A multi-patch-based deep learning model with VGG19 for breast cancer classifications in the pathology images

Area of Science:

• Medical Imaging
• Computational Pathology
• Artificial Intelligence in Oncology

Background:

• Accurate breast cancer stratification is essential due to distinct subtypes and prognoses.
• Current gene expression methods have limitations in capturing tissue complexity.
• Effective detection and classification in medical imaging are crucial for patient outcomes.

Purpose of the Study:

• To introduce a novel method, MPa-DCAE, for breast cancer detection and classification in histopathology images.
• To leverage a multi-patch deep convolutional auto-encoder (DCAE) framework combined with VGG19.
• To enhance the accuracy and efficiency of breast cancer diagnosis through automated image analysis.

Main Methods:

• Utilized a multi-patch approach for localized feature learning from pathology images.
• Integrated VGG19 for hierarchical feature extraction within a DCAE framework.
• Employed unsupervised feature learning via the auto-encoder for adaptability.

Main Results:

• The MPa-DCAE model demonstrated superior performance compared to existing methods.
• Achieved high precision (97.96-97.99%), recall (94.85-97.2%), and accuracy (98.36-98.95%) on CBIS-DDSM and MIAS datasets.
• Validated model robustness and potential for clinical computer-assisted diagnosis.

Conclusions:

• The MPa-DCAE model offers an effective, automated solution for breast cancer diagnosis in histopathology.
• High accuracy and generalizability suggest significant potential for clinical practice.
• This approach may improve patient care in histopathology-based breast cancer diagnostics.