Brain Tumor Detection Using Deep Transfer Learning Method

الموضوعات :

Alireza Fazelnia ¹ , Hassan Masoumi ² , Mohammad Fatehi ³ , Jasem Jamali ⁴

1 - Department of Biomedical Engineering, Kazerun Branch, Islamic Azad University, Kazerun, Iran
2 - Department of Biomedical Engineering, Kazerun Branch, Islamic Azad University, Kazerun, Iran
3 - Department of Electrical Engineering, Kazerun Branch, Islamic Azad University, Kazerun, Iran
4 - Department of Electrical Engineering, Kazerun Branch, Islamic Azad University, Kazerun, Iran

تاريخ الإرسال : 03 الإثنين , رجب, 1442 تاريخ التأكيد : 02 الأربعاء , رمضان, 1442 تاريخ الإصدار : 24 الأربعاء , محرم, 1443

الکلمات المفتاحية: deep learning, Brain tumor detection, transfer learning, convolution neural networks,

ملخص المقالة :

Accurate brain tumor MR images detection plays an important role in diagnosis and treatment decision making. The machine learning methods for classification only uses low-level or high-level features, to tackle the problem of classifications using some handcrafted features. Development on deep learning, transfer learning and deep convolution neural networks (CNNs) has shown great progress and has succeeded in the image classification task. Deep learning is very powerful for feature representation. In this study, deep transfer learning method for features extraction and detection is used that it does not use any handcrafted features, and needs minimal preprocessing. Transfer learning is a method of transferring information during training and testing. In this study, features extraction from images with pre-trained CNN method, namely, GoogLeNet, VGGNet and AlexNet, for tumor detection is used. The accuracy of tumor detection is 99.84%. The results show that our method, shows the best accuracy for detections tumor