On the Automation of Radiomics-Based Identification and Characterization of NSCLC

Area of Science:

• Medical Imaging
• Radiology
• Oncology

Background:

• Accurate Non-Small Cell Lung Cancer (NSCLC) detection and characterization remain challenging in medical imaging.
• Radiomics offers potential for quantitative biomarkers from medical images, but standardization and definitive results are lacking.
• Biomedical imaging is crucial for lung cancer assessment and patient management.

Purpose of the Study:

• To design and develop LuCIFEx, a fully-automated pipeline for non-invasive, in-vivo NSCLC characterization.
• To accelerate NSCLC analysis and facilitate early tumor diagnosis.
• To assess the utility of radiomic features from PET/CT for NSCLC classification.

Main Methods:

• LuCIFEx utilizes routinely acquired [18F]FDG-PET/CT images for automated cancer lesion segmentation.
• A multi-stage segmentation process achieves a mean accuracy of 94.2±5.0% in lesion identification.
• Machine learning algorithms are employed for cancer characterization using computed radiomic features.

Main Results:

• The LuCIFEx pipeline demonstrates high accuracy in segmenting NSCLC lesions.
• [18F]FDG-PET/CT radiomic features show potential for automatically recognizing lung metastases.
• The study validates the capability of radiomic features for identifying NSCLC histological subtypes.

Conclusions:

• The LuCIFEx pipeline offers a promising automated approach for NSCLC characterization using PET/CT imaging.
• Radiomics, despite current limitations, holds significant potential for early NSCLC diagnosis and management.
• Further standardization and validation are necessary to fully realize the clinical utility of radiomics in lung cancer.