Breast tissue image classification based on Semi-supervised Locality Discriminant Projection with Kernels.

Area of Science:

• Biomedical Engineering
• Machine Learning
• Medical Imaging

Background:

• Accurate breast tissue classification is crucial for computer-aided breast cancer diagnosis.
• Existing methods may not fully leverage available unlabeled data or handle nonlinear classification challenges effectively.

Purpose of the Study:

• To develop and present a novel Semi-supervised Locality Preserving Projections with Kernels (SLPPK) framework for breast cancer classification.
• To enhance the performance of Locality Preserving Projections (LPP) by incorporating semi-supervised learning and kernel methods.

Main Methods:

• The proposed method integrates semi-supervised learning into LPP, utilizing both labeled and unlabeled training samples along with side-information.
• The kernel trick is applied to SLPP to improve its capability in addressing nonlinear classification problems.
• A comprehensive framework for breast cancer classification using the developed SLPPK is established.

Main Results:

• Experiments conducted on four distinct breast tissue databases demonstrated the feasibility and effectiveness of the proposed SLPPK scheme.
• The integration of semi-supervised learning and kernel methods significantly improved classification performance compared to existing approaches.

Conclusions:

• The developed Semi-supervised Locality Preserving Projections with Kernels offers a promising advancement in computer-aided breast cancer diagnosis.
• This method provides a robust and effective approach for classifying breast tissues, particularly in handling complex, nonlinear data patterns.