Unraveling the power of radiomics: prediction and exploration of lymph node metastasis in stage T1/2 esophageal squamous cell carcinoma

Yu Zhang ¹, Long Liu ², Mengyu Han ¹

⁰Department of Radiation Therapy, The First Affiliated Hospital of Anhui Medical University, Hefei, Anhui, China.

Npj Precision Oncology +

|

June 16, 2025

View abstract on PubMed

Summary

Machine learning radiomics accurately predicts lymph node metastasis (LNM) in early esophageal squamous cell carcinoma (ESCC). This approach aids treatment planning and reveals molecular pathways involved in ESCC LNM.

Area Of Science

Oncology
Radiology
Bioinformatics

Background

Accurate lymph node metastasis (LNM) assessment is crucial for T1/2-stage esophageal squamous cell carcinoma (ESCC) treatment planning.
Current diagnostic methods for LNM in ESCC face challenges with accuracy and understanding metastatic mechanisms.

Purpose Of The Study

To predict LNM in T1/2-stage ESCC using machine learning-based radiomics.
To elucidate the biological underpinnings of LNM in ESCC.

Main Methods

Retrospective analysis of 374 surgically treated ESCC patients from two centers.
Development of an optimal radiomics score using six machine-learning algorithms, with a focus on decision tree (DT) models.
Bioinformatics analysis and experimental validation to identify key pathways and genes associated with LNM.

Main Results

The DT-based radiomics model achieved high predictive performance: AUCs of 0.933 (training), 0.887 (validation), and 0.845 (test).
Bioinformatics analysis identified tumor-lymphatic invasion pathways as significant.
EFNA1 was highlighted as a potential key regulator in ESCC LNM.

Conclusions

Machine learning-driven radiomics offers significant clinical utility for predicting LNM in early-stage ESCC.
The study provides novel insights into the molecular mechanisms driving LNM in ESCC, particularly implicating EFNA1.