Predictive IDH Genotyping Based on the Evaluation of Spatial Metabolic Heterogeneity by Compartmental Uptake Characteristics in Preoperative Glioma Using¹⁸F-FET PET

Abstract

Molecular markers are of increasing importance for classifying, treating, and determining the prognosis for central nervous system tumors. Isocitrate dehydrogenase (IDH) is a critical regulator of glucose and amino acid metabolism. Our objective was to investigate metabolic reprogramming of glioma using compartmental uptake (CU) characteristics in <i>O</i>-(2-¹⁸F-fluoroethyl)-l-tyrosine (FET) PET and to evaluate its diagnostic potential for IDH genotyping. <b>Methods:</b> Between 2017 and 2022, patients with confirmed glioma were preoperatively investigated using static ¹⁸F-FET PET. Metabolic tumor volume (MTV), MTV for 60%-100% uptake (MTV₆₀), and T2-weighted and contrast-enhancing lesion volumes were automatically segmented using U-Net neural architecture and isocontouring. Volume intersections were determined using the Dice coefficient. Uptake characteristics were determined for metabolically defined compartments (central [80%-100%] and peripheral [60%-75%] areas of ¹⁸F-FET uptake). CU ratio was defined as the fraction between the peripheral and central compartments. Mean target-to-background ratio was calculated. Comparisons were performed using parametric and nonparametric tests. Receiver-operating-characteristic curves, regression, and correlation were used for statistical analysis. <b>Results:</b> In total, 52 participants (male, 27, female, 25; mean age ± SD, 51 ± 16 y) were evaluated. MTV₆₀ was greater and distinct from contrast-enhancing lesion volume (<i>P</i> = 0.046). IDH-mutated tumors presented a greater volumetric CU ratio and SUV CU ratio than IDH wild-type tumors (<i>P</i> < 0.05). Volumetric CU ratio determined IDH genotype with excellent diagnostic performance (area under the curve [AUC], 0.88; <i>P</i> < 0.001) at more than 5.49 (sensitivity, 86%, specificity, 90%), because IDH-mutated tumors presented a greater peripheral metabolic compartment than IDH wild-type tumors (<i>P</i> = 0.045). MTV₆₀ and MTV were not suitable for IDH classification (<i>P</i> > 0.05). SUV CU ratio (AUC, 0.72; <i>P</i> = 0.005) and target-to-background ratio (AUC, 0.68; <i>P</i> = 0.016) achieved modest diagnostic performance-inferior to the volumetric CU ratio. Furthermore, the classification of loss of heterozygosity of chromosomes 1p and 19q (AUC, 0.75; <i>P</i> = 0.019), MGMT promoter methylation (AUC, 0.70; <i>P</i> = 0.011), and ATRX loss (AUC, 0.73; <i>P</i> = 0.004) by amino acid PET was evaluated. <b>Conclusion:</b> We proposed parametric ¹⁸F-FET PET as a noninvasive metabolic biomarker for the evaluation of CU characteristics, which differentiated IDH genotype with excellent diagnostic performance, establishing a critical association between spatial metabolic heterogeneity, mitochondrial tricarboxylic acid cycle, and genomic features with critical implications for clinical management and the diagnostic workup of patients with central nervous system cancer.