Deep learning-based phenotyping reclassifies combined hepatocellular-cholangiocarcinoma

Area of Science:

• Oncology
• Artificial Intelligence
• Genomics

Background:

• Primary liver cancer includes hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (ICCA), originating from distinct cell lineages.
• Combined hepatocellular-cholangiocarcinomas (cHCC-CCA) present mixed features, leading to diagnostic challenges and treatment uncertainties.

Purpose of the Study:

• To develop and validate a deep learning model for accurate phenotyping and classification of liver cancers.
• To reclassify biphenotypic cHCC-CCA tumors into HCC or ICCA subtypes using AI.
• To assess the correlation of AI-driven classification with clinical outcomes and molecular profiles.

Main Methods:

• Comprehensive deep learning-based phenotyping was applied to multiple patient cohorts.
• A series of 405 cHCC-CCA patients were analyzed using the deep learning model.
• Model predictions were validated against clinical outcomes, genetic alterations, and spatial gene expression data.

Main Results:

• The deep learning model achieved high performance in distinguishing HCC from ICCA.
• The model successfully reclassified cHCC-CCA tumors into HCC or ICCA categories.
• AI-based reclassification showed consistency with patient outcomes and molecular characteristics.

Conclusions:

• Deep learning offers a powerful tool for accurate diagnosis and subtyping of liver cancers, including rare biphenotypic types.
• AI-driven phenotyping can improve diagnostic accuracy for cHCC-CCA, aiding in treatment decisions.
• This approach holds potential to enhance clinical management and outcomes for patients with complex liver cancers.