Evaluation of Pulmonary Nodules by Radiologists vs. Radiomics in Stand-Alone and Complementary CT and MRI

Area of Science:

• Radiology and Imaging Science
• Oncology
• Medical Informatics

Background:

• Magnetic Resonance Imaging (MRI) is increasingly used for radiation-free screening of malignant pulmonary nodules.
• Accurate classification of pulmonary nodules is crucial for timely diagnosis and treatment of lung cancer.
• Radiomic feature analysis offers a quantitative approach to extract imaging biomarkers for disease characterization.

Purpose of the Study:

• To compare the diagnostic performance of human readers versus radiomic feature analysis in classifying pulmonary nodules.
• To evaluate the effectiveness of stand-alone and complementary CT and MRI imaging modalities.
• To determine if combining quantitative imaging data from multiple sequences enhances nodule classification accuracy.

Main Methods:

• A single-center study included patients with CT-detected pulmonary nodules who underwent additional lung MRI.
• Nodule classification (benign/malignant) was confirmed by surgical resection.
• Radiomic features were extracted from 2D segmentations of nodules on CT, T2-weighted (T2w), and diffusion-weighted imaging (DWI) sequences. Feature selection was performed using iterative backward elimination. Performance was assessed by accuracy using Clopper-Pearson confidence intervals.

Main Results:

• Analysis of 66 pulmonary nodules (40 malignant) in 50 patients showed that radiomic analysis using combined CT, T2w, and DWI datasets achieved the highest accuracy (0.83; 95% CI: 0.72, 0.91).
• Combined CT/T2w/DWI radiomics outperformed individual modalities (CT alone: 0.68, T2w alone: 0.65, DWI alone: 0.61).
• The combined radiomics approach also demonstrated superior performance compared to human readers across all evaluated imaging combinations (CT alone: 0.59, T2w alone: 0.68, DWI alone: 0.73, combined T2w/DWI: 0.70, combined CT/T2w/DWI: 0.64).

Conclusions:

• This study is the first to demonstrate that combining quantitative image information from CT, T2w, and DWI datasets via radiomics analysis significantly improves pulmonary nodule assessment.
• The integrated radiomic approach surpasses the performance of single imaging modalities.
• Quantitative radiomic analysis using multi-modal imaging data, particularly CT/T2w/DWI, exceeds the diagnostic performance of human readers for pulmonary nodule classification.