Identification and Classification of Prostate Cancer Identification and Classification Based on Improved Convolution...

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Prostate cancer diagnosis relies on pathologist scoring of tissue microarrays, which is time-consuming and subjective.
• Current methods for prostate cancer grading have low reproducibility due to inter-observer variability.
• Advancements in deep learning and computer vision offer potential for more objective pathology diagnostics.

Purpose of the Study:

• To develop an objective and repeatable computer-aided diagnosis system for prostate cancer.
• To improve the efficiency and accuracy of Gleason scoring using deep learning.
• To propose an improved U-Net based region segmentation model for prostate cancer tissue microarray analysis.

Main Methods:

• Utilized deep learning and computer vision techniques for pathology computer-aided diagnosis.
• Developed a region segmentation model based on an improved U-Net network.
• Fused deep and shallow layers using densely connected blocks and supervised multi-scale features.

Main Results:

• The improved U-Net model demonstrated reduced network parameters and enhanced computational efficiency.
• The proposed method achieved effective results on a fully annotated prostate cancer dataset.
• The AI-driven approach offers a more objective and repeatable alternative to manual Gleason scoring.

Conclusions:

• The developed deep learning model provides an objective and efficient method for prostate cancer diagnosis.
• The improved U-Net network shows promise in enhancing the reproducibility of Gleason scoring.
• This research contributes to the advancement of AI-powered tools in digital pathology for cancer detection.