Multimodal NASH prognosis using 3D imaging flow cytometry and artificial intelligence to characterize liver cells.

Area of Science:

• Cell biology
• Medical imaging
• Computational pathology

Background:

• Non-alcoholic steatohepatitis (NASH) pathogenesis involves complex cellular interactions.
• Understanding hepatic stellate cell (HSC) and liver endothelial cell (LEC) roles is crucial for NASH development.
• Current methods lack single-cell resolution for detailed cellular analysis in NASH.

Purpose of the Study:

• To characterize HSC and LEC morphology in early-stage NASH using 3D imaging flow cytometry (3D-IFC).
• To identify cellular and texture parameters associated with NASH progression.
• To develop a machine learning model for classifying NASH-related cell types.

Main Methods:

• Obtained HSCs and LECs from biopsy-proven early-stage NASH (F2-F3) subjects and healthy controls.
• Utilized 3D-IFC with transmission and side-scattered images for single-cell characterization.
• Applied 3D digital reconstructions and analyzed morphometric and texture parameters.
• Developed a convolutional neural network (CNN) autoencoder using label-free imaging data.

Main Results:

• Identified spatially resolved cellular and texture parameters that correlate with NASH disease progression.
• Demonstrated regression of specific morphometric parameters with increasing disease severity.
• Achieved superior cell classification performance for NASH-related cell types compared to conventional machine learning.

Conclusions:

• 3D-IFC provides high-resolution insights into cellular changes during NASH development.
• Morphometric and texture analysis can serve as biomarkers for NASH progression.
• CNN-based analysis of label-free images offers a powerful tool for identifying key cell types in NASH.