JBT Journal
Advanced

Raqamli Transformatsiya va Sun’iy Intellekt

HYBRID DEEP LEARNING APPROACH FOR BREAST CANCER DETECTION USING MRI IMAGES COMBINING CNN WITH SVM AND RANDOM FOREST FOR ENHANCED ACCURACY

Sayyora Iskandarova1, Sayyora Iskandarova1

1 Tashkent university of information technologies named after Muhammad al-Khwarizmi .

Abstract

This study proposes a hybrid deep learning framework for detecting breast cancer from MRI images by integrating Convolutional Neural Networks (CNNs) with Support Vector Machines (SVM) and Random Forest (RF) algorithms. The methodology leverages the feature extraction capabilities of CNNs to identify intricate patterns in MRI images, followed by classification using SVM and RF to enhance diagnostic accuracy. The process begins with preprocessing steps, including denoising and normalization, to improve image quality. A pre-trained CNN model, EfficientNet-B4, extracts high-level features from the images, which are then fed into SVM and RF for classification. The hybrid model was trained and validated on a dataset of 1,500 MRI images, achieving a validation accuracy of 96.8%, sensitivity of 94.5%, and specificity of 95.2%. Compared to standalone CNN (92% accuracy), SVM (87%), and RF (85%), the hybrid approach significantly improves performance, though it requires more computational time (1,300 seconds vs. 950 and 800 seconds for SVM and RF, respectively). The results demonstrate the model’s ability to generalize well, with a low validation loss of 0.02, indicating minimal overfitting. This hybrid framework offers a robust tool for clinical applications, assisting radiologists in early breast cancer diagnosis and potentially improving patient outcomes. Future work will focus on optimizing computational efficiency and validating the model on larger, diverse datasets to ensure real-world applicability. The combination of CNN with SVM and RF highlights the potential of hybrid algorithms in medical imaging.

Views: 82 Published: 2025-06-04 Pages: 20-25 Volume: 3 Issue: 3

References

No references listed for this article.