PT  - JOURNAL ARTICLE
AU  - Payabvash, Seyedmehdi
AU  - Sharaf, Kariem
AU  - Zeevi, Tal
AU  - Gross, Moritz
AU  - Mahajan, Amit
AU  - Kann, Benjamin H.
AU  - Judson, Benjamin L.
AU  - Schreier, Andrea
AU  - Krenn, Jasmin
AU  - Prasad, Manju L.
AU  - Burtness, Barbara
AU  - Aboian, Mariam
AU  - Canis, Martin
AU  - Baumeister, Philipp
AU  - Reichel, Christoph A.
AU  - Haider, Stefan P.
TI  - FDG-PET intensity normalization improves radiomics- based survival prediction in oropharyngeal cancer patients: a comparison of the SUV with alternative normalization techniques
AID  - 10.3174/ajnr.A8836
DP  - 2025 May 16
TA  - American Journal of Neuroradiology
PG  - ajnr.A8836
4099  - http://www.ajnr.org/content/early/2025/05/16/ajnr.A8836.short
4100  - http://www.ajnr.org/content/early/2025/05/16/ajnr.A8836.full
AB  - BACKGROUND AND PURPOSE: Despite the widespread research application of radiomics, there is a knowledge gap regarding the optimal voxel intensity normalization strategy for FDG-PET radiomics. We investigated the impact of three normalization strategies on the prognostic utility of individual radiomic features and machine learning models in oropharyngeal squamous cell carcinoma (OPSCC) patients.MATERIALS AND METHODS: We included n=330 (overall survival, OS, study group), n=335 (progression-free survival, PFS, study group) and n=309 (locoregional progression, LRP, study group) OPSCC patients. Three FDG-PET intensity normalization strategies were applied: the conventional body weight-corrected Standardized Uptake Value (SUV), and standardized uptake ratios to the lentiform nucleus and to the cerebellum. The raw PET voxel intensities were also analyzed. To quantify and compare features’ association with oncologic outcome, we fitted univariate Cox regression models, calculated Harrell’s C-index, and fitted random survival forest (RSF) machine learning algorithms.RESULTS: All normalization strategies tended to improve the prognostic value of radiomic features. Features from lentiform nucleus- normalized PET demonstrated the highest prognostic improvement, with n=750/1037, n=809/1037 and n=652/1037 primary tumor features attaining a significant association with OS, PFS, and LRP, respectively, compared to n=0, n=211, and n=1 SUV-based PET features, respectively. The median C-index of lentiform nucleus-normalized PET features was 0.64, 0.61 and 0.62 for OS, PFS, and LRP, respectively, while SUV-based PET features reached 0.59, 0.58 and 0.60, respectively. The best performing lentiform nucleus- normalization RSF model significantly outperformed the raw PET RSF model in predicting OS (C-index=0.66 vs. C-index=0.57; p=0.019), with model comparisons for PFS and LRP approaching statistical significance (p=0.053 and p=0.084, respectively). In contrast, the best performing SUV-based RSF models were not significantly different from raw PET models.CONCLUSIONS: Normalizing PET intensities, especially to the lentiform nucleus, improves the prognostic performance of individual radiomic features and machine learning models in predicting oncologic outcome.ABBREVIATIONS: FDG-PET = [18F]fluorodeoxyglucose positron emission tomography; SUV = standardized uptake value; OPSCC = oropharyngeal squamous cell carcinoma; HPV = human papillomavirus; VOI = volume of interest; OS = overall survival; PFS = progression-free survival; LRP = locoregional progression; C-index = Harrell’s concordance index; RSF = random survival forest; CV = cross-validation; AUC = area under the curve; SD = standard deviation.