European Society of Neuroradiology (ESNR)

Prediction of high proliferative index in pituitary macroadenomas using MRI-based radiomics and machine learning Abstract Purpose Pituitary adenomas are among the most frequent intracranial tumors. They may exhibit clinically aggressive behavior, with recurrent disease and resistance to multimodal therapy. The ki-67 labeling index represents a proliferative marker which correlates with pituitary adenoma aggressiveness. Aim of our study was to assess the accuracy of machine learning analysis of texture-derived parameters from pituitary adenomas preoperative MRI for the prediction of ki-67 proliferation index class. Methods A total of 89 patients who underwent an endoscopic endonasal procedure for pituitary adenoma removal with available ki-67 labeling index were included. From T2w MR images, 1128 quantitative imaging features were extracted. To select the most informative features, different supervised feature selection methods were employed. Subsequently, a k-nearest neighbors (k-NN) classifier was employed to predict macroadenoma high or low proliferation index. Algorithm validation was performed with a train-test approach. Results Of the 12 subsets derived from feature selection, the best performing one was constituted by the 4 highest correlating parameters at Pearson’s test. These all showed very good (ICC ≥ 0.85) inter-observer reproducibility. The overall accuracy of the k-NN in the test group was of 91.67% (33/36) of correctly classified patients. Conclusions Machine learning analysis of texture-derived parameters from preoperative T2 MRI has proven to be effective for the prediction of pituitary macroadenomas ki-67 proliferation index class. This might aid the surgical strategy making a more accurate preoperative lesion classification and allow for a more focused and cost-effective follow-up and long-term management.

Standardized image evaluation in patients with idiopathic normal pressure hydrocephalus: consistency and reproducibility Abstract Purpose Assess the agreement for two investigators between computed tomography (CT) and magnetic resonance imaging (MRI) for seven imaging features included in the iNPH Radscale, a radiological screening tool. Methods The study included 35 patients with idiopathic normal pressure hydrocephalus (iNPH) who were treated surgically from 2011 to 2015 at Uppsala University Hospital with preoperative CT and MRI performed with maximum 3 months between scans. Seven features were assessed: Evans’ index, temporal horn size, callosal angle, periventricular white matter changes, narrow high convexity sulci, focally enlarged sulci, and enlarged Sylvian fissures. All scans were assessed by two investigators who were blinded to each other’s results and to clinical data. Results The agreement between CT and MRI was almost perfect for Evans’ index, temporal horns, narrow sulci, and Sylvian fissures (kappa and intraclass correlation, 0.84–0.91, p ≤ 0.001). There was substantial to almost perfect agreement for callosal angle and focally enlarged sulci. The concordance between modalities was fair for changes in periventricular white matter. Conclusion CT and MRI are equally good for assessing radiological signs associated with iNPH except for periventricular white matter changes, as MRI has superior soft tissue contrast. The other imaging features can be evaluated consistently, and assessments are reproducible independent of modality. Therefore, the iNPH Radscale is applicable to both CT and MRI and may become an important tool for standardized evaluation in the workup in patients with suspected iNPH.

Endovascular treatment in patients with large vessel occlusion: reduced mortality despite minimal penumbra Abstract Purpose In patients with acute ischemic stroke (AIS) caused by large vessel occlusion (LVO), endovascular treatment (EVT) is highly effective for emergency revascularization. However, data on functional outcome are lacking for patients, which show no or minimal mismatch between ischemic core and penumbra. Methods Forty-five patients with AIS due to LVO of the anterior circulation were retrospectively analyzed within 6 h since onset when administered to our department. In all patients, there was no relevant penumbra according to CT perfusion (CTP). Functional outcome, defined by the modified Rankin Scale (mRS) at 30 and 90 days, was analyzed according to LVO treatment (EVT versus non-EVT). Confounding was addressed by multivariable regression analyses. Results mRS values at 30 days (p = 0.002) and 90 days (p = 0.005) after AIS occurrence were significantly lower in patients who had received EVT. There was no significant difference regarding good functional outcome, as measured by mRS of 0–2 at 30 (p = 0.432) and 90 days, respectively (p = 0.186). Mortality was significantly reduced in patients undergoing EVT at 30-day (p < 0.001) and at 90-day follow-up (p = 0.003), respectively. Multivariable regression analyses revealed that EVT was associated with reduced mortality at 30 (OR 0.091; CI (0.013–0.612); p = 0.014) and 90 days (OR 0.134; CI (0.021–0.857); p = 0.034) after AIS. Conclusions Despite a small and highly selected patient collective, our study indicates that AIS patients with minimal penumbra in CTP might benefit from EVT in terms of reduced mortality at 30 and 90 days after AIS. However, in this group of patients, we could not prove favorable functional outcome at 30 and 90 days, despite receiving EVT.

White matter alterations in adult with autism spectrum disorder evaluated using diffusion kurtosis imaging Abstract Purpose Autism spectrum disorder (ASD) is related to impairment in various white matter (WM) pathways. Utility of the recently developed two-compartment model of diffusion kurtosis imaging (DKI) to analyse axial diffusivity of WM is restricted by several limitations. The present study aims to validate the utility of model-free DKI in the evaluation of WM alterations in ASD and analyse the potential relationship between DKI-evident WM alterations and personality scales. Methods Overall, 15 participants with ASD and 15 neurotypical (NT) controls were scanned on a 3 T magnetic resonance (MR) scanner, and scores for autism quotient (AQ), systemising quotient (SQ) and empathising quotient (EQ) were obtained for both groups. Multishell diffusion-weighted MR data were acquired using two b-values (1000 and 2000 s/mm2). Differences in mean kurtosis (MK), radial kurtosis (RK) and axial kurtosis (AK) between the groups were evaluated using tract-based spatial statistics (TBSS). Finally, the relationships between the kurtosis indices and personality quotients were examined. Results The ASD group demonstrated significantly lower AK in the body and splenium of corpus callosum than the NT group; however, no other significant differences were identified. Negative correlations were found between AK and AQ or SQ, predominantly in WM areas related to social–emotional processing such as uncinate fasciculus, inferior fronto-occipital fasciculus, and inferior and superior longitudinal fasciculi. Conclusions Model-free DKI and its indices may represent a novel, objective method for detecting the disease severity and WM alterations in patients with ASD.

Diagnostic accuracy of flat-panel computed tomography in assessing cerebral perfusion in comparison with perfusion computed tomography and perfusion magnetic resonance: a systematic review Abstract Purpose Flat-panel computed tomography (FP-CT) is increasingly available in angiographic rooms and hybrid OR’s. Considering its easy access, cerebral imaging using FP-CT is an appealing modality for intra-procedural applications. The purpose of this systematic review is to assess the diagnostic accuracy of FP-CT compared with perfusion computed tomography (CTP) and perfusion magnetic resonance (MRP) in cerebral perfusion imaging. Methods We performed a systematic literature search in the Cochrane Library, MEDLINE, Embase, and Web of Science up to June 2019 for studies directly comparing FP-CT with either CTP or MRP in vivo. Methodological quality was assessed using the QUADAS-2 tool. Data on diagnostic accuracy was extracted and pooled if possible. Results We found 11 studies comparing FP-CT with CTP and 5 studies comparing FP-CT with MRP. Most articles were pilot or feasibility studies, focusing on scanning and contrast protocols. All patients studied showed signs of cerebrovascular disease. Half of the studies were animal trials. Quality assessment showed unclear to high risks of bias and low concerns regarding applicability. Five studies reported on diagnostic accuracy; FP-CT shows good sensitivity (range 0.84–1.00) and moderate specificity (range 0.63–0.88) in detecting cerebral blood volume (CBV) lesions. Conclusions Even though FP-CT provides similar CBV values and reconstructed blood volume maps as CTP in cerebrovascular disease, additional studies are required in order to reliably compare its diagnostic accuracy with cerebral perfusion imaging.

Impact of hypertension on cerebral microvascular structure in CPAP-treated obstructive sleep apnoea patients: a diffusion magnetic resonance imaging study Abstract Purpose Obstructive sleep apnoea (OSA) is a highly prevalent sleep-related breathing disorder associated with hypertension, impaired peripheral vascular function and an increased risk of stroke. Evidence suggests that abnormalities of the cerebral microcirculation, such as capillary rarefication, may be present in these patients. We evaluated whether the presence of hypertension may affect the cerebral capillary architecture and function assessed by Intravoxel Incoherent Motion (IVIM) magnetic resonance imaging (MRI) in patients with continuous positive airway pressure (CPAP)-treated OSA. Methods Forty-one patients (88% male, mean age 57 ± 10 years) with moderate-to-severe OSA were selected and divided into two groups (normotensive vs. hypertensive). All hypertensive OSA patients were adherent with their antihypertensive medication. Cerebral microvascular structure was assessed in grey (GM) and white matter (WM) using an echo-planar diffusion imaging sequence with 14 different b values. A step-wise IVIM analysis algorithm was applied to compute true diffusion (D), perfusion fraction (f) and pseudo-diffusion (D*) values. Group comparisons were performed with the Wilcoxon-Mann-Whitney-Test. Regression analysis was adjusted for age. Results Diffusion- and perfusion-related indexes in middle-aged OSA normotensive patients were quantified in both tissue types (D [10−3 mm2/s]: GM = 0.83 ± 0.03; WM = 0.72 ± 0.03; f (%) GM = 0.09 ± 0.01; WM = 0.06 ± 0.01; D* [10−3 mm2/s]: GM = 7.72 ± 0.89; WM = 7.38 ± 0.98). In the examined tissue types, hypertension did not result in changes on the estimated MRI IVIM index values. Conclusion Based on IVIM analysis, cerebral microvascular structure and function showed no difference between hypertensive and normotensive patients with moderate-to-severe OSA treated with CPAP. Treatment adherence with antihypertensive drug regime and, in turn, controlled hypertension seems not to affect microvascular structure and perfusion of the brain. Trial registration ClinicalTrials.gov Identifier: NCT02493673

Discrimination of HPV status using CT texture analysis: tumour heterogeneity in oropharyngeal squamous cell carcinomas Abstract Purpose To evaluate the diagnostic performance of texture analysis for discriminating human papillomavirus (HPV) status in patients with oropharyngeal squamous cell carcinoma (OPSCC) in the primary tumours and metastatic lymph nodes. Methods Ninety-five patients with primary tumour and 91 with metastatic lymph nodes with confirmed HPV status, who underwent pretreatment contrast-enhanced CT (CECT), were included as the discovery population. CT texture analysis was performed using commercially available software. Differences between HPV-positive and HPV-negative groups were analysed using the χ2 test (or Mann-Whitney U test) and independent t test (or Fisher’s exact test). ROC curve analysis was performed to discriminate HPV status according to heterogeneity parameters. Diagnostic accuracy was evaluated in the separate validation population (n = 36) from an outside hospital. Results HPV positivity was 52.6% for primary tumours and 56.0% for metastatic lymph nodes. The entropy and standard deviation (SD) values in the HPV-positive group were significantly lower. Entropy using the medium filter was the best discriminator between HPV-positive and HPV-negative primary OPSCCs (AUC, 0.85) and SD without the filter for metastatic lymph nodes (AUC, 0.82). Diagnostic accuracy of entropy for the primary tumour was 80.0% in the discovery group and 75.0% in the validation group. In cases of metastatic lymph node, the accuracy of SD was 79.1% and 78.8%, respectively. Conclusion Significant differences were found in heterogeneity parameters from texture analysis of pretreatment CECT, according to HPV status. Texture analysis could be used as an adjunctive tool for diagnosis of HPV status in clinical practice.

Convolutional neural network-based segmentation can help in assessing the substantia nigra in neuromelanin MRI Abstract Purpose This study aimed to evaluate the accuracy and diagnostic test performance of the U-net-based segmentation method in neuromelanin magnetic resonance imaging (NM-MRI) compared to the established manual segmentation method for Parkinson’s disease (PD) diagnosis. Methods NM-MRI datasets from two different 3T-scanners were used: a “principal dataset” with 122 participants and an “external validation dataset” with 24 participants, including 62 and 12 PD patients, respectively. Two radiologists performed SNpc manual segmentation. Inter-reader precision was determined using Dice coefficients. The U-net was trained with manual segmentation as ground truth and Dice coefficients used to measure accuracy. Training and validation steps were performed on the principal dataset using a 4-fold cross-validation method. We tested the U-net on the external validation dataset. SNpc hyperintense areas were estimated from U-net and manual segmentation masks, replicating a previously validated thresholding method, and their diagnostic test performances for PD determined. Results For SNpc segmentation, U-net accuracy was comparable to inter-reader precision in the principal dataset (Dice coefficient: U-net, 0.83 ± 0.04; inter-reader, 0.83 ± 0.04), but lower in external validation dataset (Dice coefficient: U-net, 079 ± 0.04; inter-reader, 0.85 ± 0.03). Diagnostic test performances for PD were comparable between U-net and manual segmentation methods in both principal (area under the receiver operating characteristic curve: U-net, 0.950; manual, 0.948) and external (U-net, 0.944; manual, 0.931) datasets. Conclusion U-net segmentation provided relatively high accuracy in the evaluation of the SNpc in NM-MRI and yielded diagnostic performance comparable to that of the established manual method.

Fractional anisotropy of the optic radiations correlates with the visual field after epilepsy surgery Abstract Purpose This study assessed whether optic radiations (OR) microstructure after temporal lobe epilepsy (TLE) surgery correlated with visual field defects (VFD). Methods Patients were subjected to diffusion tensor imaging (DTI) tractography of the OR and Humphrey perimetry after TLE surgery. We used Spearman’s test to verify correlations between tractographic parameters and perimetry mean deviation. Tractographic variables were compared between patients with VFD or intact perimetry. Multiple logistic regression was applied between DTI and perimetry values. DTI sensitivity and specificity were assessed with a receiver operating characteristic (ROC) curve to evaluate VFD. Results Thirty-nine patients had reliable perimetry and OR tractography. There was a significant correlation between (1) fractional anisotropy (FA) and both total (rho = 0.569, p = 0.0002) and quadrant (rho = 0.453, p = 0.0037) mean deviation and (2) radial diffusivity and total mean deviation (rho = − 0.350, p = 0.0286). There was no other significant correlation. Patients with VFD showed a significantly lower FA compared with patients with normal perimetry (p = 0.0055), and a 0.01 reduction in FA was associated with a 44% increase in presenting VFD after surgery (confidence interval, CI = 1.10–1.88; p = 0.0082). Using a FA of 0.457, DTI tractography showed a specificity of 95.2% and a sensitivity of 50% to detect VFD after surgery (area under the curve = 0.7619, CI = 0.6020–0.9218). Conclusion The postoperative OR microstructure correlated with visual loss after epilepsy surgery. DTI postoperative OR tractography may be helpful in evaluating VFD.