Pretreatment Staging FDG PET-Derived Radiomic Score Predicts Progression-Free Survival in Locally Advanced Rectal Cancer Treated with Total Neoadjuvant Therapy
Background/Objectives: Clinical outcomes in locally advanced rectal cancer (LARC) treated with total neoadjuvant therapy (TNT) remain heterogeneous, and available clinicopathological parameters provide limited prognostic accuracy. This study aimed to develop a radiomic score (RadScore) from baseline staging 18F-FDG PET/CT images using LASSO-penalized Cox regression, and to evaluate its ability to independently predict progression-free survival (PFS) in LARC patients treated with TNT followed by surgery. Methods: Eighty-five patients with LARC who underwent surgical resection after TNT between May 2018 and December 2024 were retrospectively analyzed. Primary tumor volumes were semi-automatically segmented on pretreatment FDG PET/CT images, and 140 radiomic features were extracted in accordance with Image Biomarker Standardisation Initiative (IBSI) guidelines. After correlation-based filtering, 23 non-redundant features were subjected to five-fold cross-validated LASSO-Cox regression for feature selection and RadScore construction. Internal validation used 1000-repetition bootstrap resampling. Results: Disease progression was recorded in 33 of 85 patients (mean PFS: 57.5 months). The final RadScore comprised two features—Intensity Histogram Kurtosis and GLCM Difference Entropy—and was identified as an independent prognostic factor in multivariable Cox analysis (HR = 1.81; 95% CI: 1.23–2.67; p = 0.003). The model achieved a C-index of 0.674, with a bootstrap-corrected C-index of 0.650 (95% CI: 0.568–0.741). Kaplan–Meier analysis demonstrated significantly shorter PFS in the high-risk group (log-rank p = 0.006). Conclusions: A pretreatment FDG PET-derived RadScore independently predicts PFS in LARC patients treated within a TNT framework, offering preliminary evidence for a potential imaging biomarker for preoperative risk stratification and individualized treatment planning—to our knowledge, among the first such evidence in a TNT-treated rectal cancer cohort.
Authors
- Adem Şengül (ORCID: https://orcid.org/0000-0003-1952-0287)
- Kadir Alper Küçüker (ORCID: https://orcid.org/0000-0003-2535-7498)
- Ahmet Alacacıoğlu (ORCID: https://orcid.org/0000-0003-3428-5932)
- Bülent Turgut (ORCID: https://orcid.org/0000-0001-7443-1100)
- Ayşegül Aksu (ORCID: https://orcid.org/0000-0002-6239-0660)
- Onder Kaan Vezir (ORCID: https://orcid.org/0009-0005-8686-0902)
Institutions
- Izmir Kâtip Çelebi University (TR)
Publication Details
- Journal
- Diagnostics
- Published
- 2026-07-22
- DOI
- https://doi.org/10.3390/diagnostics16142294
- Primary Topic
- Radiomics and Machine Learning in Medical Imaging
- Type
- article
- Field-Weighted Citation Impact
- 0.00