Classification of breast tumors by using a novel approach based on deep learning methods and feature selection

Area of Science:

• Medical image analysis
• Computational pathology
• Artificial intelligence in oncology

Background:

• Early cancer diagnosis is crucial for effective treatment and patient outcomes.
• Manual histopathological image analysis by experts is time-consuming and prone to errors.
• Computer-aided systems offer potential for enhanced accuracy and efficiency in cancer detection.

Purpose of the Study:

• To develop and evaluate a computer-aided system for accurate cancer tissue classification.
• To improve the efficiency and reduce errors in cancer diagnosis using deep learning.
• To investigate the efficacy of a novel feature selection method in conjunction with deep learning models for histopathological image analysis.

Main Methods:

• Employed deep learning models including ResNet-50, GoogLeNet, InceptionV3, and MobilNetV2.
• Implemented a novel feature selection method to enhance classification performance.
• Utilized both a local binary class dataset and the multi-class BACH dataset for evaluation.

Main Results:

• Achieved 98.89% classification accuracy on the local binary class dataset.
• Attained 92.17% classification accuracy on the multi-class BACH dataset.
• Demonstrated superior performance compared to existing methods in the literature.

Conclusions:

• The proposed feature selection integrated with deep learning effectively detects and classifies cancer types in tissues.
• The developed methods show high accuracy and efficiency, supporting their clinical utility.
• Computer-aided histopathological analysis holds significant promise for advancing cancer diagnostics.