18F-FDG PET/CT Radiomics Models for Precision Diagnosis and Prognosis in NSCLC

The study consists of three main parts based on a multicenter retrospective cohort:

Staging and Typing: Developing radiomics models to distinguish histological subtypes (Adenocarcinoma vs. Squamous Cell Carcinoma) and predict TNM staging preoperatively.

Gene Mutation Prediction: Analyzing radiomics signatures to predict EGFR mutation status (Mutant vs. Wild-type) non-invasively.

Prognostic Assessment: Evaluating the prognostic value of radiomics features by analyzing their association with Disease-Free Survival (DFS) and Overall Survival (OS).

High-throughput radiomics features will be extracted from standardized PET/CT images and analyzed using machine learning algorithms.

Inclusion criteria

• Age >= 18 years.
• Underwent standard whole-body 18F-FDG PET/CT scan within 30 days before surgery.
• Histopathologically confirmed Non-Small Cell Lung Cancer (NSCLC) with clear histological subtyping and complete postoperative TNM staging.
• Primary tumor SUVmax > 2.5 and maximum diameter > 1.0 cm on CT.
• Complete clinical, pathological, and imaging data available.
• (For Gene Sub-study) Known EGFR gene mutation status.
• (For Prognosis Sub-study) Complete follow-up data available (minimum 12 months or until endpoint event).

Exclusion criteria

• History of other malignancies.
• Received any anti-tumor treatment (chemotherapy, radiotherapy, targeted therapy, immunotherapy) prior to PET/CT.
• Severe image artifacts or indistinct tumor boundaries affecting ROI delineation.
• Missing key clinical or pathological data.
• Baseline PET/CT evaluated recurrent or metastatic tumors instead of primary NSCLC.
• Extremely short life expectancy due to severe comorbidities.