Citation

Rahman, Sayedur and Alam, Touhidul and Aziz, Shusmita Anjum and Rahman, Md Arifur and Haque, B M Taslimul and Liew, Tze Hui (2025) Advancing Prostate Cancer Diagnosis with DCGAN-Generated Synthetic Histopathology Images. In: 4th International Conference on Smart Cities, Automation, and Intelligent Computing Systems, ICON-SONICS 2025, 14 October 2025 - 17 October 2025, Hybrid, Malacca.

Text IEEE Xplore Full-Text PDF_18.pdf - Published Version Restricted to Repository staff only Download (1MB)

Abstract

Deep learning models for prostate cancer diagnosis are often hindered by the scarcity and imbalance of medical imaging datasets. This study addresses this challenge by developing a Deep Convolutional Generative Adversarial Network (DCGAN) to produce high-fidelity synthetic histopathology images. Leveraging the Cancer-Net PCa dataset, our DCGAN, featuring over 51 million parameters, was trained for 200 epochs to generate realistic images. The model’s performance was rigorously evaluated, demonstrating stable training dynamics with generator and discriminator losses converging at 0.6526 and 0.6953, respectively. A significant improvement in image quality was confirmed by the Frechet Inception Distance (FID) score, ´ which dropped from an initial value of approximately 400 to 211.89. While the model showed high precision, limitations in image diversity were identified, indicating an area for future enhancement. Futhermore, we trained several transfer learning models on the augmented dataset using the proposed DCGAN model where the EfficientNetV2 performed the best at 97.02% accuracy and an improvement of 5.42% over the original dataset, demonstrating a state-of-the-art result among the other established GAN models. The results affirm the potential of DCGANs to effectively augment medical datasets, paving the way for more robust and accurate diagnostic tools in the fight against prostate cancer.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Prostate cancer, synthetic image generation
Subjects:	R Medicine > R Medicine (General) > R858-859.7 Computer applications to medicine. Medical informatics
Divisions:	Faculty of Information Science and Technology (FIST)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	17 Mar 2026 05:31
Last Modified:	17 Mar 2026 05:31
URII:	http://shdl.mmu.edu.my/id/eprint/15483

Downloads

Edit (login required)