Purpose or Case Report: It has been known that the pituitary gland volume (PV) in idiopathic central precocious puberty (IPP) is significantly higher than in healthy children. However, most PV measurements rely on manual quantitative methods, which are time-consuming and labor-intensive. This study aims to automatically measure the PV of patients with IPP using artificial intelligence to accurately quantify the correlation between IPP and PV, and to improve the efficiency of diagnosing IPP.
Methods & Materials: From July 2016 to February 2024, 226 patients diagnosed with IPP and underwent brain MR were included (117 males and 109 females, median age, 8 years). A control group of 52 patients who underwent brain MR imaging without symptoms of precocious puberty was also included (37 males and 15 females, median age, 8 years). Measurement variability was examined between manual and automatic measurements (n=57). The pituitary gland volume was measured using 1-3 mm thickness T1 sagittal images from non-enhanced brain MR imaging, analyzed with the MA-net artificial intelligence learning method. Physical characteristics (height, weight, and age) were correlated with PV, and the difference in PV between the IPP group and the control group was evaluated.
Results: The intraclass correlation coefficient was 0.993 for agreement between manual and automatic measurement. Confounding bias was reduced by Propensity Score Matching. PV was positively correlated with age and body weight in the IPP group (17.4%, P =0.009, and 14.0%, P =.037). The median values of PV were 432 cubic millimeters in the IPP group and 380 cubic millimeters in the control group, showing a significant difference of 52 cubic millimeters (P <0.05).
Conclusions: The PV in the IPP group was significantly higher than in the control group. Automatically measuring PV along with assessing hormone levels could enable a faster and more straightforward diagnosis of the IPP.