Machine Learning-Driven Radiomics for an Early-Stage Predictive Model of Nodal Upstaging in Thoracic Oncology

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Precise lymph node staging is crucial for non-small cell lung cancer (NSCLC) management.
• Occult nodal metastases detected post-surgery can alter prognosis and treatment.
• Current 2-[18F]FDG PET/CT has limitations in identifying all nodal involvement.

Purpose of the Study:

• To evaluate radiomic features from preoperative 2-[18F]FDG PET/CT for predicting nodal involvement in early-stage NSCLC.
• To identify imaging biomarkers for non-invasive nodal staging.
• To assess the potential of radiomics in refining surgical planning and patient selection.

Main Methods:

• Retrospective analysis of 124 patients with cT1N0 NSCLC.
• Preoperative 2-[18F]FDG PET/CT scans were analyzed for radiomic features.
• Comparison between patients with and without unexpected nodal metastasis post-resection.

Main Results:

• Several radiomic parameters significantly correlated with lymph node upstaging.
• Metabolic tumor volume (MTV), total lesion glycolysis (TLG), run-length non-uniformity (RLNU), and gray-level non-uniformity (GLNU) were key predictors.
• These features demonstrated strong association with pathological nodal involvement.

Conclusions:

• Radiomic analysis of 2-[18F]FDG PET/CT shows promise as a non-invasive tool for predicting nodal involvement in early-stage NSCLC.
• Integrating radiomics may enhance preoperative staging accuracy.
• This approach could lead to improved surgical decision-making and reduced unexpected upstaging.