Classification of Multiclass Histopathological Breast Images Using Residual Deep Learning

Area of Science:

• Computational pathology
• Artificial intelligence in medical imaging
• Oncology

Background:

• Histopathological investigation is crucial for cancer diagnosis but requires significant pathologist expertise and time.
• Breast cancer diagnosis relies heavily on histopathological images, which present challenges due to complex tissue textures.
• Existing computer-assisted diagnosis (CAD) systems need enhancement to support radiologists effectively.

Purpose of the Study:

• To develop an advanced AI-driven system for accurate and efficient breast cancer histopathology image classification.
• To address the challenge of limited labeled data in training deep learning models for medical image analysis.
• To create a self-training deep learning model that functions as a reliable second opinion for radiologists.

Main Methods:

• Implementation of a novel self-training learning method utilizing a deep learning neural network.
• Integration of residual learning principles to enhance model performance.
• Development and training of the proposed model from scratch for breast cancer histopathology classification.

Main Results:

• The proposed self-training deep learning model effectively classifies breast cancer histopathology images.
• The method overcomes the limitation of requiring large annotated datasets for training.
• Demonstrated potential for improved accuracy and efficiency in histopathological diagnoses.

Conclusions:

• The developed AI system shows promise in supporting pathologists and improving breast cancer diagnosis.
• Self-training deep learning offers a viable solution for data-scarce scenarios in computational pathology.
• This approach can enhance the reliability and speed of computer-assisted diagnosis in medical imaging.