Longitudinal structural and perfusion MRI enhanced by machine learning outperforms standalone modalities and radiological expertise in high-grade glioma surveillance.

PURPOSE: Surveillance of patients with high-grade glioma (HGG) and identification of disease progression remain a major challenge in neurooncology. This study aimed to develop a support vector machine (SVM) classifier, employing combined longitudinal structural and perfusion MRI studies, to classify between stable disease, pseudoprogression and progressive disease (3-class problem). METHODS: Study participants were separated into two groups: group I (total cohort: 64 patients) with a single DSC time point and group II (19 patients) with longitudinal DSC time points (2-3). We retrospectively analysed 269 structural MRI and 92 dynamic susceptibility contrast perfusion (DSC) MRI scans. The SVM classifier was trained using all available MRI studies for each group. Classification accuracy was assessed for different feature dataset and time point combinations and compared to radiologists' classifications. RESULTS: SVM classification based on combined perfusion and structural features outperformed radiologists' classification across all groups. For the identification of progressive disease, use of combined features and longitudinal DSC time points improved classification performance (lowest error rate 1.6%). Optimal performance was observed in group II (multiple time points) with SVM sensitivity/specificity/accuracy of 100/91.67/94.7% (first time point analysis) and 85.71/100/94.7% (longitudinal analysis), compared to 60/78/68% and 70/90/84.2% for the respective radiologist classifications. In group I (single time point), the SVM classifier also outperformed radiologists' classifications with sensitivity/specificity/accuracy of 86.49/75.00/81.53% (SVM) compared to 75.7/68.9/73.84% (radiologists). CONCLUSION: Our results indicate that utilisation of a machine learning (SVM) classifier based on analysis of longitudinal perfusion time points and combined structural and perfusion features significantly enhances classification outcome (p value= 0.0001).

World Health Organization Grade II/III Glioma Molecular Status: Prediction by MRI Morphologic Features and Apparent Diffusion Coefficient.

Background A readily implemented MRI biomarker for glioma genotyping is currently lacking. Purpose To evaluate clinically available MRI parameters for predicting isocitrate dehydrogenase (IDH) status in patients with glioma. Materials and Methods In this retrospective study of patients studied from July 2008 to February 2019, untreated World Health Organization (WHO) grade II/III gliomas were analyzed by three neuroradiologists blinded to tissue results. Apparent diffusion coefficient (ADC) minimum (ADCmin) and mean (ADCmean) regions of interest were defined in tumor and normal appearing white matter (ADCNAWM). A visual rating of anatomic features (T1 weighted, T1 weighted with contrast enhancement, T2 weighted, and fluid-attenuated inversion recovery) was performed. Interobserver comparison (intraclass correlation coefficient and Cohen κ) was followed by nonparametric (Kruskal-Wallis analysis of variance) testing of associations between ADC metrics and glioma genotypes, including Bonferroni correction for multiple testing. Descriptors with sufficient concordance (intraclass correlation coefficient, >0.8; κ > 0.6) underwent univariable analysis. Predictive variables (P < .05) were entered into a multivariable logistic regression and tested in an additional test sample of patients with glioma. Results The study included 290 patients (median age, 40 years; interquartile range, 33-52 years; 169 male patients) with 82 IDH wild-type, 107 IDH mutant/1p19q intact, and 101 IDH mutant/1p19q codeleted gliomas. Two predictive models incorporating ADCmean-to-ADCNAWM ratio, age, and morphologic characteristics, with model A mandating calcification result and model B recording cyst formation, classified tumor type with areas under the receiver operating characteristic curve of 0.94 (95% confidence interval [CI]: 0.91, 0.97) and 0.96 (95% CI: 0.93, 0.98), respectively. In the test sample of 49 gliomas (nine IDH wild type, 21 IDH mutant/1p19q intact, and 19 IDH mutant/1p19q codeleted), the classification accuracy was 40 of 49 gliomas (82%; 95% CI: 71%, 92%) for model A and 42 of 49 gliomas (86%; 95% CI: 76%, 96%) for model B. Conclusion Two algorithms that incorporated apparent diffusion coefficient values, age, and tumor morphologic characteristics predicted isocitrate dehydrogenase status in World Health Organization grade II/III gliomas on the basis of standard clinical MRI sequences alone. © RSNA, 2020 Online supplemental material is available for this article.

Evaluation of the stability of the stereotactic Leksell Frame G in Gamma Knife radiosurgery.

The purpose of this study was to evaluate the stability of the Leksell Frame G in Gamma Knife radiosurgery (GKR). Forty patients undergoing GKR underwent pretreatment stereotactic MRI for GKR planning and stereotactic CT immediately after GKR. The stereotactic coordinates of four anatomical landmarks (cochlear apertures and the summits of the anterior post of the superior semicircular canals, bilaterally) were measured by two evaluators on two separate occasions in the pre-treatment MRI and post-treatment CT scans and the absolute distance between the observations is reported. The measurement method was validated with an indepen-dent group of patients who underwent both stereotactic MRI and CT imaging before treatment (negative controls; n: 5). Patients undergoing GKR for arteriovenous malformations (AVM) also underwent digital subtraction angiography (DSA), which could result in extra stresses on the frame. The distance between landmark local-ization in the scans for the negative control group (0.63 mm; 95% CI: 0.57-0.70; SD: 0.29) represents the overall consistency of the evaluation method and provides an estimate of the minimum displacement that could be detected by the study. Two patients in the study group had the fiducial indicator box accidentally misplaced at post-treatment CT scanning. This simulated the scenario of a frame displacement, and these cases were used as positive controls to demonstrate that the evaluation method is capable of detecting a discrepancy between the MRI and CT scans, if there was one. The mean distance between the location of the landmarks in the pretreatment MRI and post-treatment CT scans for the study group was 0.71 mm (95% CI: 0.68-0.74; SD:0.32), which was not statistically different from the over-all uncertainty of the evaluation method observed in the negative control group (p = 0.06). The subgroup of patients with AVM (n: 9), who also underwent DSA, showed a statistically significant difference between the location of the landmarks compared to subjects with no additional imaging: 0.78 mm (95% CI: 0.72-0.84) vs. 0.69 mm (95% CI: 0.66-0.72), p = 0.016. This is however a minimal differ-ence (0.1 mm) and the mean difference in landmark location for each AVM patient remained submillimeter. This study demonstrates submillimeter stability of the Leksell Frame G in GKR throughout the treatment procedure.