Leveraging deep learning to discover interpretable cellular spatial biomarkers for prognostic predictions based on hepatocellular carcinoma histology

Huijuan Hu ¹, Tianhua Tan ^2,3, Yerong Liu ¹

⁰State Key Laboratory of Quantitative Synthetic Biology, Shenzhen Institute of Synthetic Biology, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, PR China.

The Journal of Pathology. Clinical Research +

|

June 13, 2025

View abstract on PubMed

Summary

Researchers developed a computational method to analyze cell spatial organization in hepatocellular carcinoma (HCC) tumors. This analysis identified novel spatial biomarkers from H&E-stained images that predict patient survival and improve prognostic accuracy.

Area Of Science

Oncology
Computational Pathology
Bioinformatics

Background

The tumor microenvironment (TME) spatial structure offers insights into disease progression.
Identifying cell spatial organization linked to patient prognosis in HCC is challenging.

Purpose Of The Study

To develop a computational pipeline for quantifying cell spatial distribution features in HCC.
To identify spatial features correlating with patient survival and prognostic accuracy.

Main Methods

Deep learning for cell segmentation and recognition in H&E-stained HCC images.
Systematic quantification of spatial distribution features for tumor cells, stromal cells, and lymphocytes.
Validation in two independent HCC patient cohorts (TCGA and Beijing Hospital).

Main Results

Six cellular spatial features significantly correlated with overall survival in both cohorts.
Features like StrDiv-M, CellDis-MED, and CvStrDis-MED improved patient stratification.
Combining spatial features with microvascular invasion enhanced prognostic accuracy.

Conclusions

Quantifying cellular spatial organization in HCC TME reveals novel prognostic biomarkers.
These biomarkers can evaluate tumor prognosis and guide clinical treatment strategies.
Findings support further mechanistic studies on spatial organization in HCC TME.

Keywords:

Delaunay network cell classification cell segmentation cellular diversity coefficient of variation