Attention-Guided Multi-Branch Convolutional Neural Network for Mitosis Detection From Histopathological Images

Area of Science:

• Digital Pathology
• Computational Biology
• Oncology

Background:

• Manual mitotic counting in breast cancer histopathology is crucial for assessing invasiveness but is labor-intensive and time-consuming.
• Accurate mitosis detection is vital for reliable cancer grading and treatment planning.

Purpose of the Study:

• To develop and validate a fast and accurate automated method for mitosis detection in histopathological images.
• To improve the efficiency and objectivity of mitosis screening in breast cancer diagnosis.

Main Methods:

• Utilized deep convolutional neural networks (CNNs) for high-level feature extraction of mitotic cells.
• Incorporated spatial attention modules to refine and enhance mitotic feature representation.
• Employed multi-branch classification subnets for the final screening of mitotic candidates.

Main Results:

• The proposed method demonstrated superior performance in automatically identifying mitotic candidates from histological sections.
• Achieved state-of-the-art detection results on the ICPR 2012 Mitosis Detection Competition dataset.
• The automated approach offers a significant improvement over existing manual and computational methods.

Conclusions:

• The developed deep learning model provides an efficient and accurate solution for automated mitosis detection.
• This technology has the potential to streamline pathological workflows and enhance breast cancer assessment.
• The availability of the code facilitates further research and clinical application.