Improved prostate cancer diagnosis using a modified ResNet50-based deep learning architecture.

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Prostate cancer is the most common cancer in men, influenced by various risk factors.
• Early detection improves patient outcomes, but balancing screening with overdiagnosis is challenging.
• Deep Learning (DL) offers potential for accurate and efficient prostate cancer diagnosis, especially with difficult imaging.

Purpose of the Study:

• To propose a Deep Learning (DL) model, the Prostate Cancer Detection Model (PCDM), for automatic prostate cancer diagnosis.
• To evaluate the clinical applicability of the PCDM for early detection and management in healthcare settings.
• To enhance prostate cancer detection accuracy and efficiency using advanced DL techniques.

Main Methods:

• Developed a modified ResNet50-based architecture incorporating faster R-CNN and dual optimizers.
• Trained the Prostate Cancer Detection Model (PCDM) on a large dataset of annotated medical images.
• Compared the performance of the PCDM against standard ResNet50 and VGG19 architectures.

Main Results:

• The proposed PCDM model demonstrated superior performance compared to ResNet50 and VGG19.
• Achieved high diagnostic metrics: 97.40% sensitivity, 97.09% specificity, 97.56% precision, and 95.24% accuracy.
• The model's architecture and training strategy significantly improved detection capabilities.

Conclusions:

• The Prostate Cancer Detection Model (PCDM) is a clinically applicable tool for aiding in the early detection of prostate cancer.
• DL algorithms, like the proposed PCDM, can significantly enhance the accuracy and efficiency of prostate cancer diagnosis.
• The model's high performance suggests its potential for real-world healthcare integration to improve patient management.