Brain Tumor Detection in MRI Images Using ResNet۱۸ Convolutional Neural Network and Transfer Learning

In this study, a deep learning-based brain tumor detection model is proposed using a Convolutional Neural Network (CNN) architecture, specifically the ResNet۱۸ model. The aim is to develop an automated and accurate system capable of detecting brain tumors from MRI images, classifying them into two categories: “tumor present” and “no tumor.” To enhance performance and reduce the need for large-scale annotated medical datasets, the model employs transfer learning by initializing with pre-trained weights from the ImageNet dataset. The final fully connected layers of the ResNet۱۸ network are fine-tuned to adapt to the specific binary classification task. The MRI dataset is divided into training and test sets, and preprocessing steps such as image resizing and normalization are applied to standardize inputs. After training for ten epochs, the model achieved promising results, including an accuracy of ۸۴.۳۱%, a precision of ۷۹.۳۱%, a recall of ۹۲.۰۰%, and an F۱ score of ۸۵.۱۹%. These metrics indicate the model’s robustness in detecting tumors with high sensitivity and specificity. The experimental results suggest that the proposed method can effectively extract and interpret critical features from MRI scans, offering a reliable tool for assisting radiologists in early diagnosis and reducing the risk of human error in clinical decision-making.

Keywords:

Brain tumor detection , MRI Images , convolutional neural network , ResNet۱۸ , Transfer learning , Automatic classification , Deep Learning