Artificial intelligence to identify genetic alterations in conventional histopathology

Area of Science:

• Computational pathology
• Oncology
• Genomics

Background:

• Precision oncology requires identifying targetable molecular alterations in tumors.
• Current molecular testing is limited by cost and availability, especially for rare alterations.
• Artificial intelligence (AI) methods show potential for predicting genetic alterations from H&E slides.

Purpose of the Study:

• To systematically review the state-of-the-art AI methods for predicting molecular alterations from H&E tissue slides.
• To assess the performance and limitations of AI in this domain.
• To discuss the future implementation of AI-based surrogate tests in diagnostics.

Main Methods:

• Systematic literature review of studies published between 2017 and 2021.
• Analysis of AI methods predicting genetic alterations from H&E slides.
• Evaluation of predictability across different tumor types and genetic alterations.

Main Results:

• AI methods demonstrate reasonable performance across multiple tumor types for predicting genetic alterations from H&E slides.
• Specific alterations like FGFR, IDH, PIK3CA, BRAF, TP53, and DNA repair pathways are predictable.
• Few AI algorithms have undergone broad validation, and many alterations remain under-investigated or poorly predictable.

Conclusions:

• AI holds promise as a pre-screening tool to reduce the scope of genetic analyses in oncology.
• Further validation of AI algorithms is needed for clinical implementation.
• AI-based surrogate testing could enhance diagnostic workflows in precision oncology.