Breast cancer histopathological image classification using a hybrid deep neural network

Area of Science:

• * Computational pathology
• * Medical imaging analysis
• * Artificial intelligence in oncology

Background:

• * Histopathological diagnosis is the gold standard for cancer detection but is time-consuming and subjective.
• * Increasing workload and image complexity necessitate automated analysis methods.
• * Accurate classification of breast cancer histopathology is crucial for effective treatment.

Purpose of the Study:

• * To develop an automated method for precise histopathological image analysis.
• * To improve the accuracy and efficiency of breast cancer diagnosis.
• * To propose a hybrid deep neural network integrating convolutional and recurrent architectures.

Main Methods:

• * Development of a hybrid convolutional and recurrent deep neural network.
• * Feature extraction from multilevel representations of histopathological image patches.
• * Preservation of short-term and long-term spatial correlations between image patches.

Main Results:

• * Achieved an average accuracy of 91.3% for 4-class breast cancer classification.
• * Demonstrated superior performance compared to state-of-the-art methods.
• * Released a diverse dataset of 3771 breast cancer histopathological images.

Conclusions:

• * The proposed hybrid deep learning model offers a precise and efficient approach to histopathological image classification.
• * The new method enhances diagnostic accuracy, potentially reducing pathologist subjectivity.
• * The publicly available dataset supports further research in breast cancer image analysis.