Related Experiment Video
Updated: Jan 7, 2026
05:33
Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
Published on: July 11, 2025
741
Enhanced classification prostate cancer based on generative adversarial networks and integrated deep learning with
Wessam M Salama1, Moustafa H Aly2
1Department of Computer Engineering, Faculty of Engineering, Pharos University in Alexandria, Canal El Mahmoudia Street, Beside Green Plaza Complex 21648, Alexandria, Egypt.
Scientific Reports
|December 24, 2025
Summary
This study presents a secure, coverless image steganography method using a hybrid Generative Adversarial Network (GAN) and Support Vector Machine (SVM). The ViT-GAN-SVM model enhances security and diagnostic accuracy for medical imaging, particularly prostate cancer detection.
Area of Science:
- Computer Science
- Medical Imaging
- Cybersecurity
Background:
- Steganography techniques often require altering source images, increasing vulnerability to detection.
- Existing methods struggle to balance steganographic security with image quality and diagnostic utility.
- Diffusion Weighted Imaging (DWI) presents unique challenges for secure data embedding.
Purpose of the Study:
- To develop a secure, coverless image steganography technique resistant to steganalysis.
- To enhance diagnostic accuracy in medical imaging, specifically for prostate cancer detection using DWI.
- To evaluate a novel hybrid Generative Adversarial Network (GAN) and Support Vector Machine (SVM) model.
Main Methods:
- A hybrid Generative Adversarial Network (GAN) integrated with a Support Vector Machine (SVM) was employed.
- Feature extraction was performed using Deep Learning Models (DLMs) including EfficientNet-B4, DenseNet121, and Residual Network-18 (ResNet-18), combined with Vision Transformer (ViT).
- The model was trained and validated on a Diffusion Weighted Imaging (DWI) dataset for prostate cancer identification.
Main Results:
- The ViT-GAN-SVM model achieved superior steganographic quality with a Peak Signal-to-Noise Ratio (PSNR) of 45.87 dB and Structural Similarity Index (SSIM) of 0.98.
- Diagnostic accuracy metrics were exceptionally high: 99.78% accuracy, 99.85% sensitivity, 98.99% precision, and 99.85% F1-Score.
- Performance improvements in diagnostic metrics ranged from 5.55% to 6.36% compared to other models.
Conclusions:
- The ViT-GAN-SVM model offers a secure and effective coverless image steganography solution.
- The model demonstrates significant potential for improving diagnostic accuracy in medical tasks, particularly prostate cancer detection on DWI.
- This approach provides a robust method for embedding information while maintaining image integrity and enhancing security against steganalysis.
