AI-based synthetic CT attenuation correction enables reliable quantitative SPECT in unilateral condylar hyperplasia.

Abstract

Purpose: Unilateral condylar hyperplasia (UCH) is a rare mandibular growth disorder in which accurate assessment of condylar metabolic activity is essential for surgical decision-making. Quantitative ^99mTc-methylene diphosphonate (MDP) SPECT/CT is commonly used for this purpose; however, CT-based attenuation correction (CTAC) increases radiation exposure and may be affected by registration errors. Artificial intelligence (AI)-generated synthetic CT (SyCT) has been proposed as a CT-independent alternative. This study aimed to evaluate the agreement between SyCT-based attenuation correction (SyCTAC) and conventional CTAC in quantitative SPECT imaging of UCH.

Methods: This retrospective study included 14 patients with UCH who underwent ^99mTc-MDP SPECT/CT. SPECT images were reconstructed with identical parameters using either CTAC or AI-generated SyCTAC. Spherical volumes of interest were placed over both mandibular condyles and the clivus, and maximum and mean standardized uptake values were measured. Relative uptake fractions normalized to the summed condylar activity were calculated for the affected side, while condylar uptake normalized to the clivus was evaluated for both sides. Agreement between CTAC- and SyCTAC-derived indices was assessed using Bland-Altman analysis and linear mixed-effects models.

Results: Visual evaluation revealed no relevant differences in image quality between the two reconstructions. Relative uptake fractions normalized to the summed condylar activity showed good agreement between methods. Clivus-normalized ratios demonstrated a small positive bias.

Conclusion: AI-generated SyCT provides attenuation correction comparable to CTAC for clinically relevant relative condylar uptake assessment in UCH, supporting its use as a low-radiation alternative for quantitative mandibular SPECT imaging.