Quantification of T2-FLAIR Mismatch in Nonenhancing Diffuse Gliomas Using Digital Subtraction

Representative false-negative and false-positive cases with T2-FLAIR subtraction maps and tumor segmentations. A, Patient A is a 38-year-old man from the TCIA cohort diagnosed with grade 2 IDHm-A with no T2-FLAIR mismatch on the subtraction map (false-negative result). B, Patient B is a 36-year-old woman from the institutional cohort diagnosed with grade 2 IDHm-A with no T2-FLAIR mismatch on the subtraction map excluding the cystic portion (false-negative result). C, Patient C is a 45-year-old man from the TCIA cohort diagnosed with grade 2 IDHm-O demonstrating 35.0% T2FM volume, which was assessed as a false-positive for the T2FM sign by 1 reader using subtraction maps (false-positive result). D, Patient D is a 35-year-old woman diagnosed with IDHwt glioblastoma demonstrating 12.5% T2FM volume, which was assessed as a false-positive for the T2-FLAIR mismatch sign by both readers using subtraction maps. Corresponding tumor-segmentation VOIs of T2FM (pink) and T2FNM (green) subregions are shown (A–D) (false-positive result).

Tumor percentage of T2FM volume and diagnostic performance for IDHm-A classification. Results are shown for both the TCIA cohort (A and B) and the institutional cohort (C and D). IDHm-A had a higher percentage T2FM volume (% T2FM volume) compared with the other molecular signatures (A and C). ROC curves show that a specificity of 100% for the IDHm-A diagnosis could be achieved with a threshold of ≥42% T2FM volume in both cohorts (B and D), though in the validation cohort, a threshold of ≥25% T2FM volume sufficed (D). The summary results when pooling the 2 cohorts are also shown (E and F). INST indicates institutional; Spec, specificity; Sens, sensitivity; *, P < .05; ***, P < .001; ****, P < .0001.