Clinical performance of a machine learning-based model for detecting lymph node metastasis in papillary thyroid...

Area of Science:

• Oncology
• Genomics
• Artificial Intelligence

Background:

• Papillary thyroid carcinoma (PTC) is a common endocrine malignancy.
• Lymph node metastasis (LNM) in PTC complicates treatment and increases recurrence risk.
• Current preoperative LNM detection methods like ultrasound have limitations, potentially leading to overtreatment.

Purpose of the Study:

• To develop and validate a non-invasive, AI-driven predictive model for LNM in PTC using gene expression data.
• To identify key genes and pathways associated with LNM in PTC.
• To provide a tool for improved preoperative assessment of LNM to guide surgical management.

Main Methods:

• Gene expression data from 157 PTC patients were analyzed to develop a predictive model.
• A random forest algorithm was employed, focusing on RPS4Y1, PKHD1L1, and CRABP1 genes.
• The model's performance was validated using qRT-PCR in an independent cohort of 807 patients from multiple centers.

Main Results:

• The AI model demonstrated high accuracy in predicting LNM, with an AUROC of 0.992 in training and 0.911-0.953 in external validation.
• RPS4Y1 was identified as a highly significant predictor of LNM.
• Immune-related pathways, including TGF-β signaling and cancer-associated fibroblast activation, were implicated in PTC metastasis.

Conclusions:

• A non-invasive, gene expression-based AI model can accurately predict LNM in PTC.
• This model offers a cost-effective tool to assist in preoperative surgical planning, potentially reducing unnecessary surgeries.
• The findings provide valuable insights into the molecular mechanisms of LNM in PTC, guiding future research and clinical strategies.