Preoperative prediction of IDH genotypes and prognosis in adult-type diffuse gliomas: intratumor heterogeneity habitat analysis using dynamic contrast-enhanced MRI and diffusion-weighted imaging

Abstract

Background: Intratumor heterogeneity (ITH) is a key biological characteristic of gliomas. This study aimed to characterize ITH in adult-type diffuse gliomas and assess the feasibility of using habitat imaging based on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and diffusion-weighted imaging (DWI) to preoperatively predict isocitrate dehydrogenase (IDH) genotypes and prognosis.

Methods: Sixty-three adult-type diffuse gliomas with known IDH genotypes were enrolled. Volume transfer constant (K^trans) and apparent diffusion coefficient (ADC) maps were acquired from DCE-MRI and DWI, respectively. After tumor segmentation, the k-means algorithm clustered K^trans and ADC image voxels to generate spatial habitats and extract quantitative image features. Receiver operating characteristic (ROC) curves and area under the curve (AUC) were used to evaluate IDH predictive performance. Multivariable logistic regression models were constructed and validated using leave-one-out cross-validation, and the contrast-enhanced subgroup was analyzed independently. Kaplan-Meier and Cox proportional hazards regression analyses were used to investigate the relationship between tumor habitats and progression-free survival (PFS) in the two IDH groups.

Results: Three habitats were identified: Habitat 1 (hypo-vasopermeability and hyper-cellularity), Habitat 2 (hypo-vasopermeability and hypo-cellularity), and Habitat 3 (hyper-vasopermeability). Compared to the IDH wild-type group, the IDH mutant group exhibited lower mean K^trans values in Habitats 1 and 2 (both P < 0.001), higher volume (P < 0.05) and volume percentage (pVol, P < 0.01) of Habitat 2, and lower volume and pVol of Habitat 3 (both P < 0.001). The optimal logistic regression model for IDH prediction yielded an AUC of 0.940 (95% confidence interval [CI]: 0.880-1.000), which improved to 0.948 (95% CI: 0.890-1.000) after cross-validation. Habitat 2 contributed the most to the model, consistent with the findings in the contrast-enhanced subgroup. In IDH wild-type group, pVol of Habitat 2 was identified as a significant risk factor for PFS (high- vs. low-pVol subgroup, hazard ratio = 2.204, 95% CI: 1.061-4.580, P = 0.034), with a value below 0.26 indicating a 5-month median survival benefit.

Conclusions: Habitat imaging employing DCE-MRI and DWI may facilitate the characterization of ITH in adult-type diffuse gliomas and serve as a valuable adjunct in the preoperative prediction of IDH genotypes and prognosis.

Clinical trial number: Not applicable.

Keywords: Adult-type diffuse glioma; Diffusion-weighted imaging; Dynamic contrast-enhanced perfusion; Intratumor heterogeneity; Isocitrate dehydrogenase; Progression-free survival.

Fig. 1

Flow diagram of the patient inclusion process. DCE, dynamic contrast-enhanced; IDH, isocitrate dehydrogenase; T1WI, T1-weighted imaging

Fig. 2

Demonstration of habitat segmentation slices in two patients with IDH mutant and wild-type gliomas. Spatial habitats are labeled with different colors. Contrast-enhanced T1-weighted imaging shows (A) a non-enhanced IDH mutant glioma, (B) an obviously enhanced IDH mutant glioma, (C) an obviously enhanced IDH wild-type glioma, and (D) a mildly enhanced IDH wild-type glioma

Fig. 3

Receiver operating characteristic (ROC) curves of habitat features and tumor volume of interest (VOI)-based features with an area under the ROC curve (AUC) value exceeding 0.7 in discriminating IDH genotypes. Features include K^trans_Mean (AUC: 0.827, 95% CI: 0.726–0.929, P < 0.001) in Habitat 1, K^trans_Mean (AUC:0.825, 95% CI:0.721–0.929, P < 0.001) and pVol (AUC: 0.718, 95% CI: 0.589–0.848, P = 0.005) of Habitat 2, Volume (AUC: 0.756, 95% CI: 0.634–0.878, P = 0.001) and pVol (AUC: 0.805, 95% CI: 0.696–0.914, P < 0.001) of Habitat 3, and K^trans_Mean (AUC: 0.810, 95% CI: 0.703–0.918, P < 0.001) in tumor VOI. Ktrans = K^trans

Fig. 4

Receiver operating characteristic (ROC) curves of the multivariable logistic regression models for IDH prediction in entire cohort (A) and contrast-enhanced subgroup (B). Model CM represents the multivariable logistic regression model based solely on clinical and morphological features. Models TV, H1, H2, H3, and AHs correspond to multivariable logistic regression models based on quantitative features from the tumor volume of interest (VOI), Habitats 1, 2, and 3, and all habitats, each incorporating clinical and morphological features

Fig. 5

Kaplan-Meier curves for paired patient subgroups based on all quantitative metrics from Habitat 2 in (A) the IDH wild-type group and (B) the IDH mutant group.