| 作成者 |
|
|
|
| 本文言語 |
|
| 出版者 |
|
|
|
| 発行日 |
|
| 収録物名 |
|
| 巻 |
|
| 開始ページ |
|
| 終了ページ |
|
| 会議情報 |
|
| 出版タイプ |
|
| アクセス権 |
|
| Crossref DOI |
|
| 権利関係 |
|
| 権利関係 |
|
| 概要 |
A brain tumor is a cancerous disease that can be fatal. The classification of brain tumors in clinical methods is time-consuming and error-prone. This paper proposes three deep convolutional neural ne...twork (CNN)-based architectures, ConVGXNet, ConResXNet, and ConIncXNet, by utilizing the concept of transfer learning. The proposed models extricate diversified features from magnetic resonance images (MRI). Many multi-class brain MRIs corresponding to glioma, pituitary, and no tumor patients are used to train the proposed models. This study also carries out data pre-processing and data augmentation for the models' effectiveness. A proper training configuration monitors the performance of the models. ConIncXNet is the most accurate of the architectures proposed. The accuracy of the ConIncXNet architecture is 97 percent. The models are also evaluated based on weighted F1 score, precision, and recall. This proposed CNN architecture can improve the classification of brain tumors in medical image diagnosis systems.続きを見る
|
Mendeley出力
