• Swetav Suman
• Ritesh Jha
• Vandana Bhattacharjee

This study explores three distinct deep learning-based feature extraction methodologies for brain tumor classification using MRI images. It combines features from Autoencoders, VGG16, and AlexNet, using Recursive Feature Elimination (RFE) and Principal Component Analysis (PCA) for dimensionality reduction. The extracted features were evaluated using K-Nearest Neighbors (KNN) and Support Vector Machines (SVM). The first method combines autoencoder and deep features; the second uses autoencoder features passed through AlexNet and VGG16 and then simply combines them; the third explores selective layer-wise feature fusion from VGG16 and AlexNet. Experimental results show that first method achieves the highest accuracy of 93.5%, and the hybrid method (Method 3) with selective deep features achieves the accuracy of 91%, with a precision of 89% and an F1-score of 92% and we get to know how deep layers work. The proposed methods highlight the potential of deep feature fusion and dimensionality reduction techniques for effective brain tumor detection.

Autoencoder, VGG16, AlexNet, Brain Tumor Classification, MRI, PCA, RFE, KNN, SVM

"Classification of Brain MRI using the Hypercolumn technique with a Convolutional Neural Network and a feature selection method", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.12, Issue 4, page no.m13-m18, April-2025, Available :http://www.jetir.org/papers/JETIR2504C02.pdf

"Classification of Brain MRI using the Hypercolumn technique with a Convolutional Neural Network and a feature selection method", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.12, Issue 4, page no. ppm13-m18, April-2025, Available at : http://www.jetir.org/papers/JETIR2504C02.pdf