PAIP 2020: Microsatellite instability prediction in colorectal cancer

Area of Science:

• Computational pathology
• Oncology
• Genomics

Background:

• Microsatellite instability (MSI) is crucial for colorectal cancer prognosis and treatment selection.
• MSI-high status indicates a favorable prognosis in Stages II/III and predicts response to immunotherapy in Stage IV.
• Accurate MSI status determination is vital for personalized colorectal cancer treatment protocols.

Purpose of the Study:

• To evaluate the effectiveness of artificial intelligence (AI) in predicting MSI status from colorectal cancer slide images.
• To summarize the methodologies employed in the Pathology Artificial Intelligence Platform (PAIP) 2020 challenge.
• To demonstrate the utility of digital pathology in assessing MSI characteristics in colorectal cancer.

Main Methods:

• The PAIP 2020 challenge tasked AI researchers with predicting MSI status (MSI-high vs. microsatellite-stable) from colorectal cancer histology slides.
• A secondary task involved tumor area segmentation.
• Deep learning models, primarily convolutional neural networks like EfficientNet and UNet, were utilized by participants.

Main Results:

• 23 teams submitted results, with the top system achieving an F1 score of 0.9231.
• Seven of the top 10 teams shared their algorithms, predominantly based on deep learning.
• The challenge highlighted the potential of AI in analyzing histopathological features related to MSI.

Conclusions:

• AI models, particularly deep learning approaches, show high accuracy in predicting MSI status from colorectal cancer digital slides.
• Digital pathology combined with AI offers a promising tool for identifying MSI characteristics.
• These findings support the integration of AI in pathology workflows for improved cancer diagnosis and treatment guidance.