An efficient brain tumor classification system with accurate recognition and complementary downsampling blocks
- 1College of Information Engineering, Henan University of Science and Technology, Luoyang, China
- 2The First Affiliated Hospital, and College of Clinical Medicine of Henan University of Science and Technology, Luoyang, China
Abstract
Due to the substantial impact of brain tumors on human health, precise identification of tumor types is vital for patient prognosis and guiding treatment decisions. Magnetic Resonance Imaging (MRI) technology, owing to its non-invasive nature and high resolution, is indispensable in accurately classifying brain tumor types. However, manual interpretation of MRI images is not only time-consuming and labor-intensive but also susceptible to subjective bias. Hence, many automatic brain tumor diagnosis systems utilizing deep learning techniques have been developed. This paper introduces an advanced modularized Convolutional Neural Network (CNN) that incorporates ConvDAtt block, Complementary Information (CI) downsampling block, and Spatial Pyramid Pooling (SPP) block. Firstly, the ConvDAtt block combines convolution, depth-separable convolution and Coordinated Attention (CA) mechanism. The module alleviates the gradient vanishing problem in deep neural network training, enhancing the training stability and performance of the model. In the ConvDAtt block, the CA mechanism can not only capture information across channels, but also sense direction and position sensitive information, improving the accuracy of the model’s identification of objects of interest. Secondly, the CI block is a downsampling block with convolution and pooling functions. In the CI block, there is a complementary relationship between the pooling layer and the convolution layer, which is mainly reflected in two aspects, feature extraction and dimensionality reduction, enlargement of receptive field. Moreover, the complementary relationship can solve the problem of information loss in the process of downsampling, while also significantly reducing computational cost. Thirdly, the SPP block extracts features from different scales of the input data and integrates these features, thus enhancing the performance of the network, and enhancing the robustness of the model. The combination of ConvDAtt block, CI block and SPP block can effectively extract the information features and discriminant features of brain tumor MRI images, which has the advantages of parameter reduction, stable training and clear structure.
Related
- HayabusaNet: Hybrid attention-based multiscale fusion CNN for accurate and efficient brain tumor classification in MRI scans
- Edge context fusion transformer for accurate and efficient photovoltaic electroluminescence defect detection
- EPRA U-Net: an efficient pyramid residual attention framework for accurate infarct segmentation in diffusion-weighted MRI.
- Feature-aware attention convolutional neural network for efficient pulmonary fibrosis classification from chest CT images with FPGA deployment
- Efficient Alzheimer’s disease classification with robust rank pooling
BibTeX
@article{zhang2026efficient,
title = {{An efficient brain tumor classification system with accurate recognition and complementary downsampling blocks}},
author = {Jincan Zhang and Chuanqi Cai and XingHua Tan and Wenna Chen and Xin Zhao and Ganqin Du and Hongwei Jiang},
journal = {PLOS One},
year = {2026},
month = sep,
doi = {10.1371/journal.pone.0343457},
url = {https://doi.org/10.1371/journal.pone.0343457},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - An efficient brain tumor classification system with accurate recognition and complementary downsampling blocks
AU - Jincan Zhang
AU - Chuanqi Cai
AU - XingHua Tan
AU - Wenna Chen
AU - Xin Zhao
AU - Ganqin Du
AU - Hongwei Jiang
PY - 2026
DA - 2026/09/30
JO - PLOS One
DO - 10.1371/journal.pone.0343457
UR - https://doi.org/10.1371/journal.pone.0343457
AB - Due to the substantial impact of brain tumors on human health, precise identification of tumor types is vital for patient prognosis and guiding treatment decisions. Magnetic Resonance Imaging (MRI) technology, owing to its non-invasive nature and high resolution, is indispensable in accurately classifying brain tumor types. However, manual interpretation of MRI images is not only time-consuming and labor-intensive but also susceptible to subjective bias. Hence, many automatic brain tumor diagnosis systems utilizing deep learning techniques have been developed. This paper introduces an advanced modularized Convolutional Neural Network (CNN) that incorporates ConvDAtt block, Complementary Information (CI) downsampling block, and Spatial Pyramid Pooling (SPP) block. Firstly, the ConvDAtt block combines convolution, depth-separable convolution and Coordinated Attention (CA) mechanism. The module alleviates the gradient vanishing problem in deep neural network training, enhancing the training stability and performance of the model. In the ConvDAtt block, the CA mechanism can not only capture information across channels, but also sense direction and position sensitive information, improving the accuracy of the model’s identification of objects of interest. Secondly, the CI block is a downsampling block with convolution and pooling functions. In the CI block, there is a complementary relationship between the pooling layer and the convolution layer, which is mainly reflected in two aspects, feature extraction and dimensionality reduction, enlargement of receptive field. Moreover, the complementary relationship can solve the problem of information loss in the process of downsampling, while also significantly reducing computational cost. Thirdly, the SPP block extracts features from different scales of the input data and integrates these features, thus enhancing the performance of the network, and enhancing the robustness of the model. The combination of ConvDAtt block, CI block and SPP block can effectively extract the information features and discriminant features of brain tumor MRI images, which has the advantages of parameter reduction, stable training and clear structure.
SN - 1932-6203
ER -