Deep learning-driven macroscopic AI segmentation model for brain tumor detection via digital pathology: Foundations...

Area of Science:

• Digital Pathology
• Artificial Intelligence
• Oncology

Background:

• Accurate cancer delineation in digital pathology images is crucial for diagnosis and treatment planning.
• Manual annotation of cancerous regions is time-consuming and requires specialized expertise.
• Developing automated methods can improve efficiency and consistency in cancer diagnosis.

Purpose of the Study:

• To develop and validate a deep learning-based AI model for autonomous delineation of cancerous regions in H&E-stained digital pathology images.
• To create a robust and scalable AI training pipeline using DEEP:PHI and efficient data handling techniques.
• To facilitate the development of AI-driven cancer diagnosis technologies.

Main Methods:

• Utilized deep learning algorithms on a dataset of 187 H&E-stained images from a transgenic brain tumor model.
• Employed the DEEP:PHI platform to simplify AI model training and execution.
• Implemented Image Crop with Mask and patch generation techniques for data balancing and resource optimization.

Main Results:

• Successfully developed an AI model capable of autonomously segmenting cancerous areas in digital pathology images.
• The AI model provides guiding images, reducing the need for extensive neuropathologist assistance.
• A high-quality, large dataset was curated, supporting further AI development in cancer diagnosis.

Conclusions:

• Deep learning offers a powerful approach for automated cancer region delineation in digital pathology.
• The developed AI model enhances diagnostic efficiency and accuracy, supporting neuropathologists.
• The curated dataset and methodology contribute to advancing AI-based cancer diagnosis, including terahertz imaging applications.