{"title":"Brain Tumor Type Recognition Algorithm Fused with Double Residual Structure and Attention Mechanism","authors":"Shubin Yang, Feng-ge Wang, Chunlin Dong","doi":"10.1109/AICIT55386.2022.9930300","DOIUrl":null,"url":null,"abstract":"Brain tumor type recognition plays an important role in brain tumor diagnosis, In order to solve the problems of low accuracy and poor real-time performance of existing recognition algorithms, this paper combines double residual structure and attention mechanism to accurately and real-time brain tumor type recognition based on ResNet34. The first step is to reduce the number of parameters while enhancing the algorithm’s ability to extract multi-scale features by replacing the original convolution in the algorithm with a multi-scale convolution to avoid the loss of too large or too small features; Secondly, the loss of features due to deep layers is mitigated by changing the original structure to a double residual structure to prevent degradation of the algorithm. Finally, in order to enhance the weight of important features, an attention mechanism module is embedded in the sidechain residual structure to avoid the impact of redundant features on recognition accuracy. The extracted features are then passed to a classifier for accurate identification of the type of brain tumour. The improved algorithm was validated under the kaggle public dataset, and its accuracy reached 97.4% and the number of parameters was 7. 59M, which is 2% more accurate than the original model and 33% of the number of parameters, and outperformed some existing classical and mainstream algorithms. The experimental results show that the algorithm can accurately and quickly identify the type of brain tumour, which can help doctors in the subsequent treatment.","PeriodicalId":231070,"journal":{"name":"2022 International Conference on Artificial Intelligence and Computer Information Technology (AICIT)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 International Conference on Artificial Intelligence and Computer Information Technology (AICIT)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/AICIT55386.2022.9930300","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Brain tumor type recognition plays an important role in brain tumor diagnosis, In order to solve the problems of low accuracy and poor real-time performance of existing recognition algorithms, this paper combines double residual structure and attention mechanism to accurately and real-time brain tumor type recognition based on ResNet34. The first step is to reduce the number of parameters while enhancing the algorithm’s ability to extract multi-scale features by replacing the original convolution in the algorithm with a multi-scale convolution to avoid the loss of too large or too small features; Secondly, the loss of features due to deep layers is mitigated by changing the original structure to a double residual structure to prevent degradation of the algorithm. Finally, in order to enhance the weight of important features, an attention mechanism module is embedded in the sidechain residual structure to avoid the impact of redundant features on recognition accuracy. The extracted features are then passed to a classifier for accurate identification of the type of brain tumour. The improved algorithm was validated under the kaggle public dataset, and its accuracy reached 97.4% and the number of parameters was 7. 59M, which is 2% more accurate than the original model and 33% of the number of parameters, and outperformed some existing classical and mainstream algorithms. The experimental results show that the algorithm can accurately and quickly identify the type of brain tumour, which can help doctors in the subsequent treatment.