Pituitary MRI features in identifying idiopathic short stature from growth hormone deficiency in children with short stature.

Objective: This study examined the diagnostic performance of the prediction models using baseline characteristics, biochemical indicators, and adenohypophysis MRI features for differentiating idiopathic short stature (ISS) from growth hormone deficiency (GHD).

Methods: A total of 96 patients with short stature underwent hypophysial CUBE T1-weighted imaging on 3.0-T scanner and GH stimulation testing between February 2021 and February 2024 and were classified into ISS and GHD groups according to GH stimulation testing results. Two-independent-sample T-test was tested for the differences between groups. The partial correlation analysis was conducted after controlling for demographic data. The prediction models were established using stepwise binary logistic regression method. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the diagnostic efficacy of the models.

Results: ISS group (29 boys and 29 girls) and GHD group (21 boys and 17 girls) showed significant differences in pituitary height (aPH) and pituitary volume (aPV), GH_max, LH_max and cortisol but no difference in demographic data (gender, age, height, weight and BMI). After controlling for those demographic data, aPH, aPV, FSH_max and LH_max showed positive correlation with GH_max. Using binary logistic regression, three prediction models were built; Model 1 had the highest AUC value (0.862) followed by Model 3 with AUC value of 0.858 and Model 2 with the lowest AUC value (0.725).

Conclusion: A useful prediction model using adenohypophysis MRI features, age, and BMI had great potential in differentiation between GHD and ISS.

Key points: Question Distinguishing growth hormone deficiency (GHD) from idiopathic short stature (ISS) among prepubescent children is important but difficult, time-consuming, and costly with growth hormone stimulation testing. Findings The proposed model using clinical and radiomics features offered sufficient diagnostic performance on differentiating GHD and ISS. Clinical relevance We propose a low-cost, time-saving, and non-invasive model based on adenohypophysis MRI features and baseline characteristics to differentiate GHD and ISS.