Improving cervical cancer classification in PAP smear images with enhanced segmentation and deep progressive...

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Cervical cancer is a leading cause of death among women globally.
• Pap smear tests are crucial for early detection but suffer from human error and false positives.
• Machine learning and deep learning offer solutions for automated image analysis in diagnostics.

Purpose of the Study:

• To develop and evaluate a deep learning model for accurate cervical cancer cell classification.
• To improve the precision of cervical cancer screening using automated image analysis.
• To assist medical professionals in diagnosing cervical cancer stages from cytology and colposcopy images.

Main Methods:

• Utilized the ResNet-50 deep learning architecture for image classification.
• Implemented a three-step method: preprocessing, k-means clustering for segmentation, and cell classification.
• Assessed model performance using metrics including accuracy, precision, sensitivity, specificity, and kappa score.

Main Results:

• The ResNet-50 model effectively pinpointed infected cervical regions through spatial k-means clustering and preprocessing.
• A progressive learning technique was applied across four stages with increasing image resolutions (64x64 to 1024x1024).
• Achieved high performance with 97.4% accuracy and approximately 98% kappa score in analyzing Pap smear tests.

Conclusions:

• The proposed deep learning model, ResNet-50, demonstrates significant effectiveness in analyzing Pap smear images.
• The AI-driven approach offers a valuable tool for the timely and accurate detection of cervical cancer.
• This technology has the potential to reduce diagnostic errors and improve patient outcomes in cervical cancer screening.