1Department of Computer Science and Engineering, Government College of Engineering, Kalahandi, Bhawanipatna, Odisha, 766003, India. Email: drnayak@gcekbpatna.ac.in, gbehera@gcekbpatna.ac.in, bkswain@gcekbpatna.ac.in, akbhoi@gcekbpatna.ac.in
2Department of Mechanical Engineering, Government College of Engineering, Kalahandi, Bhawanipatna, Odisha, 766003, India. Email: dkbagal@gcekbpatna.ac.in
Background: Every day, the cancer cases increase rapidly. Early detection is the crucial step for the save of life. In this paper, we proposed a novel hybrid model that consist of VGG16 convolutional neural network (CNN) with fuzzy membership function and wavelet transforms to identify automatically tumor from MRI images. VGG16 learn detailed pattern from MRI Image, on other hand fuzzy membership function enhance the quality of image and then finetuning VGG16 model through wavelet transfer learning to differentiating between tumor and non-tumor cases.
Methodology: To justify the model's performance, we carried out several experiments on real world dataset and found that the model accuracy is improved approximately 3% compared to the state-of-art model. The outcome competently proves the potential of the proposed model precise brain tumor classification. In this paper, authors proposed the stimulating task of brain tumor detection from MRI images using 1225 brain tumor MRI images. This study explores the use of the modified VGG16 convolutional neural network (CNN) model combined with fuzzy membership function and wavelet transforms to automatically identify brain tumors from MRI image. Wavelet transforms break down MRI images into different frequency components, highlighting important features that may indicate the presence of a tumour. By Integrating these processed images allows the VGG16 model to learn detailed patterns differentiating between tumour and non-tumor cases. The research involved preprocessing MRI images to enhance their quality by fuzzy INT function and applying transfer learning techniques to fine-tune the VGG16 model using yes and no data set.
Results: Performance metrics such as accuracy, precision, recall, and F1-score were used to evaluate the model's effectiveness. The results demonstrated an accuracy of approximately 99.18%, indicating the model's potential in supporting healthcare professionals in diagnosing brain tumors.
Conclusion: These outcomes competently prove the potential of the proposed model precise brain tumor classification.
Keywords: Brain tumor; Convolutional Neural Network; Deep learning; MRI images; Transfer learning; VGG16; Wavelet transform.
How to cite this article: Nayak DR, Behera G, Swain BK, Bhoi AK, Bagal DK. Enhancing Brain Tumor Classification using Berkeley Wavelet Transformation and Improved CNN. Int J Drug Deliv Technol. 2026;16(6s): 1044-1054. DOI: 10.25258/ijddt.16.6s.136
Source of support: Nil.
Conflict of interest: Nil