal performance on two publicly available concrete pavement crack datasets, thereby indicating its adaptability to crack detection across diverse scenarios.

Audio-visual saliency prediction with multisensory perception and integration

Audio-visual saliency prediction (AVSP) is a task that aims to model human attention patterns in the perception of auditory and visual scenes. Given the challenges associated with perceiving and combining multi-modal saliency features from videos, this paper presents a multi-sensory framework for AVSP. This framework is designed to extract audio, motion and image saliency features and integrate them effectively, which can then serve as a general architecture for the AVSP task. To obtain multi-sensory information, we develop a three-stream encoder that extracts audio, motion and image saliency features. In particular, we utilize a pre-trained encoder with knowledge related to image saliency to extract saliency features for each frame. The image saliency features are then incorporated with motion features using a spatial attention module. For motion features, 3D convolutional neural networks (CNNs) like S3D are commonly used in AVSP models. However, these networks are unable to effectively capture the global motion relationship in videos. To tackle this problem, we incorporate Transformer- and MLP-based motion encoders into the AVSP models. To learn joint audio-visual representations, an audio-visual fusion block is exploited to enhance the correlation between audio and visual motion features under the supervision of a cosine similarity loss in a self-supervised manner. Finally, a multi-stage decoder integrates audio, motion and image saliency features to generate the final saliency map. We evaluate our methods on six audio-visual eye-tracking datasets. Experimental results demonstrate that our method achieves compelling performance compared to the state-of-the-art methods. The source code is available at https://github.com/oraclefina/MSPI.

Relation fusion propagation network for transductive few-shot learning

Previous graph-based meta-learning approaches have explored pairwise feature similarity to learn instance-level relations of samples, however, the gap between the sample relations in feature and label spaces is often ignored. It is empirically observed that instances with different labels may display considerable similarity in visual characteristics, making it challenging to distinguish between them in the feature space. To this end, we propose a dual-branch Relation Fusion Propagation Network (RFPN) for transductive few-shot learning, which explicitly models both feature and label relations across support-query pairs. Specifically, we design a Relation Fusion Block (RFB) to fuse instance-level and class-level relations, thus obtaining more robust fusion-level relations to guide feature propagation. In addition to the feature propagation branch, we encode the label relations and construct a label shortcut branch for label propagation. To alleviate the error accumulation during propagation, which is caused by uncertain pseudo-labels for query samples, we propose a Label Shortcut Mechanism (LSM) to progressively update the sample relations with the initial labels. Our full method is plug-and-play and can be easily applied in existing graph-based approaches for transductive few-shot learning. Extensive experiments demonstrate that our proposed RFPN yields significant improvements over the baselines and achieves promising performance on four popular few-shot classification benchmarks.

ERCP-Net: a channel extension residual structure and adaptive channel attention mechanism for plant leaf disease classification network

Plant leaf diseases are a major cause of plant mortality, especially in crops. Timely and accurately identifying disease types and implementing proper treatment measures in the early stages of leaf diseases are crucial for healthy plant growth. Traditional plant disease identification methods rely heavily on visual inspection by experts in plant pathology, which is time-consuming and requires a high level of expertise. So, this approach fails to gain widespread adoption. To overcome these challenges, we propose a channel extension residual structure and adaptive channel attention mechanism for plant leaf disease classification network (ERCP-Net). It consists of channel extension residual block (CER-Block), adaptive channel attention block (ACA-Block), and bidirectional information fusion block (BIF-Block). Meanwhile, an application for the real-time detection of plant leaf diseases is being created to assist precision agriculture in practical situations. Finally, experiments were conducted to compare our model with other state-of-the-art deep learning methods on the PlantVillage and AI Challenger 2018 datasets. Experimental results show that our model achieved an accuracy of 99.82% and 86.21%, respectively. Also, it demonstrates excellent robustness and scalability, highlighting its potential for practical implementation.

Multi-Modal Enhancement Transformer Network for Skeleton-Based Human Interaction Recognition.

Skeleton-based human interaction recognition is a challenging task in the field of vision and image processing. Graph Convolutional Networks (GCNs) achieved remarkable performance by modeling the human skeleton as a topology. However, existing GCN-based methods have two problems: (1) Existing frameworks cannot effectively take advantage of the complementary features of different skeletal modalities. There is no information transfer channel between various specific modalities. (2) Limited by the structure of the skeleton topology, it is hard to capture and learn the information about two-person interactions. To solve these problems, inspired by the human visual neural network, we propose a multi-modal enhancement transformer (ME-Former) network for skeleton-based human interaction recognition. ME-Former includes a multi-modal enhancement module (ME) and a context progressive fusion block (CPF). More specifically, each ME module consists of a multi-head cross-modal attention block (MH-CA) and a two-person hypergraph self-attention block (TH-SA), which are responsible for enhancing the skeleton features of a specific modality from other skeletal modalities and modeling spatial dependencies between joints using the specific modality, respectively. In addition, we propose a two-person skeleton topology and a two-person hypergraph representation. The TH-SA block can embed their structural information into the self-attention to better learn two-person interaction. The CPF block is capable of progressively transforming the features of different skeletal modalities from low-level features to higher-order global contexts, making the enhancement process more efficient. Extensive experiments on benchmark NTU-RGB+D 60 and NTU-RGB+D 120 datasets consistently verify the effectiveness of our proposed ME-Former by outperforming state-of-the-art methods.

LA-YOLO: an effective detection model for multi-UAV under low altitude background

Detecting unmanned aerial vehicles (UAVs) in various environments and conditions is highly demanded in applications, and for solving the problem of detecting UAVs under low altitude background, we propose a high performance and effective LA-YOLO network by integrating the SimAM attention mechanism and introducing a fusion block with the normalized Wasserstein distance. By recording images of multi-UAV under low altitude background and annotating them, we construct a dataset called GUET-UAV-LA to evaluate the performance of the proposed network. Using the GUET-UAV-LA dataset and public datasets, the experiments validate the effectiveness of the proposed network and show that LA-YOLO can improve mAP by up to 5.9% compared to the existing networks.

Volumetric Imitation Generative Adversarial Networks for Anatomical Human Body Modeling.

Volumetric representation is a technique used to express 3D objects in various fields, such as medical applications. On the other hand, tomography images for reconstructing volumetric data have limited utilization because they contain personal information. Existing GAN-based medical image generation techniques can produce virtual tomographic images for volume reconstruction while preserving the patient's privacy. Nevertheless, these images often do not consider vertical correlations between the adjacent slices, leading to erroneous results in 3D reconstruction. Furthermore, while volume generation techniques have been introduced, they often focus on surface modeling, making it challenging to represent the internal anatomical features accurately. This paper proposes volumetric imitation GAN (VI-GAN), which imitates a human anatomical model to generate volumetric data. The primary goal of this model is to capture the attributes and 3D structure, including the external shape, internal slices, and the relationship between the vertical slices of the human anatomical model. The proposed network consists of a generator for feature extraction and up-sampling based on a 3D U-Net and ResNet structure and a 3D-convolution-based LFFB (local feature fusion block). In addition, a discriminator utilizes 3D convolution to evaluate the authenticity of the generated volume compared to the ground truth. VI-GAN also devises reconstruction loss, including feature and similarity losses, to converge the generated volumetric data into a human anatomical model. In this experiment, the CT data of 234 people were used to assess the reliability of the results. When using volume evaluation metrics to measure similarity, VI-GAN generated a volume that realistically represented the human anatomical model compared to existing volume generation methods.

RII-GAN: Multi-scaled Aligning-Based Reversed Image Interaction Network for Text-to-Image Synthesis

The text-to-image (T2I) model based on a single-stage generative adversarial network (GAN) has significantly succeeded in recent years. However, the generation model based on GAN has two disadvantages: the generator does not introduce any image feature manifold structure, which makes it challenging to align the image and text features. Another is the image’s diversity; the text’s abstraction will prevent the model from learning the actual image distribution. This paper proposes a reversed image interaction generative adversarial network (RII-GAN), which consists of four components: text encoder, reversed image interaction network (RIIN), adaptive affine-based generator, and dual-channel feature alignment discriminator (DFAD). RIIN indirectly introduces the actual image distribution into the generation network, thus overcoming the problem that the network lacks the learning of the actual image feature manifold structure and generating the distribution of text-matching images. Each adaptive affine block (AAB) in the proposed affine-based generator can adaptively enhance text information, establishing an updated relation between original independent fusion blocks and the image feature. Moreover, this study designs a DFAD to capture important feature information of images and text in two channels. Such a dual-channel backbone improves semantic consistency by utilizing a particular synchronized bi-modal information extraction structure. We have performed experiments on publicly available datasets to prove the effectiveness of our model.

An intelligent wildfire identification method based on weighted boxes fusion and convolutional block attention module

Identifying wildfires is a key task to ensure timely and effective response and prevent the spread of wildfires. The widely used methods for identifying wildfires based on infrared and satellite remote sensing still have problems such as poor identification accuracy and high deployment costs. To address the high deployment costs, this paper adopted the method of deep learning network identifying visible light images.To address the issue of poor accuracy, this paper proposed an intelligent wildfire identification method based on Weighted Boxes Fusion (WBF) and Convolutional Block Attention Module (CBAM). In order to enhance the feature expression of flames and smoke, this paper adopted two different scale SSD networks combined with CBAM to adjust the feature weights of relevant regions and channels adaptively and uses them as the fusion base model. In order to further improve the performance of the model and integrate two different feature extraction strategies, this paper adopted the WBF method to weigh and fuse the identification results of the base model. The method was evaluated on a self-built dataset. The experimental results showed that the mAP of this method reached 96.19%, which is the optimal performance method compared to existing methods. The model proposed in this paper can effectively detect wildfires from images and has practical application potential in wildfire monitoring and management.

TSoft-Net: A novel transfer soft thresholding network based on self-attention for intelligent fault diagnosis of rotating machinery

In the industrial scenes, the machinery operates under diverse working conditions and generates varying levels of noise, which can hinder the performance of the intelligent fault diagnosis model that is trained in the laboratory. This is because the different working conditions and environmental noise change the distribution of laboratory vibration signal. To tackle this issue, we proposed a novel transfer soft thresholding network (TSoft-Net) based on self-attention for intelligent fault diagnosis of rotating machinery, which has good anti-noise performance and can be transferred to different working conditions with excellent accuracy. This paper constructs a soft residual block for extracting the fault representation and enhancing the residual learning. In this block, we propose a residual factor to learn and enhance domain-invariant fault representation. Furthermore, a representation soft fusion block is built for extracting and fusing the different scale fault representation. In this block, we propose a scale-attention weight to dynamically fuse the different scale fault representation. The experiments show that the TSoft-Net, compared with six existing methods on eight sub-datasets, has better anti-noise ability and achieves an accuracy of up to 100% on the target domain.

Crop plant automatic detecting based on in-field images by lightweight DFU-Net model

Efficiently and accurately extracting crop plants from authentic and complex field images is essential to ensure the economic efficiency of agricultural production and enhance farming operations. However, many existing deep Convolutional Neural Networks (CNNs) prioritize accuracy over efficiency in crop detection, often lacking real-time capabilities. Therefore, this study aims to investigate the automated detection of various crops, such as cotton plants and seed melons, in natural environments using a U-Net established with a Double-depth convolutional and Fusion block (DFU-Net). This involves exploring compound loss functions for improved accuracy. The DFU-Net model combines the strengths of lightweight CNNs and the U-Net model, constructing a Lightweight Crop with a Double-Depth Convolution backbone (LC-DDC). This is achieved by introducing a Double-Depth Convolutional block (DDC block) and incorporating initial and fusion blocks at the first and last positions of the encoder, respectively. The effectiveness of these designs is confirmed through ablation experiments and interpretability analyses. The experimental results indicate that the DFU-Net model achieved Pixel Accuracy (PA), mean Intersection over Union (mIoU), and F1 metrics exceeding 92.0 % for the cotton plant, seed melon, and Chloroplast Viability Prediction in Plant Phenotyping (CVPPP) dataset. Particularly, on the CVPPP dataset, DFU-Net outperformed the Pyramid Scene Parsing Network (PSPNet) model, demonstrating 16.4 %, 17.2 %, and 13.2 % higher PA, mIoU, and F1 scores, respectively. In addition, DFU-Net demonstrated efficient model space utilization, requiring only 0.975 MB and 1.68 GB for parameters and floating-point operations (FLOPs), respectively. The performance of low-performance computers is impressive, achieving a detection speed of 10.5 Frames Per Second (FPS), which is 6.9 times faster than that of the Deeplabv3 + model. This achievement demonstrates a balanced compromise between detection speed and accuracy. This study provides novel approaches for optimizing crop-detection algorithms, providing valuable technical support for intelligent crop management.

Research on the short-term wind power prediction with dual branch multi-source fusion strategy

The smooth operation of the power system imposes high requirements on wind power forecasting. Given the forecasting challenges brought by the coexistence of strong and weak disturbances in wind power data, this paper proposes a hierarchical anti-disturbance mechanism to raise the accuracy of single-point wind power prediction. In the first-level disturbance processing, we prioritize the treatment of strong disturbances in wind power data through a dual-branch multi-source fusion prediction structure. The wind power subsequences with distinct time–frequency characteristics are divided into two sets, processed by the stationary set prediction branch and the drastic set prediction branch, respectively. For the drastic set, we incorporate microenvironmental multi-source data surrounding the wind turbine as input and propose a parallel-structured attention fusion module CA (Convolution attention fusion block) to provide effective data support for the prediction module, then extract rich fusion features through deep temporal LSTM algorithm and ultimately achieve fine-grained forecasting. For weak disturbances that have not been adequately addressed in the previous step, we conduct a secondary treatment through error correction strategy. By employing the hierarchical processing approach, this paper accomplishes a comprehensive handling of disturbances with different intensities. Experimental results on four data sets show that the proposed model DEWFM (Dual-branch and error correction wind power forecast) demonstrates remarkable performance advantages in comparing with the baseline model and other advanced models, the average values of MAE reaching 1.679, revealing the effectiveness of the proposed method in enhancing the precision of short-term wind power prediction.

Multi-modality frequency-aware cross attention network for fake news detection

An increasing number of fake news combining text, images and other forms of multimedia are spreading rapidly across social platforms, leading to misinformation and negative impacts. Therefore, the automatic identification of multimodal fake news has become an important research hotspot in academia and industry. The key to multimedia fake news detection is to accurately extract features of both text and visual information, as well as to mine the correlation between them. However, most of the existing methods merely fuse the features of different modal information without fully extracting intra- and inter-modal connections and complementary information. In this work, we learn physical tampered cues for images in the frequency domain to supplement information in the image space domain, and propose a novel multimodal frequency-aware cross-attention network (MFCAN) that fuses the representations of text and image by jointly modelling intra- and inter-modal relationships between text and visual information whin a unified deep framework. In addition, we devise a new cross-modal fusion block based on the cross-attention mechanism that can leverage inter-modal relationships as well as intra-modal relationships to complement and enhance the features matching of text and image for fake news detection. We evaluated our approach on two publicly available datasets and the experimental results show that our proposed model outperforms existing baseline methods.

Efficient cross-information fusion decoder for semantic segmentation

For fine-scale prediction tasks such as semantic segmentation, existing segmentation models cannot support detailed segmentation due to the difficulty of assigning deep feature semantics generated by the encoder to shallow features, thus making the segmentation of details ambiguous in semantic segmentation scenarios. In addition, high-precision models often require large quantities of computational resources. To solve the above problems, we design an efficient cross-information fusion decoder (ECFD). In the ECFD, we design a cross-information fusion block (CFB), and contextual information is used to assign semantic information to the shallow features in spatial domain, thus facilitating the classification of the details of segmented objects. To reduce the computational effort of the model, we choose the same decoder structure as used by the efficient SenFormer: the feature pyramid structure. Compared with SenFormer, ECFD-Swin-Large reduces the numbers of parameters and floating-point operations by 1/3, and achieves 83.61% and 64.98% of mIoU values for the benchmark datasets Cityscapes and Pascal Context, respectively, outperforming SenFormer, especially for in detailed segmentation. In addition, 69.19% is obtained on BDD100K. The code is publicly available at https://github.com/songyang-xiaobai/ECFD-main.

TSCA-Net: Transformer based spatial-channel attention segmentation network for medical images

Deep learning architectures based on convolutional neural network (CNN) and Transformer have achieved great success in medical image segmentation. Models based on the encoder–decoder framework like U-Net have been successfully employed in many realistic scenarios. However, due to the low contrast between object and background, various shapes and scales of objects, and complex background in medical images, it is difficult to locate targets and obtain better segmentation performance by extracting effective information from images. In this paper, an encoder–decoder architecture based on spatial and channel attention modules built by Transformer is proposed for medical image segmentation. Concretely, spatial and channel attention modules based on Transformer are utilized to extract spatial and channel global complementary information at different layers in U-shape network, which is beneficial to learn the detail features in different scales. To fuse better spatial and channel information from Transformer features, a spatial and channel feature fusion block is designed for the decoder. The proposed network inherits the advantages of both CNN and Transformer with the local feature representation and long-range dependency for medical images. Qualitative and quantitative experiments demonstrate that the proposed method outperforms against eight state-of-the-art segmentation methods on five publicly medical image datasets including different modalities, such as 80.23% and 93.56% Dice value, 67.13% and 88.94% Intersection over Union (IoU) value on the Multi-organ Nucleus Segmentation (MoNuSeg) and Combined Healthy Abdominal Organ Segmentation with Computed Tomography scans (CHAOS-CT) datasets.

A predictive model for daylight performance based on multimodal generative adversarial networks at the early design stage

In the early building design stage, machine learning-based surrogate models demonstrate significant advantages in rapidly evaluating design performance compared to conventional simulation software. However, due to network limitations, current models struggle to incorporate comprehensive building information, leading to undesired generalizability in adapting to design variations. This study proposes a multimodal Generative Adversarial Network (GAN)-based surrogate model, representing building features with combined multimodal data of images and vectors. Geometric features on the plan and façade are translated into grayscale maps, indicating the potential for daylight reception from each viewpoint, while material properties and weather features are extracted as vectors, such as reflectance, transmittance, sun position and sky conditions. To satisfy the demands of multimodal inputs, the multimodal feature fusion block, the vector-based feature encoding block, and the feature reinforcement block are introduced. These components aim to promote the deep integration of information from different modalities, balance the influences of features across different dimensional scales, and ensure the integrity of the information. The model's validity was verified on the RPLAN database, predicting illumination distribution at the whole-building hierarchy instead of the single-room hierarchy. The results indicate that the multimodal GAN has a Mean Squared Error of 8.129, a Mean Absolute Percentage Error of 0.135, and a Structural Similarity Index of 0.907 on the test set. Meanwhile, this approach saved 73.03% of computing time compared to simulation methods. Boasting strengths in generalizability, speed, and accuracy, the model provides effective support for daylight organization in the early stage of building design, without being confined by variable design scenarios.

Enhancing fine-detail image synthesis from text descriptions by text aggregation and connection fusion module

Synthesizing images with fine details from text descriptions is a challenge. The existing single-stage generative adversarial networks (GANs) fuse sentence features into the image generation process through affine transformation, which alleviate the problems of missing details and large computation from stacked networks. However, existing single-stage networks ignore the word features in the text description, resulting in a lack of detail in the generated image. To address this issue, we proposed a text aggregation module (TAM) to fuse sentence features and word features in a text by a simple spatial attention mechanism. Then we built a text connection fusion (TCF) block consisting mainly of gated recurrent unit (GRU) and up-sampled block. It can connect text features used in the up-sampled blocks to improve text utilization. Besides, to further improve the semantic consistency between text and the generated images, we introduce the deep attentional multimodal similarity model (DAMSM) loss, which monitors the similarity between text and improves semantic consistency. Experimental results prove that our method is superior to the state-of-the-art models on the CUB and COCO datasets, regarding both image fidelity and semantic consistency with the text.

Improved brain tumour segmentation using modified U-Net model with inception and attention modules on multimodal MRI images

ABSTRACT Brain tumours are currently recognised as one of the most dangerous diseases worldwide. Manual segmentation of brain tumours poses a challenging task heavily reliant on individual expertise, highlighting the need for the development of fully automated approaches. Deep learning techniques have gained significant popularity in recent years for medical image segmentation due to their exceptional accuracy and efficiency. In this study, we present a new brain tumour segmentation model based on the U-Net architecture. Our model incorporates an innovative inception module and a multi-level attention gate, effectively enhancing local feature expression and achieving superior segmentation accuracy. Additionally, we integrated a multiscale prediction fusion block to leverage global information across multiple scales. The performance of our proposed model was evaluated using the BRATS 2020 and BRATS 2018 datasets. On the BRATS 2018 dataset, our system achieved dice scores of 0.899, 0.851, and 0.838 for the whole tumour, tumour core, and enhancing tumour regions, respectively. Similarly, on the BRATS 2020 dataset, we obtained dice scores of 0.907, 0.842, and 0.829, respectively. Our proposed system demonstrated competitive performance across all brain tumour regions, as shown by the comparative analysis against state-of-the-art CNN models.