Assessing the application of radiomics on post radio-iodine therapy whole body scan – A preliminary study

Purpose: Radiomics has been principally applied for characterization of tumor heterogeneity which by virtue of its predictive power in the prognostication, claims further implication in the management. As a practitioner of radioactive iodine therapy, we have observed the characteristic of radiomic features extracted from post radioiodine therapy whole body scan. Patients and methods: Anterior planar images from 39 patients who underwent radioiodine therapy after thyroidectomy due to differentiated papillary thyroid carcinoma were analyzed in LIFEX software to generate radiomic feature (RF) data from two regions of interest (ROI): neck and liver per patient. Correlation of the RF was checked against two demographic variables, patient’s body weight and patient’s age and three clinical variables, administered radioiodine dose during therapy, pre therapy TSH level, pre therapy thyroglobulin level. After discovering features with significant correlation with demographic and clinical variables, the RF that correlated in each of the two ROI were shortlisted. Results: A total of 141 RF per ROI per patient were extracted using LIFEX while 32 Radiomics features from neck ROI and 15 radiomic features from liver ROI were discovered to have significant correlation with two demographic variables and three clinical variables (R ranged from -0.5 to 0.49, p < 0.05). Four radiomic features were common for both neck and liver ROI that have significant correlations. Among those four, one was correlated with body weight and three were with therapy dose. Pre therapy TSH level was found to have correlation with 10 features from neck ROI and no feature from liver ROI. Pre therapy Tg level was found to have correlation with two features from neck ROI and one feature from liver ROI. Conclusions: With further exploration, the RF from post radioiodine therapy whole body scan may evolve as a useful aid to the management of differentiated thyroid cancer by providing prediction about clinical outcome as well as by playing as surrogates of histopathological and immunohistochemical markers. Bangladesh J. Nuclear Med. 27(2): 241-244, 2024