Choosing the right model for breast cancer classification depends on several factors, including the specific context, available datasets, and research goals and requirements. You can begin your exploration from here:
Chapter Deep Convolutional Comparison Architecture for Breast Cancer...
Article Classifying breast cancer using multi-view graph neural netw...
Recent advancements in breast cancer binary classification have introduced several innovative models and techniques that significantly enhance the accuracy and efficiency of diagnosis. Here are some of the new and notable models:
Dual-Activated Lightweight Attention ResNet50: This model incorporates an attention mechanism within the ResNet50 architecture. The attention mechanism helps the model focus on the most relevant areas of medical images, thereby improving the precision of identifying malignant and benign tissues. The model uses adaptive average pooling and dual activation functions (LeakyReLU and ReLU) to enhance feature extraction and adaptability, achieving high-precision results on datasets like BreakHis (https://ar5iv.labs.arxiv.org/html/2308.13150).
Multi-Branch Spectral Channel Attention Network (MbsCANet): MbsCANet utilizes a multi-branch approach to combine low and high-frequency features extracted through two-dimensional discrete cosine transform (DCT). This model aims to prevent the loss of phase information and provide richer contextual data for classification. It has demonstrated optimal results on the BreakHis dataset, significantly outperforming traditional spatial-domain models in terms of accuracy (
Article Multi-Branch Spectral Channel Attention Network for Breast C...
).
Deep Convolutional Comparison Architecture: This approach leverages multiple pre-trained convolutional neural networks (CNNs) such as Inception-V3, ResNet-50, VGG-16, and VGG-19. These models are fine-tuned for the specific task of breast cancer binary classification, offering robust feature extraction and classification capabilities. Comparative studies show that these models can achieve performance levels comparable to or better than human experts in histopathology image analysis (SpringerLink:
Chapter Deep Convolutional Comparison Architecture for Breast Cancer...
To cite my article, I need reference papers based on CNN models using 5 or 7 convolution layers for the breast cancer binary classification in the BreakHis histopathology image dataset.