Construction and evaluation of a SPECT-based radiomics nomogram for predicting the therapeutic response to radioactive iodine in patients with differentiated thyroid carcinoma

Differentiated thyroid carcinoma (DTC) frequently requires radioactive iodine therapy, yet accurately predicting short-term therapeutic response remains challenging. Radiomics offers a means to quantify intratumoral heterogeneity from routine imaging. We aimed to develop and internally validate a Single-photon emission computed tomography (SPECT)-based radiomics nomogram for predicting 6-month response after RAI in DTC. We conducted a single-center retrospective study of 420 patients with DTC. Post-therapy 131I SPECT/CT images were processed to extract image biomarker standardisation initiative-aligned radiomic features in MATLAB; features passing reproducibility (ICC ≥ 0.80) and redundancy (|r|<0.85) filters were entered into a nested-cross-validation (CV) least absolute shrinkage and selection operator to derive a radiomics score, which was combined with clinicopathologic variables to build a nomogram. Model performance was assessed for discrimination (receiver operating characteristic area under the curve [AUC]), calibration (concordance index [C-index]; calibration intercept/slope), and clinical utility (decision-curve analysis); internal validation used outer 10-fold CV and 1000 bootstrap with permutation testing. Multicollinearity was checked by variance inflation factor. Interobserver reproducibility was high: 79/107 (73.8%) features achieved ICC ≥ 0.80 (median ICC 0.88), and the median mask-level Dice was 0.87. The nomogram showed AUC = 0.853 (95% confidence interval: 0.786-0.892) and C-index = 0.889 (95% confidence interval: 0.812-0.934) with good calibration. Internal validation yielded CV-AUC = 0.846 (SD 0.037) and bootstrap optimism-corrected C-index = 0.876/AUC = 0.848; permutation AUCs centered near 0.50. Decision-curve analysis indicated clinical net benefit across thresholds 0.15 to 0.95. All validation was internal; no external cohort was available. Radiomics score, age, regional lymph node metastasis, extrathyroidal extension, and tumor diameter independently influence the therapeutic outcome of RAI in DTC patients. In this exploratory, single-center study, the SPECT-based radiomics nomogram demonstrates promising internal performance; however, it must undergo external, multi-center validation and, if necessary, local recalibration before it can be considered for routine clinical use.