Association of quantitative MRI-based radiomic features with prognostic factors and recurrence rate in oropharyngeal squamous cell carcinoma

Radiomics focuses on extracting a large number of quantitative imaging features and testing both their correlation with clinical characteristics prognostic predictive values. We propose radiomic approach using magnetic resonance (MRI) to decode tumor phenotype local recurrence in oropharyngeal squamous cell carcinoma (OPSCC). The contrast-enhanced T1-weighted sequences from baseline MRI examinations OPSCC patients treated between 2008 2016 were retrospectively selected. Radiomic extracted the IBEX software, hierarchical clustering was applied reduce redundancy. association each feature grading stage, HPV status, loco-regional within 2 years, considered as main endpoints, assessed by univariate analysis then corrected for multiple testing. Statistical performed SAS/STAT® software. Thirty-two eligible cases identified. For patient, 1286 extracted, subsequently grouped into 16 clusters. Higher (G3 vs. G1/G2) associated lower values GOH/65Percentile GOH/85Percentile (p = 0.04 0.01, respectively). Positive status higher GOH/10Percentile 0.03) GOH/90Percentile 0.03). Loco-regional years GLCM3/4-7Correlation 0.04) GLCM3/2-1InformationMeasureCorr1 0.04). Results lost statistical significance after correction T stage significantly correlated 9 features, 4 which (GLCM25/180-4InformationMeasureCorr2, Shape/MeanBreadth, GLCM25/90-1InverseDiffMomentNorm GLCM3/6-1InformationMeasureCorr1) retained False Discovery Rate correction. MRI-based radiomics is feasible promising prediction OPSCC. Some seem be oncologic outcome however, larger collaborative studies are warranted order increase power obtain robust validated results.