Performance Evaluation of Deep Learning-Based Prostate Cancer Screening Methods in Histopathological Images:...

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Prostate cancer (PCa) is a leading cause of cancer death globally, with projected mortality doubling by 2040, especially in low-resource settings.
• Deep learning (DL) models, particularly convolutional neural networks (CNNs), show promise for PCa detection in histopathological images.
• Current DL approaches often prioritize accuracy, demanding significant computational resources for training and inference, which is a barrier in resource-limited areas.

Purpose of the Study:

• To evaluate the performance of state-of-the-art DL models for PCa detection.
• To introduce and assess PROMETEO, a custom-designed DL architecture for PCa screening.
• To establish a benchmark for comparing PCa detection model efficiency and effectiveness.

Main Methods:

• A novel benchmark was developed to assess PCa detection models.
• Performance metrics were measured for current state-of-the-art DL models.
• The proposed PROMETEO architecture was compared against existing models using the benchmark.

Main Results:

• Comprehensive performance comparison of various DL models for PCa detection was conducted.
• PROMETEO demonstrated competitive or superior performance in specific application contexts.
• The study highlights the trade-offs between model complexity, computational requirements, and diagnostic accuracy.

Conclusions:

• Dedicated, optimized DL models can be highly effective for specific applications like PCa screening.
• Reducing prediction time and computational load is critical for widespread adoption in resource-limited healthcare settings.
• Future research should focus on developing efficient DL architectures tailored for specific diagnostic tasks to improve accessibility and timeliness of cancer detection.