Tri-Model Integration: Advancing Breast Cancer Immunohistochemical Image Generation through Multi-Method Fusion.

Area of Science:

• Computational pathology
• Digital pathology
• Medical image analysis

Background:

• Immunohistochemical (IHC) staining is vital for breast cancer diagnosis and treatment planning, assessing biomarkers like human epidermal growth factor receptor-2.
• Current IHC staining is costly and complex, motivating research into generating IHC images from Hematoxylin and Eosin (H&E) stained images via image-to-image (I2I) translation.

Purpose of the Study:

• To develop an improved method for generating high-quality synthetic IHC images.
• To enhance the reliability and accuracy of IHC image synthesis using an ensemble approach.

Main Methods:

• A novel approach combining three state-of-the-art I2I models was proposed.
• A Convolutional Neural Network was designed to fuse the outputs of three distinct I2I models into a single, consensus IHC image.
• The method utilizes a four-dimensional input comprising the RGB outputs from each individual model.

Main Results:

• The proposed ensemble method demonstrated superior performance compared to single-model approaches.
• Experiments on the BCI dataset showed improved Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index (SSIM) metrics.
• The fusion mechanism resulted in more robust and accurate synthetic IHC image generation.

Conclusions:

• The ensemble approach effectively leverages the strengths of individual I2I models for enhanced synthetic IHC image quality.
• This technique has the potential to reduce diagnostic costs and streamline the breast cancer assessment process.
• The developed code is publicly available for further research and application.