Deep learning for automatic segmentation of vestibular schwannoma: a retrospective study from multi-center routine MRI

Automatic segmentation of vestibular schwannoma (VS) from routine clinical MRI has potential to improve clinical workflow, facilitate treatment decisions, and assist patient management. Previous work demonstrated reliable automatic segmentation performance on datasets of standardized MRI images acquired for stereotactic surgery planning. However, diagnostic clinical datasets are generally more diverse and pose a larger challenge to automatic segmentation algorithms, especially when post-operative images are included. In this work, we show for the first time that automatic segmentation of VS on routine MRI datasets is also possible with high accuracy. We acquired and publicly release a curated multi-center routine clinical (MC-RC) dataset of 160 patients with a single sporadic VS. For each patient up to three longitudinal MRI exams with contrast-enhanced T1-weighted (ceT1w) (n = 124) and T2-weighted (T2w) (n = 363) images were included and the VS manually annotated. Segmentations were produced and verified in an iterative process: (1) initial segmentations by a specialized company; (2) review by one of three trained radiologists; and (3) validation by an expert team. Inter- and intra-observer reliability experiments were performed on a subset of the dataset. A state-of-the-art deep learning framework was used to train segmentation models for VS. Model performance was evaluated on a MC-RC hold-out testing set, another public VS datasets, and a partially public dataset. The generalizability and robustness of the VS deep learning segmentation models increased significantly when trained on the MC-RC dataset. Dice similarity coefficients (DSC) achieved by our model are comparable to those achieved by trained radiologists in the inter-observer experiment. On the MC-RC testing set, median DSCs were 86.2(9.5) for ceT1w, 89.4(7.0) for T2w, and 86.4(8.6) for combined ceT1w+T2w input images. On another public dataset acquired for Gamma Knife stereotactic radiosurgery our model achieved median DSCs of 95.3(2.9), 92.8(3.8), and 95.5(3.3), respectively. In contrast, models trained on the Gamma Knife dataset did not generalize well as illustrated by significant underperformance on the MC-RC routine MRI dataset, highlighting the importance of data variability in the development of robust VS segmentation models. The MC-RC dataset and all trained deep learning models were made available online.

Structural- and DTI- MRI enable automated prediction of IDH Mutation Status in CNS WHO Grade 2-4 glioma patients: a deep Radiomics Approach

BACKGROUND

The role of isocitrate dehydrogenase (IDH) mutation status for glioma stratification and prognosis is established. While structural magnetic resonance image (MRI) is a promising biomarker, it may not be sufficient for non-invasive characterisation of IDH mutation status. We investigated the diagnostic value of combined diffusion tensor imaging (DTI) and structural MRI enhanced by a deep radiomics approach based on convolutional neural networks (CNNs) and support vector machine (SVM), to determine the IDH mutation status in Central Nervous System World Health Organization (CNS WHO) grade 2-4 gliomas.

METHODS

This retrospective study analyzed the DTI-derived fractional anisotropy (FA) and mean diffusivity (MD) images and structural images including fluid attenuated inversion recovery (FLAIR), non-enhanced T1-, and T2-weighted images of 206 treatment-naïve gliomas, including 146 IDH mutant and 60 IDH-wildtype ones. The lesions were manually segmented by experienced neuroradiologists and the masks were applied to the FA and MD maps. Deep radiomics features were extracted from each subject by applying a pre-trained CNN and statistical description. An SVM classifier was applied to predict IDH status using imaging features in combination with demographic data.

RESULTS

We comparatively assessed the CNN-SVM classifier performance in predicting IDH mutation status using standalone and combined structural and DTI-based imaging features. Combined imaging features surpassed stand-alone modalities for the prediction of IDH mutation status [area under the curve (AUC) = 0.846; sensitivity = 0.925; and specificity = 0.567]. Importantly, optimal model performance was noted following the addition of demographic data (patients' age) to structural and DTI imaging features [area under the curve (AUC) = 0.847; sensitivity = 0.911; and specificity = 0.617].

CONCLUSIONS

Imaging features derived from DTI-based FA and MD maps combined with structural MRI, have superior diagnostic value to that provided by standalone structural or DTI sequences. In combination with demographic information, this CNN-SVM model offers a further enhanced non-invasive prediction of IDH mutation status in gliomas.

Fluid suppression in amide proton transfer-weighted (APTw) CEST imaging: New theoretical insights and clinical benefits

PURPOSE

Amide proton transfer-weighted (APTw) MRI at 3T provides a unique contrast for brain tumor imaging. However, APTw imaging suffers from hyperintensities in liquid compartments such as cystic or necrotic structures and provides a distorted APTw signal intensity. Recently, it has been shown that heuristically motivated fluid suppression can remove such artifacts and significantly improve the readability of APTw imaging.

THEORY AND METHODS

In this work, we show that the fluid suppression can actually be understood by the known concept of spillover dilution, which itself can be derived from the Bloch-McConnell equations in comparison to the heuristic approach. Therefore, we derive a novel post-processing formula that efficiently removes fluid artifact, and explains previous approaches. We demonstrate the utility of this APTw assessment in silico, in vitro, and in vivo in brain tumor patients acquired at MR scanners from different vendors.

RESULTS

Our results show a reduction of the CEST signals from fluid environments while keeping the APTw-CEST signal intensity almost unchanged for semi-solid tissue structures such as the contralateral normal appearing white matter. This further allows us to use the same color bar settings as for conventional APTw imaging.

CONCLUSION

Fluid suppression has considerable value in improving the readability of APTw maps in the neuro-oncological field. In this work, we derive a novel post-processing formula from the underlying Bloch-McConnell equations that efficiently removes fluid artifact, and explains previous approaches which justify the derivation of this metric from a theoretical point of view, to reassure the scientific and medical field about its use.

Intraoperative MRI-Guided Resection Is Not Superior to 5-Aminolevulinic Acid Guidance in Newly Diagnosed Glioblastoma: A Prospective Controlled Multicenter Clinical Trial

PURPOSE

Prospective data suggested a superiority of intraoperative MRI (iMRI) over 5-aminolevulinic acid (5-ALA) for achieving complete resections of contrast enhancement in glioblastoma surgery. We investigated this hypothesis in a prospective clinical trial and correlated residual disease volumes with clinical outcome in newly diagnosed glioblastoma.

METHODS

This is a prospective controlled multicenter parallel-group trial with two center-specific treatment arms (5-ALA and iMRI) and blinded evaluation. The primary end point was complete resection of contrast enhancement on early postoperative MRI. We assessed resectability and extent of resection by an independent blinded centralized review of preoperative and postoperative MRI with 1-mm slices. Secondary end points included progression-free survival (PFS) and overall survival (OS), patient-reported quality of life, and clinical parameters.

RESULTS

We recruited 314 patients with newly diagnosed glioblastomas at 11 German centers. A total of 127 patients in the 5-ALA and 150 in the iMRI arm were analyzed in the as-treated analysis. Complete resections, defined as a residual tumor ≤0.175 cm³, were achieved in 90 patients (78%) in the 5-ALA and 115 (81%) in the iMRI arm (P = .79). Incision-suture times (P < .001) were significantly longer in the iMRI arm (316 v 215 [5-ALA] minutes). Median PFS and OS were comparable in both arms. The lack of any residual contrast enhancing tumor (0 cm³) was a significant favorable prognostic factor for PFS (P < .001) and OS (P = .048), especially in methylguanine-DNA-methyltransferase unmethylated tumors (P = .006).

CONCLUSION

We could not confirm superiority of iMRI over 5-ALA for achieving complete resections. Neurosurgical interventions in newly diagnosed glioblastoma shall aim for safe complete resections with 0 cm³ contrast-enhancing residual disease, as any other residual tumor volume is a negative predictor for PFS and OS.

Investigating the value of radiomics stemming from DSC quantitative biomarkers in IDH mutation prediction in gliomas

Objective

This study aims to assess the value of biomarker based radiomics to predict IDH mutation in gliomas. The patient cohort consists of 160 patients histopathologicaly proven of primary glioma (WHO grades 2-4) from 3 different centers.

Methods

To quantify the DSC perfusion signal two different mathematical modeling methods were used (Gamma fitting, leakage correction algorithms) considering the assumptions about the compartments contributing in the blood flow between the extra- and intra vascular space.

Results

The Mean slope of increase (MSI) and the K1 parameter of the bidirectional exchange model exhibited the highest performance with (ACC 74.3% AUROC 74.2%) and (ACC 75% AUROC 70.5%) respectively.

Conclusion

The proposed framework on DSC-MRI radiogenomics in gliomas has the potential of becoming a reliable diagnostic support tool exploiting the mathematical modeling of the DSC signal to characterize IDH mutation status through a more reproducible and standardized signal analysis scheme for facilitating clinical translation.

Multi-class glioma segmentation on real-world data with missing MRI sequences: comparison of three deep learning algorithms

This study tests the generalisability of three Brain Tumor Segmentation (BraTS) challenge models using a multi-center dataset of varying image quality and incomplete MRI datasets. In this retrospective study, DeepMedic, no-new-Unet (nn-Unet), and NVIDIA-net (nv-Net) were trained and tested using manual segmentations from preoperative MRI of glioblastoma (GBM) and low-grade gliomas (LGG) from the BraTS 2021 dataset (1251 in total), in addition to 275 GBM and 205 LGG acquired clinically across 12 hospitals worldwide. Data was split into 80% training, 5% validation, and 15% internal test data. An additional external test-set of 158 GBM and 69 LGG was used to assess generalisability to other hospitals' data. All models' median Dice similarity coefficient (DSC) for both test sets were within, or higher than, previously reported human inter-rater agreement (range of 0.74-0.85). For both test sets, nn-Unet achieved the highest DSC (internal = 0.86, external = 0.93) and the lowest Hausdorff distances (10.07, 13.87 mm, respectively) for all tumor classes (p < 0.001). By applying Sparsified training, missing MRI sequences did not statistically affect the performance. nn-Unet achieves accurate segmentations in clinical settings even in the presence of incomplete MRI datasets. This facilitates future clinical adoption of automated glioma segmentation, which could help inform treatment planning and glioma monitoring.

Diffusion MRI of the facial-vestibulocochlear nerve complex: a prospective clinical validation study

OBJECTIVES

Surgical planning of vestibular schwannoma surgery would benefit greatly from a robust method of delineating the facial-vestibulocochlear nerve complex with respect to the tumour. This study aimed to optimise a multi-shell readout-segmented diffusion-weighted imaging (rs-DWI) protocol and develop a novel post-processing pipeline to delineate the facial-vestibulocochlear complex within the skull base region, evaluating its accuracy intraoperatively using neuronavigation and tracked electrophysiological recordings.

METHODS

In a prospective study of five healthy volunteers and five patients who underwent vestibular schwannoma surgery, rs-DWI was performed and colour tissue maps (CTM) and probabilistic tractography of the cranial nerves were generated. In patients, the average symmetric surface distance (ASSD) and 95% Hausdorff distance (HD-95) were calculated with reference to the neuroradiologist-approved facial nerve segmentation. The accuracy of patient results was assessed intraoperatively using neuronavigation and tracked electrophysiological recordings.

RESULTS

Using CTM alone, the facial-vestibulocochlear complex of healthy volunteer subjects was visualised on 9/10 sides. CTM were generated in all 5 patients with vestibular schwannoma enabling the facial nerve to be accurately identified preoperatively. The mean ASSD between the annotators' two segmentations was 1.11 mm (SD 0.40) and the mean HD-95 was 4.62 mm (SD 1.78). The median distance from the nerve segmentation to a positive stimulation point was 1.21 mm (IQR 0.81-3.27 mm) and 2.03 mm (IQR 0.99-3.84 mm) for the two annotators, respectively.

CONCLUSIONS

rs-DWI may be used to acquire dMRI data of the cranial nerves within the posterior fossa.

CLINICAL RELEVANCE STATEMENT

Readout-segmented diffusion-weighted imaging and colour tissue mapping provide 1-2 mm spatially accurate imaging of the facial-vestibulocochlear nerve complex, enabling accurate preoperative localisation of the facial nerve. This study evaluated the technique in 5 healthy volunteers and 5 patients with vestibular schwannoma.

KEY POINTS

• Readout-segmented diffusion-weighted imaging (rs-DWI) with colour tissue mapping (CTM) visualised the facial-vestibulocochlear nerve complex on 9/10 sides in 5 healthy volunteer subjects. • Using rs-DWI and CTM, the facial nerve was visualised in all 5 patients with vestibular schwannoma and within 1.21-2.03 mm of the nerve's true intraoperative location. • Reproducible results were obtained on different scanners.

Assessment of Computed Tomography Perfusion Research Landscape: A Topic Modeling Study

The number of scholarly articles continues to rise. The continuous increase in scientific output poses a challenge for researchers, who must devote considerable time to collecting and analyzing these results. The topic modeling approach emerges as a novel response to this need. Considering the swift advancements in computed tomography perfusion (CTP), we deem it essential to launch an initiative focused on topic modeling. We conducted a comprehensive search of the Scopus database from 1 January 2000 to 16 August 2023, to identify relevant articles about CTP. Using the BERTopic model, we derived a group of topics along with their respective representative articles. For the 2020s, linear regression models were used to identify and interpret trending topics. From the most to the least prevalent, the topics that were identified include "Tumor Vascularity", "Stroke Assessment", "Myocardial Perfusion", "Intracerebral Hemorrhage", "Imaging Optimization", "Reperfusion Therapy", "Postprocessing", "Carotid Artery Disease", "Seizures", "Hemorrhagic Transformation", "Artificial Intelligence", and "Moyamoya Disease". The model provided insights into the trends of the current decade, highlighting "Postprocessing" and "Artificial Intelligence" as the most trending topics.

Adaptive Peer Tutoring and Insights From a Neurooncology Course

Peer teaching in medicine is a valuable educational approach that benefits students and tutors alike. The COVID-19 pandemic has significantly impacted the advancement of remote education in the medical field. In response, the Cerrahpasa Neuroscience Society organized a web-based, volunteer-based peer tutoring program to introduce students to central nervous system tumors. This viewpoint examines our peer mentoring experience in medical education. We discussed how we shaped the course, its positive effects, and the flexible nature of the course, which brought medical students from different regions together. In addition to evaluating academic results, we examined the social relations made possible by this unique teaching method by analyzing student feedback and test scores. Finally, we discussed the promise of global web-based mentoring, highlighting its significance in the dynamic and global context of medicine.

The Feasibility of Dose-Escalated Radiation Therapy for Glioblastoma Using Biological Image Guided Adaptive Radiotherapy (BIGART)

PURPOSE/OBJECTIVE(S)

Radiotherapy dose escalation (DE) for glioblastoma (GBM) has been an area of active research. We aimed to study the imaging changes within residual gross tumor volumes (GTV) through the course of concurrent chemo-radiotherapy (CCRT) using multiparametric magnetic resonance imaging (mpMRI). Diffusion, perfusion and chemical exchange saturation transfer (CEST) characteristics of the tumor and its microenvironment were investigated to identify a GTV subvolume potentially associated with radio resistance. We used biological image-guided adaptive radiotherapy (BIGART) to study the feasibility of DE to this GTV subvolume using either photon or proton beam therapy (PBT).

MATERIALS/METHODS

We prospectively identified GBM patients with >5cc residual tumor post-resection who were candidates for radical CCRT (60 Gy in 30 daily fractions over 6-weeks with concurrent temozolomide 75mg/m2 daily). We observed the imaging changes with serial mpMRI scans done at baseline, after 2, 4 and 6-weeks of CCRT. Regions of interest (ROIs) within the GTV associated with the following abnormal values at week 2 were identified: apparent diffusion coefficient (ADC): 750-1000 ×10-6 mm2/s; relative cerebral blood volume (rCBV): 1.75-6; and APT-w CEST signal intensity >1.79%. The overlap regions of these ROIs were defined as a novel biological target volume (BTV), identifying the potential area of maximal radioresistance. An in silico study was performed using a technology company's treatment planning system to evaluate the feasibility of planning adaptive treatment to the BTV to total dose of 75 Gy in 30 fractions. This is given in two phases: 30 Gy/15# to the whole PTV as per standard practice, followed by a simultaneous integrated boost (SIB) of 45 Gy/15 fractions while maintaining the dose to the rest of the of the PTV to 60 Gy. Either photons or PBT were used to keep doses to organs at risk (OARs) within standard clinical tolerances.

RESULTS

Nine patients were recruited for this analysis and a total of 27 mpMRI scans were studied. Median BTVs to GTVs ratio was 35% (range 22-47%). Volumetric-modulated arc (VMAT) photon and PBT adaptive plans for dose escalation to BTVs were created in all cases whilst maintaining OAR tolerances. Both VMAT and PBT provided acceptable target coverage with average BTV-PTV D98% of 73 Gy (range 71.5-73.8 Gy) and average D2% of 76 Gy (range 75.4-77 Gy) while effectively sparing OARs. Sharper dose gradient between DE-BTV and PTV was achieved with VMAT. PBT was particularly advantageous in minimizing the low-dose spillage outside the BTV.

CONCLUSION

We hereby propose a platform for adaptive radiotherapy to GBM tumors with biological-image guidance through the utilization of mpMRI to evaluate the tumor and its microenvironment during CCRT. We identified thresholds for tumor sub volumes showing the most resistant imaging features and created precise BTVs that allowed for dose escalation. PBT represents an additional useful tool for BIGART planning that will be investigated further in our ongoing work.

TISS-net: Brain tumor image synthesis and segmentation using cascaded dual-task networks and error-prediction consistency

Accurate segmentation of brain tumors from medical images is important for diagnosis and treatment planning, and it often requires multi-modal or contrast-enhanced images. However, in practice some modalities of a patient may be absent. Synthesizing the missing modality has a potential for filling this gap and achieving high segmentation performance. Existing methods often treat the synthesis and segmentation tasks separately or consider them jointly but without effective regularization of the complex joint model, leading to limited performance. We propose a novel brain Tumor Image Synthesis and Segmentation network (TISS-Net) that obtains the synthesized target modality and segmentation of brain tumors end-to-end with high performance. First, we propose a dual-task-regularized generator that simultaneously obtains a synthesized target modality and a coarse segmentation, which leverages a tumor-aware synthesis loss with perceptibility regularization to minimize the high-level semantic domain gap between synthesized and real target modalities. Based on the synthesized image and the coarse segmentation, we further propose a dual-task segmentor that predicts a refined segmentation and error in the coarse segmentation simultaneously, where a consistency between these two predictions is introduced for regularization. Our TISS-Net was validated with two applications: synthesizing FLAIR images for whole glioma segmentation, and synthesizing contrast-enhanced T1 images for Vestibular Schwannoma segmentation. Experimental results showed that our TISS-Net largely improved the segmentation accuracy compared with direct segmentation from the available modalities, and it outperformed state-of-the-art image synthesis-based segmentation methods.

Responsible AI practice and AI education are central to AI implementation: a rapid review for all medical imaging professionals in Europe

Artificial intelligence (AI) has transitioned from the lab to the bedside, and it is increasingly being used in healthcare. Radiology and Radiography are on the frontline of AI implementation, because of the use of big data for medical imaging and diagnosis for different patient groups. Safe and effective AI implementation requires that responsible and ethical practices are upheld by all key stakeholders, that there is harmonious collaboration between different professional groups, and customised educational provisions for all involved. This paper outlines key principles of ethical and responsible AI, highlights recent educational initiatives for clinical practitioners and discusses the synergies between all medical imaging professionals as they prepare for the digital future in Europe. Responsible and ethical AI is vital to enhance a culture of safety and trust for healthcare professionals and patients alike. Educational and training provisions for medical imaging professionals on AI is central to the understanding of basic AI principles and applications and there are many offerings currently in Europe. Education can facilitate the transparency of AI tools, but more formalised, university-led training is needed to ensure the academic scrutiny, appropriate pedagogy, multidisciplinarity and customisation to the learners' unique needs are being adhered to. As radiographers and radiologists work together and with other professionals to understand and harness the benefits of AI in medical imaging, it becomes clear that they are faced with the same challenges and that they have the same needs. The digital future belongs to multidisciplinary teams that work seamlessly together, learn together, manage risk collectively and collaborate for the benefit of the patients they serve.

The Advent of Generative Language Models in Medical Education

Artificial intelligence (AI) and generative language models (GLMs) present significant opportunities for enhancing medical education, including the provision of realistic simulations, digital patients, personalized feedback, evaluation methods, and the elimination of language barriers. These advanced technologies can facilitate immersive learning environments and enhance medical students' educational outcomes. However, ensuring content quality, addressing biases, and managing ethical and legal concerns present obstacles. To mitigate these challenges, it is necessary to evaluate the accuracy and relevance of AI-generated content, address potential biases, and develop guidelines and policies governing the use of AI-generated content in medical education. Collaboration among educators, researchers, and practitioners is essential for developing best practices, guidelines, and transparent AI models that encourage the ethical and responsible use of GLMs and AI in medical education. By sharing information about the data used for training, obstacles encountered, and evaluation methods, developers can increase their credibility and trustworthiness within the medical community. In order to realize the full potential of AI and GLMs in medical education while mitigating potential risks and obstacles, ongoing research and interdisciplinary collaboration are necessary. By collaborating, medical professionals can ensure that these technologies are effectively and responsibly integrated, contributing to enhanced learning experiences and patient care.

The role of DSC MR perfusion in predicting IDH mutation and 1p19q codeletion status in gliomas: meta-analysis and technical considerations

PURPOSE

Isocitrate dehydrogenase (IDH) mutation and 1p19q codeletion status are important for managing glioma patients. However, current practice dictates invasive tissue sampling for histomolecular classification. We investigated the current value of dynamic susceptibility contrast (DSC) MR perfusion imaging as a tool for the non-invasive identification of these biomarkers.

METHODS

A systematic search of PubMed, Medline, and Embase up to 2023 was performed, and meta-analyses were conducted. We removed studies employing machine learning models or using multiparametric imaging. We used random-effects standardized mean difference (SMD) and bivariate sensitivity-specificity meta-analyses, calculated the area under the hierarchical summary receiver operating characteristic curve (AUC) and performed meta-regressions using technical acquisition parameters (e.g., time to echo [TE], repetition time [TR]) as moderators to explore sources of heterogeneity. For all estimates, 95% confidence intervals (CIs) are provided.

RESULTS

Sixteen eligible manuscripts comprising 1819 patients were included in the quantitative analyses. IDH mutant (IDHm) gliomas had lower rCBV values compared to their wild-type (IDHwt) counterparts. The highest SMD was observed for rCBV_mean, rCBV_max, and rCBV 75^th percentile (SMD≈ - 0.8, 95% CI ≈ [- 1.2, - 0.5]). In meta-regression, shorter TEs, shorter TRs, and smaller slice thicknesses were linked to higher absolute SMDs. When discriminating IDHm from IDHwt, the highest pooled specificity was observed for rCBV_mean (82% [72, 89]), and the highest pooled sensitivity (i.e., 92% [86, 93]) and AUC (i.e., 0.91) for rCBV 10^th percentile. In the bivariate meta-regression, shorter TEs and smaller slice gaps were linked to higher pooled sensitivities. In IDHm, 1p19q codeletion was associated with higher rCBVmean (SMD = 0.9 [0.2, 1.5]) and rCBV 90^th percentile (SMD = 0.9 [0.1, 1.7]) values.

CONCLUSIONS

Identification of vascular signatures predictive of IDH and 1p19q status is a novel promising application of DSC perfusion. Standardization of acquisition protocols and post-processing of DSC perfusion maps are warranted before widespread use in clinical practice.

Phenotyping Superagers Using Resting-State fMRI

BACKGROUND AND PURPOSE

Superagers are defined as older adults with episodic memory performance similar or superior to that in middle-aged adults. This study aimed to investigate the key differences in discriminative networks and their main nodes between superagers and cognitively average elderly controls. In addition, we sought to explore differences in sensitivity in detecting these functional activities across the networks at 3T and 7T MR imaging fields.

MATERIALS AND METHODS

Fifty-five subjects 80 years of age or older were screened using a detailed neuropsychological protocol, and 31 participants, comprising 14 superagers and 17 cognitively average elderly controls, were included for analysis. Participants underwent resting-state-fMRI at 3T and 7T MR imaging. A prediction classification algorithm using a penalized regression model on the measurements of the network was used to calculate the probabilities of a healthy older adult being a superager. Additionally, ORs quantified the influence of each node across preselected networks.

RESULTS

The key networks that differentiated superagers and elderly controls were the default mode, salience, and language networks. The most discriminative nodes (ORs > 1) in superagers encompassed areas in the precuneus posterior cingulate cortex, prefrontal cortex, temporoparietal junction, temporal pole, extrastriate superior cortex, and insula. The prediction classification model for being a superager showed better performance using the 7T compared with 3T resting-state-fMRI data set.

CONCLUSIONS

Our findings suggest that the functional connectivity in the default mode, salience, and language networks can provide potential imaging biomarkers for predicting superagers. The 7T field holds promise for the most appropriate study setting to accurately detect the functional connectivity patterns in superagers.

Clinical indications and acquisition protocol for the use of dynamic contrast-enhanced MRI in head and neck cancer squamous cell carcinoma: recommendations from an expert panel

BACKGROUND

The clinical role of perfusion-weighted MRI (PWI) in head and neck squamous cell carcinoma (HNSCC) remains to be defined. The aim of this study was to provide evidence-based recommendations for the use of PWI sequence in HNSCC with regard to clinical indications and acquisition parameters.

METHODS

Public databases were searched, and selected papers evaluated applying the Oxford criteria 2011. A questionnaire was prepared including statements on clinical indications of PWI as well as its acquisition technique and submitted to selected panelists who worked in anonymity using a modified Delphi approach. Each panelist was asked to rate each statement using a 7-point Likert scale (1 = strongly disagree, 7 = strongly agree). Statements with scores equal or inferior to 5 assigned by at least two panelists were revised and re-submitted for the subsequent Delphi round to reach a final consensus.

RESULTS

Two Delphi rounds were conducted. The final questionnaire consisted of 6 statements on clinical indications of PWI and 9 statements on the acquisition technique of PWI. Four of 19 (21%) statements obtained scores equal or inferior to 5 by two panelists, all dealing with clinical indications. The Delphi process was considered concluded as reasons entered by panelists for lower scores were mainly related to the lack of robust evidence, so that no further modifications were suggested.

CONCLUSIONS

Evidence-based recommendations on the use of PWI have been provided by an independent panel of experts worldwide, encouraging a standardized use of PWI across university and research centers to produce more robust evidence.

CTNI-67. DUAL INHIBITION OF POST-RADIOGENIC ANGIO-VASCULOGENESIS BY OLAPTESED PEGOL (NOX-A12) AND BEVACIZUMAB IN GLIOBLASTOMA – INTERIM DATA FROM THE FIRST EXPANSION ARM OF THE GERMAN PHASE 1/2 GLORIA TRIAL

BACKGROUND

We recently reported favorable safety, promising clinical efficacy and immunohistochemical indicators of response after radiotherapy (RT) plus escalating doses of the CXCL12-neutralizing RNA-Spiegelmer olaptesed pegol (NOX-A12) for glioblastoma in the German multicenter phase 1/2 GLORIA trial (NCT04121455). Here, we report outcomes after RT plus dual inhibition of vasculogenesis (NOX-A12) and angiogenesis (bevacizumab).

METHODS

After establishing safety in the monotherapy arm, we enrolled six patients with incompletely resected GBM, ECOG ≤ 2, age ≥ 18 and without MGMT promoter hypermethylation into a pre-planned expansion arm. Patients received standard RT (60 Gy in 30 fractions), continuous i.v. infusions of NOX-A12 (600 mg/week) and i.v. infusions of bevacizumab (10 mg/kg q2w). The primary endpoint was safety. Secondary endpoints included radiographic response, perfusion/diffusion imaging and neurologic performance.

RESULTS

Dual treatment was well-tolerated and safe. Of all G ≥ 2 AEs (n = 37), two G2 events (5.4%) were deemed related to NOX-A12. There were no dose-limiting toxicities and no treatment-related deaths. Longitudinal NANO assessment revealed stable neurologic functioning in all patients. Five out of six patients achieved partial responses (PRs) as per mRANO in week 9. All PRs remained durable at a median follow up of 5.6 months (range 3.6 to 9.3 months). No progression occurred. The mean best response was -65.9% (-13.3% to -99.9%) for target lesion sums and -92.1% (-76.2% to -100%) for non-target lesion (NTL) sums. In all three patients with NTL at least one lesion disappeared. The mean best change from baseline of the highly perfused-tumor fraction was -84.5% (-51.9% to -100%) and the mean best change of the apparent diffusion coefficient was 20.1% (-24.5% to 59.1%).

CONCLUSION

Interim data of the ongoing trial confirm the previously established safety profile of NOX-A12 and suggest improved efficacy of dual inhibition of post-radiogenic angio- and vasculogenesis by the addition of bevacizumab.

Deep learning for prediction of isocitrate dehydrogenase mutation in gliomas: a critical approach, systematic review and meta-analysis of the diagnostic test performance using a Bayesian approach

Background

Conventionally, identifying isocitrate dehydrogenase (IDH) mutation in gliomas is based on histopathological analysis of tissue specimens acquired via stereotactic biopsy or definitive resection. Accurate pre-treatment prediction of IDH mutation status using magnetic resonance imaging (MRI) can guide clinical decision-making. We aim to evaluate the diagnostic performance of deep learning (DL) to determine IDH mutation status in gliomas.

Methods

A systematic search of Cochrane Library, Web of Science, Medline, and Scopus was conducted to identify relevant publications until August 1, 2021. Articles were included if all the following criteria were met: (I) patients with histopathologically confirmed World Health Organization (WHO) grade II, III, or IV gliomas; (II) histopathological examination with the IDH mutation; (III) DL was used to predict the IDH mutation status; (IV) sufficient data for reconstruction of confusion matrices in terms of the diagnostic performance of the DL algorithms; and (V) original research articles. Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) and Checklist for Artificial Intelligence in Medical Imaging (CLAIM) was used to assess the studies' quality. Bayes theorem was utilized to calculate the posttest probability.

Results

Four studies with a total of 1,295 patients were included. In the training set, the pooled sensitivity, specificity, and area under the summary receiver operating characteristic (SROC) curve were 93.9%, 90.9% and 0.958, respectively. In the validation set, the pooled sensitivity, specificity, and area under the SROC curve were 90.8%, 85.5% and 0.939, respectively. With a known pretest probability of 80.2%, the Bayes theorem yielded a posttest probability of 97.6% and 96.0% for a positive test and 27.0% and 30.6% for a negative test for training sets and validation sets, respectively.

Discussion

This is the first meta-analysis that summarizes the diagnostic performance of DL in predicting IDH mutation status in gliomas via the Bayes theorem. DL algorithms demonstrate excellent diagnostic performance in predicting IDH mutation in gliomas. Radiomic features associated with IDH mutation, and its underlying pathophysiology extracted from advanced MRI may improve prediction probability. However, more studies are required to optimize and increase its reliability. Limitations include obtaining some data via email and lack of training and test sets statistics.

Guidelines for magnetic resonance imaging in pediatric head and neck pathologies: a multicentre international consensus paper

The use of standardized imaging protocols is paramount in order to facilitate comparable, reproducible images and, consequently, to optimize patient care. Standardized MR protocols are lacking when studying head and neck pathologies in the pediatric population. We propose an international, multicenter consensus paper focused on providing the best combination of acquisition time/technical requirements and image quality. Distinct protocols for different regions of the head and neck and, in some cases, for specific pathologies or clinical indications are recommended. This white paper is endorsed by several international scientific societies and it is the result of discussion, in consensus, among experts in pediatric head and neck imaging.

The Utility of Conventional Amino Acid PET Radiotracers in the Evaluation of Glioma Recurrence also in Comparison with MRI

AIM

In this comprehensive review we present an update on the most relevant studies evaluating the utility of amino acid PET radiotracers for the evaluation of glioma recurrence as compared to magnetic resonance imaging (MRI).

METHODS

A literature search extended until June 2020 on the PubMed/MEDLINE literature database was conducted using the terms "high-grade glioma", "glioblastoma", "brain tumors", "positron emission tomography", "PET", "amino acid PET", "[¹¹C]methyl-l-methionine", "[¹⁸F]fluoroethyl-tyrosine", "[¹⁸F]fluoro-l-dihydroxy-phenylalanine", "MET", "FET", "DOPA", "magnetic resonance imaging", "MRI", "advanced MRI", "magnetic resonance spectroscopy", "perfusion-weighted imaging", "diffusion-weighted imaging", "MRS", "PWI", "DWI", "hybrid PET/MR", "glioma recurrence", "pseudoprogression", "PSP", "treatment-related change", and "radiation necrosis" alone and in combination. Only original articles edited in English and about humans with at least 10 patients were included.

RESULTS

Forty-four articles were finally selected. Conventional amino acid PET tracers were demonstrated to be reliable diagnostic techniques in differentiating tumor recurrence thanks to their high uptake from tumor tissue and low background in normal grey matter, giving additional and early information to standard modalities. Among them, MET-PET seems to present the highest diagnostic value but its use is limited to on-site cyclotron facilities. [¹⁸F]labelled amino acids, such as FDOPA and FET, were developed to provide a more suitable PET tracer for routine clinical applications, and demonstrated similar diagnostic performance. When compared to the gold standard MRI, amino acid PET provides complementary and comparable information to standard modalities and seems to represent an essential tool in the differentiation between tumor recurrence and other entities such as pseudoprogression, radiation necrosis, and pseudoresponse.

CONCLUSIONS

Despite the introduction of new advanced imaging techniques, the diagnosis of glioma recurrence remains challenging. In this scenario, the growing knowledge about imaging techniques and analysis, such as the combined PET/MRI and the application of artificial intelligence (AI) and machine learning (ML), could represent promising tools to face this difficult and debated clinical issue.

Dynamic contrast-enhanced MRI in malignant pleural mesothelioma: prediction of outcome based on DCE-MRI measurements in patients undergoing cytotoxic chemotherapy

Background

The malignant pleural mesothelioma (MPM) response rate to chemotherapy is low. The identification of imaging biomarkers that could help guide the most effective therapy approach for individual patients is highly desirable. Our aim was to investigate the dynamic contrast-enhanced (DCE) MR parameters as predictors for progression-free (PFS) and overall survival (OS) in patients with MPM treated with cisplatin-based chemotherapy.

Methods

Thirty-two consecutive patients with MPM were enrolled in this prospective study. Pretreatment and intratreatment DCE-MRI were scheduled in each patient. The DCE parameters were analyzed using the extended Tofts (ET) and the adiabatic approximation tissue homogeneity (AATH) model. Comparison analysis, logistic regression and ROC analysis were used to identify the predictors for the patient’s outcome.

Results

Patients with higher pretreatment ET and AATH-calculated K^trans and v_e values had longer OS (P≤.006). Patients with a more prominent reduction in ET-calculated K^trans and k_ep values during the early phase of chemotherapy had longer PFS (P =.008). No parameter was identified to predict PFS. Pre-treatment ET-calculated K^trans was found to be an independent predictive marker for longer OS (P=.02) demonstrating the most favourable discrimination performance compared to other DCE parameters with an estimated sensitivity of 89% and specificity of 78% (AUC 0.9, 95% CI 0.74-0.98, cut off > 0.08 min^-1).

Conclusions

In the present study, higher pre-treatment ET-calculated K^trans values were associated with longer OS. The results suggest that DCE-MRI might provide additional information for identifying MPM patients that may respond to chemotherapy.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-022-09277-x.

Repeatability of perfusion measurements in adult gliomas using pulsed and pseudo-continuous arterial spin labelling MRI

OBJECTIVES

To investigate the repeatability of perfusion measures in gliomas using pulsed- and pseudo-continuous-arterial spin labelling (PASL, PCASL) techniques, and evaluate different regions-of-interest (ROIs) for relative tumour blood flow (rTBF) normalisation.

MATERIALS AND METHODS

Repeatability of cerebral blood flow (CBF) was measured in the Contralateral Normal Appearing Hemisphere (CNAH) and in brain tumours (aTBF). rTBF was normalised using both large/small ROIs from the CNAH. Repeatability was evaluated with intra-class-correlation-coefficient (ICC), Within-Coefficient-of-Variation (WCoV) and Coefficient-of-Repeatability (CR).

RESULTS

PASL and PCASL demonstrated high reliability (ICC > 0.9) for CNAH-CBF, aTBF and rTBF. PCASL demonstrated a more stable signal-to-noise ratio (SNR) with a lower WCoV of the SNR than that of PASL (10.9-42.5% vs. 12.3-29.2%). PASL and PCASL showed higher WCoV in aTBF and rTBF than in CNAH CBF in WM and GM but not in the caudate, and higher WCoV for rTBF than for aTBF when normalised using a small ROI (PASL 8.1% vs. 4.7%, PCASL 10.9% vs. 7.9%, respectively). The lowest CR was observed for rTBF normalised with a large ROI.

DISCUSSION

PASL and PCASL showed similar repeatability for the assessment of perfusion parameters in patients with primary brain tumours as previous studies based on volunteers. Both methods displayed reasonable WCoV in the tumour area and CNAH. PCASL's more stable SNR in small areas (caudate) is likely to be due to the longer post-labelling delays.

CEST MRI provides amide/amine surrogate biomarkers for treatment-naïve glioma sub-typing

PURPOSE

Accurate glioma classification affects patient management and is challenging on non- or low-enhancing gliomas. This study investigated the clinical value of different chemical exchange saturation transfer (CEST) metrics for glioma classification and assessed the diagnostic effect of the presence of abundant fluid in glioma subpopulations.

METHODS

Forty-five treatment-naïve glioma patients with known isocitrate dehydrogenase (IDH) mutation and 1p/19q codeletion status received CEST MRI (B_1rms = 2μT, T_sat = 3.5 s) at 3 T. Magnetization transfer ratio asymmetry and CEST metrics (amides: offset range 3-4 ppm, amines: 1.5-2.5 ppm, amide/amine ratio) were calculated with two models: 'asymmetry-based' (AB) and 'fluid-suppressed' (FS). The presence of T2/FLAIR mismatch was noted.

RESULTS

IDH-wild type had higher amide/amine ratio than IDH-mutant_1p/19q^codel (p < 0.022). Amide/amine ratio and amine levels differentiated IDH-wild type from IDH-mutant (p < 0.0045) and from IDH-mutant_1p/19q^ret (p < 0.021). IDH-mutant_1p/19q^ret had higher amides and amines than IDH-mutant_1p/19q^codel (p < 0.035). IDH-mutant_1p/19q^ret with AB/FS mismatch had higher amines than IDH-mutant_1p/19q^ret without AB/FS mismatch ( < 0.016). In IDH-mutant_1p/19q^ret, the presence of AB/FS mismatch was closely related to the presence of T2/FLAIR mismatch (p = 0.014).

CONCLUSIONS

CEST-derived biomarkers for amides, amines, and their ratio can help with histomolecular staging in gliomas without intense contrast enhancement. T2/FLAIR mismatch is reflected in the presence of AB/FS CEST mismatch. The AB/FS CEST mismatch identifies glioma subgroups that may have prognostic and clinical relevance.

Artificial Intelligence in Medicine: A Multinational Multi-Center Survey on the Medical and Dental Students' Perception

Background: The emerging field of artificial intelligence (AI) will probably affect the practice for the next generation of doctors. However, the students' views on AI have not been largely investigated. Methods: An anonymous electronic survey on AI was designed for medical and dental students to explore: (1) sources of information about AI, (2) AI applications and concerns, (3) AI status as a topic in medicine, and (4) students' feelings and attitudes. The questionnaire was advertised on social media platforms in 2020. Security measures were employed to prevent fraudulent responses. Mann-Whitney U-test was employed for all comparisons. A sensitivity analysis was also performed by binarizing responses to express disagreement and agreement using the Chi-squared test. Results: Three thousand one hundred thirty-three respondents from 63 countries from all continents were included. Most respondents reported having at least a moderate understanding of the technologies underpinning AI and of their current application, with higher agreement associated with being male (p < 0.0001), tech-savvy (p < 0.0001), pre-clinical student (p < 0.006), and from a developed country (p < 0.04). Students perceive AI as a partner rather than a competitor (72.2%) with a higher agreement for medical students (p = 0.002). The belief that AI will revolutionize medicine and dentistry (83.9%) with greater agreement for students from a developed country (p = 0.0004) was noted. Most students agree that the AI developments will make medicine and dentistry more exciting (69.9%), that AI shall be part of the medical training (85.6%) and they are eager to incorporate AI in their future practice (99%). Conclusion: Currently, AI is a hot topic in medicine and dentistry. Students have a basic understanding of AI principles, a positive attitude toward AI and would like to have it incorporated into their training.

CTNI-45. RESPONSE EVALUATION OF AR-67 FROM A PHASE-2 RECURRENT GLIOBLASTOMA TRIAL BY ARTIFICIAL INTELLIGENCE ASSISTED TUMOR VOLUMETRIC ESTIMATION: COMPARISON WITH THE SUM OF THE PERPENDICULAR DIAMETERS PRODUCT

BACKGROUND

Modified Radiographic Response Assessment in Neuro-Oncology (“mRANO”) criteria based on SPDP form the basis for assessing treatment response in Glioblastoma but are subject to sampling bias and difficulty in differentiating between pseudo- and true disease progression. Volumetric image analysis using AI may overcome these limitations of standard techniques and improve our ability to detect changes earlier and more accurately.

METHODS

Images from eight reGBM patients enrolled in a Phase-2 reGBM study of Vivacitas Oncology’s drug, AR-67, were re-assessed using IAG’s AI-assisted volumetric measurements. A median of five MRI time points from each patient were included. The mRANO response was determined by central reading and tumor volumetric measurement using IAG’s proprietary platform. Statistical significance was set at p&lt; .0001.

RESULTS

Four patients showed responses, two patients showed stable disease, and two patients showed progressive disease. Tumor volume was correlated (r=0.97) with SPDP, but was driven by high coefficients in large lesions. Standard SPDP overestimated tumor size in larger tumors using the Bland-Altman analysis (mean difference: 829; 95% CI: 704 to -2362) leading to discrepancies in response rates. For example, the mean response rate based on IAG’s volumetric criteria was +22% (1.29) compared with +17% using SPDP (0.81). Eight out of 45 time-points also differed in the directionality of responses (e.g., increase vs. decrease) with SPDP underestimating the positive effects of AR-67 compared to AI analysis.

CONCLUSIONS

IAG’s AI-assisted tumor volumetric analysis is feasible for clinical trials and may be more sensitive for evaluating treatment-related response rates vs. SPDP methodology. This is particularly true for measuring large lesions, and may also allow for more accurately differentiating between pseudo- and true disease progression. The data included eight patients’ MRI images from a Phase-2 reGBM study, showing that five patients achieved the primary end-point of six months Progression-Free Survival, suggesting AR-67’s therapeutic potential.

CTNI-43. CXCL12 INHIBITION IN MGMT UNMETHYLATED GLIOBLASTOMA – RESULTS OF AN EARLY PROOF-OF-CONCEPT ASSESSMENT IN THE MULTICENTRIC PHASE I/II GLORIA TRIAL (NCT04121455)

BACKGROUND

Preclinical studies showed that CXCL12-mediated influx of highly angiogenic monocytes/macrophages is a key driver of tumor re-vascularization and re-growth after radiotherapy (RT) of glioblastoma (GBM). We report findings from a phase I/II proof-of concept (PoC) study on CXCL12 inhibition during and after RT of GBM.

METHODS

Patients ≥18years with incompletely or unresected GBM without MGMT promoter hypermethylation and ECOG≤2 were eligible to participate. Patients received continuous (24/7) i.v. infusions of 200mg/week (n=3), 400mg/week (n=3) or 600mg/week (n=3) of the CXCL12 inhibitor olaptesed pegol (OLA) for 26 weeks during and after normo- or hypofractionated RT (60Gy/40.05Gy). The primary endpoint was safety as per the incidence of treatment-related adverse events. The study was accompanied by PoC-research including multiparametric MRI biomarkers (relative cerebral blood volume, rCBV; fractional tumor burden with high perfusion, FTBhigh; apparent diffusion coefficient, ADC) and of multiplexed immunofluorescence imaging (CODEX®) of reference and patient samples. Initial results of these analyses are reported for the first six patients enrolled.

RESULTS

Five of six (83%) patients assessed with advanced MRI showed response under OLA in rCBV/FTBhigh and ADC. Maximum reduction in perfusion (rCBV) from baseline was 55%, maximum reduction of FTBhigh was 55% and maximum increase in ADC was 77%. Furthermore, five of six (83%) patients analyzed showed reduction of enhancing tissue volumes in at least one scan under OLA therapy. In both one patient and two reference samples CXCL12 co-localized with endothelial cells of the microvascular proliferation zone. In a paired sample (before/during OLA) of one patient, endothelial cells stained positive for CXCL12 before but not during treatment and almost all GBM cells were negative in Ki67 staining in the sample obtained under OLA therapy.

CONCLUSIONS

Advanced MRI and multiplexed immunofluorescence suggest efficacy of combined radiotherapy and CXCL12 inhibition in unmethylated GBM.

Funded by NOXXON Pharma AG; ClinicalTrials.gov number, NCT04121455.

Unconventional non-amino acidic PET radiotracers for molecular imaging in gliomas

PURPOSE

The objective of this review was to explore the potential clinical application of unconventional non-amino acid PET radiopharmaceuticals in patients with gliomas.

METHODS

A comprehensive search strategy was used based on SCOPUS and PubMed databases using the following string: ("perfusion" OR "angiogenesis" OR "hypoxia" OR "neuroinflammation" OR proliferation OR invasiveness) AND ("brain tumor" OR "glioma") AND ("Positron Emission Tomography" OR PET). From all studies published in English, the most relevant articles were selected for this review, evaluating the mostly used PET radiopharmaceuticals in research centers, beyond amino acid radiotracers and 2-[¹⁸F]fluoro-2-deoxy-D-glucose ([¹⁸F]FDG), for the assessment of different biological features, such as perfusion, angiogenesis, hypoxia, neuroinflammation, cell proliferation, tumor invasiveness, and other biological characteristics in patients with glioma.

RESULTS

At present, the use of non-amino acid PET radiopharmaceuticals specifically designed to assess perfusion, angiogenesis, hypoxia, neuroinflammation, cell proliferation, tumor invasiveness, and other biological features in glioma is still limited.

CONCLUSION

The use of investigational PET radiopharmaceuticals should be further explored considering their promising potential and studies specifically designed to validate these preliminary findings are needed. In the clinical scenario, advancements in the development of new PET radiopharmaceuticals and new imaging technologies (e.g., PET/MR and the application of the artificial intelligence to medical images) might contribute to improve the clinical translation of these novel radiotracers in the assessment of gliomas.

Segmentation of vestibular schwannoma from MRI, an open annotated dataset and baseline algorithm

Automatic segmentation of vestibular schwannomas (VS) from magnetic resonance imaging (MRI) could significantly improve clinical workflow and assist patient management. We have previously developed a novel artificial intelligence framework based on a 2.5D convolutional neural network achieving excellent results equivalent to those achieved by an independent human annotator. Here, we provide the first publicly-available annotated imaging dataset of VS by releasing the data and annotations used in our prior work. This collection contains a labelled dataset of 484 MR images collected on 242 consecutive patients with a VS undergoing Gamma Knife Stereotactic Radiosurgery at a single institution. Data includes all segmentations and contours used in treatment planning and details of the administered dose. Implementation of our automated segmentation algorithm uses MONAI, a freely-available open-source framework for deep learning in healthcare imaging. These data will facilitate the development and validation of automated segmentation frameworks for VS and may also be used to develop other multi-modal algorithmic models.

The Brain Metabolic Signature in Superagers Using In Vivo 1H-MRS: A Pilot Study

BACKGROUND AND PURPOSE

Youthful memory performance in older adults may reflect an underlying resilience to the conventional pathways of aging. Subjects having this unusual characteristic have been recently termed "superagers." This study aimed to explore the significance of imaging biomarkers acquired by ¹H-MRS to characterize superagers and to differentiate them from their normal-aging peers.

MATERIALS AND METHODS

Fifty-five patients older than 80 years of age were screened using a detailed neuropsychological protocol, and 25 participants, comprising 12 superagers and 13 age-matched controls, were statistically analyzed. We used state-of-the-art 3T ¹H-MR spectroscopy to quantify 18 neurochemicals in the posterior cingulate cortex of our subjects. All ¹H-MR spectroscopy data were analyzed using LCModel. Results were further processed using 2 approaches to investigate the technique accuracy: 1) comparison of the average concentration of metabolites estimated with Cramer-Rao lower bounds <20%; and 2) calculation and comparison of the weighted means of metabolites' concentrations.

RESULTS

The main finding observed was a higher total N-acetyl aspartate concentration in superagers than in age-matched controls using both approaches (P = .02 and P = .03 for the weighted means), reflecting a positive association of total N-acetyl aspartate with higher cognitive performance.

CONCLUSIONS

¹H-MR spectroscopy emerges as a promising technique to unravel neurochemical mechanisms related to cognitive aging in vivo and providing a brain metabolic signature in superagers. This may contribute to monitoring future interventional therapies to avoid or postpone the pathologic processes of aging.

Multicenter DSC-MRI-Based Radiomics Predict IDH Mutation in Gliomas

Simple Summary

Significant efforts have been put toward developing MRI-based radiogenomics for IDH status subtyping predictions; however, in the vast majority of these approaches, the external validation sets are absent. Another limitation in current studies is the lack of explainability in radiomics models, which hampers clinical trust and translation. Motivated by these challenges, we proposed a multicenter DSC–MRI-based radiomics study based on an independent exploratory set, which was externally validated on two independent cohorts, for IDH mutation status prediction. Our results demonstrated that DSC–MRI radiogenomics in gliomas, coupled with dynamic-based image standardization techniques, hold the potential to provide (a) increased predictive performance by offering models that generalize well, (b) reasoning behind the IDH mutation status predictions, and (c) interpretability of the radiomics features’ impacts in model performance.

Abstract

To address the current lack of dynamic susceptibility contrast magnetic resonance imaging (DSC–MRI)-based radiomics to predict isocitrate dehydrogenase (IDH) mutations in gliomas, we present a multicenter study that featured an independent exploratory set for radiomics model development and external validation using two independent cohorts. The maximum performance of the IDH mutation status prediction on the validation set had an accuracy of 0.544 (Cohen’s kappa: 0.145, F1-score: 0.415, area under the curve-AUC: 0.639, sensitivity: 0.733, specificity: 0.491), which significantly improved to an accuracy of 0.706 (Cohen’s kappa: 0.282, F1-score: 0.474, AUC: 0.667, sensitivity: 0.6, specificity: 0.736) when dynamic-based standardization of the images was performed prior to the radiomics. Model explainability using local interpretable model-agnostic explanations (LIME) and Shapley additive explanations (SHAP) revealed potential intuitive correlations between the IDH–wildtype increased heterogeneity and the texture complexity. These results strengthened our hypothesis that DSC–MRI radiogenomics in gliomas hold the potential to provide increased predictive performance from models that generalize well and provide understandable patterns between IDH mutation status and the extracted features toward enabling the clinical translation of radiogenomics in neuro-oncology.

Magnetic resonance features and cranial nerve involvement in pediatric head and neck rhabdomyosarcomas

PURPOSE

Rhabdomyosarcoma (RMS) is a malignant tumor frequent in children. The frequency and characteristics of cranial nerve involvement in pediatric head and neck (H&N) RMS have been scarcely reported. The aim of this study is to review a large cohort of pediatric head and neck RMS with an emphasis on cranial nerve involvement.

METHODS

We retrospectively reviewed H&N RMS cases from 3 tertiary hospitals over a 10-year period. Cranial nerve involvement was defined as radiologically apparent tumor extension along a nerve and/or the presence of secondary signs. Scans were reviewed by two pediatric neuroradiologists, blinded to clinical data.

RESULTS

A total of 52 patients met the inclusion criteria. Histologically, 39/52 were embryonal RMS, while 13/52 were alveolar RMS. Regional lymph nodes metastases were present in 19.2%. Cranial nerve involvement was present in 36.5%. Nerves were mainly involved as a direct extension of the mass through skull base foramina or after invasion of cavernous sinus, Meckel's cave, orbital apex, or stylomastoid foramen.

CONCLUSION

Cranial nerve involvement is frequent in pediatric head and neck RMS and occurs secondary to "geographic" invasion due to direct extension through skull base foramina or cavernous sinus. These tumors never showed distant perineural metastatic disease as is seen in cases of adult head and neck carcinomas. This implies a different biological interaction between the nerves and these tumors in comparison to adult H&N tumors.

The role of preoperative diffusion tensor imaging in predicting and improving functional outcome in pediatric patients undergoing epilepsy surgery: a systematic review

OBJECTIVE

Diffusion tensor imaging (DTI) is a useful neuroimaging technique for surgical planning in adult patients. However, no systematic review has been conducted to determine its utility for pre-operative analysis and planning of Pediatric Epilepsy surgery. We sought to determine the benefit of pre-operative DTI in predicting and improving neurological functional outcome after epilepsy surgery in children with intractable epilepsy.

METHODS

A systematic review of articles in English using PubMed, EMBASE and Scopus databases, from inception to January 10, 2020 was conducted. All studies that used DTI as either predictor or direct influencer of functional neurological outcome (motor, sensory, language and/or visual) in pediatric epilepsy surgical candidates were included. Data extraction was performed by two blinded reviewers. Risk of bias of each study was determined using the QUADAS 2 Scoring System.

RESULTS

13 studies were included (6 case reports/series, 5 retrospective cohorts, and 2 prospective cohorts) with a total of 229 patients. Seven studies reported motor outcome; three reported motor outcome prediction with a sensitivity and specificity ranging from 80 to 85.7 and 69.6 to 100%, respectively; four studies reported visual outcome. In general, the use of DTI was associated with a high degree of favorable neurological outcomes after epilepsy surgery.

CONCLUSION

Multiple studies show that DTI helps to create a tailored plan that results in improved functional outcome. However, more studies are required in order to fully assess its utility in pediatric patients. This is a desirable field of study because DTI offers a non-invasive technique more suitable for children.

ADVANCES IN KNOWLEDGE

This systematic review analyses, exclusively, studies of pediatric patients with drug-resistant epilepsy and provides an update of the evidence regarding the role of DTI, as part of the pre-operative armamentarium, in improving post-surgical neurological sequels and its potential for outcome prediction.

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).

Integrated use of tumor AI volumetrics and advanced MRI to assess early response to Temferon in a phase 2 GBM trial: A novel paradigm

e14041

Background: mRANO criteria form the basis for assessing treatment response in GBM clinical trials. mRANO measurements show variability due to limited sampling of the tumor morphology and reader estimations. Volumetrics may overcome limitations of linear measurements and improve our ability to detect changes reliably. Advanced MRI shed light on the mechanism of action and provide biomarkers, whose changes may precede any morphological alterations in the tumor. Methods: Six consecutive patients (3M, 3F, mean age 53.5 yrs) from an ongoing Phase-2 study on the effect of Temferon on recurrent GBM, were included and 3 consecutive MRI follow-up time points from each patient were entered in this interim analysis. The mRANO response status was determined by a central reader. Integrated volumetrics, two perfusion modalities (DSC/DCE) and diffusion analysis were performed on a proprietary platform after co-registering data in a common space to eliminate user bias and variability. Volumetrics and perfusion/diffusion biomarkers were estimated using state-of-the-art Artificial Intelligence (AI)-empowered algorithms. Statistical significance was set at p < 0.001. Results: At the time of the interim analysis, 3 patients experienced stable disease (SD) / partial response (pR) and 3 patients had progressive disease (PD). The tumor volume had significant correlation (r = 0.89) with the SPDP but the SPDP showed monotonously biased overestimation of the tumor burden in the Bland-Altman analysis with mean difference = 886 (95%CI: 64-1708). Changes assessed by SPDP in the tumor burden during the treatment were underestimated for low volume tumor and underestimated for sizeable tumors compared with the volumetrics. The optimal threshold to predict tumor progression was 26% volume increase (sensitivity/specificity: 67/93%, AUC = 0.85), whereas there was no statistically significant threshold for SPDP. The perfusion fractional tumor volumes (FTV) with low and intermediately elevated rCBV showed strong correlation with the SPDP and enhancement volumetrics (r = 0.92-0.98). ROC analysis showed that 14% increase in the high perfusion tumor could significantly predict progression (sensitivity/specificity: 79/71%, AUC = 0.71). The responders showed significant increase in mean DWI values (11±0.3%) vs decrease (-4±0.05%) in non-responders. The mean rCBV and Vp values in responders showed decrease (-3±0.45% and -33±0.1%) and were increased in non-responders (25±0.1% and 51±0.66%). Conclusions: We demonstrated significant bias in estimating treatment response by means of mRANO compared to tumor volumetrics. Distinct patterns of changes in tumor perfusion, diffusion and FTV correlated highly with treatment response and enhancing tumor burden. Volumetrics and perfusion/diffusion composite biomarkers predicted accurately the response status in our patients. Clinical trial information: NCT03866109.

Automated quantitative MRI volumetry reports support diagnostic interpretation in dementia: a multi-rater, clinical accuracy study

Objectives

We examined whether providing a quantitative report (QReport) of regional brain volumes improves radiologists’ accuracy and confidence in detecting volume loss, and in differentiating Alzheimer’s disease (AD) and frontotemporal dementia (FTD), compared with visual assessment alone.

Methods

Our forced-choice multi-rater clinical accuracy study used MRI from 16 AD patients, 14 FTD patients, and 15 healthy controls; age range 52–81. Our QReport was presented to raters with regional grey matter volumes plotted as percentiles against data from a normative population (n = 461). Nine raters with varying radiological experience (3 each: consultants, registrars, ‘non-clinical image analysts’) assessed each case twice (with and without the QReport). Raters were blinded to clinical and demographic information; they classified scans as ‘normal’ or ‘abnormal’ and if ‘abnormal’ as ‘AD’ or ‘FTD’.

Results

The QReport improved sensitivity for detecting volume loss and AD across all raters combined (p = 0.015* and p = 0.002*, respectively). Only the consultant group’s accuracy increased significantly when using the QReport (p = 0.02*). Overall, raters’ agreement (Cohen’s κ) with the ‘gold standard’ was not significantly affected by the QReport; only the consultant group improved significantly (κ_s 0.41➔0.55, p = 0.04*). Cronbach’s alpha for interrater agreement improved from 0.886 to 0.925, corresponding to an improvement from ‘good’ to ‘excellent’.

Conclusion

Our QReport referencing single-subject results to normative data alongside visual assessment improved sensitivity, accuracy, and interrater agreement for detecting volume loss. The QReport was most effective in the consultants, suggesting that experience is needed to fully benefit from the additional information provided by quantitative analyses.

Key Points

• The use of quantitative report alongside routine visual MRI assessment improves sensitivity and accuracy for detecting volume loss and AD vs visual assessment alone.

• Consultant neuroradiologists’ assessment accuracy and agreement (kappa scores) significantly improved with the use of quantitative atrophy reports.

• First multi-rater radiological clinical evaluation of visual quantitative MRI atrophy report for use as a diagnostic aid in dementia.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00330-020-07455-8.

Understanding brain resilience in superagers: a systematic review

PURPOSE

Superagers are older adults presenting excellent memory performance that may reflect resilience to the conventional pathways of aging. Our contribution aims to shape the evidence body of the known distinctive biomarkers of superagers and their connections with the Brain and Cognitive Reserve and Brain Maintenance concepts.

METHODS

We performed a systematic literature search in PubMed and ScienceDirect with no limit on publication date for studies that evaluated potential biomarkers in superagers classified by validated neuropsychological tests. Methodological quality was assessed using the QUADAS-2 tool.

RESULTS

Twenty-one studies were included, the majority in neuroimaging, followed by histological, genetic, cognition, and a single one on blood plasma analysis. Superagers exhibited specific regions of cortical preservation, rather than global cortical maintenance, standing out the anterior cingulate and hippocampus regions. Both superagers and controls showed similar levels of amyloid deposition. Moreover, the functional oscillation patterns in superagers resembled those described in young adults. Most of the quality assessment for the included studies showed medium risks of bias.

CONCLUSION

This systematic review supports selective cortical preservation in superagers, comprehending regions of the default mode, and salience networks, overlapped by stronger functional connectivity. In this context, the anterior cingulate cortex is highlighted as an imaging and histologic signature of these subjects. Besides, the biomarkers included pointed out that the Brain and Cognitive Reserve and Brain Maintenance concepts are independent and complementary in the superagers' setting.

Clinical evaluation of automated quantitative MRI reports for assessment of hippocampal sclerosis

OBJECTIVES

Hippocampal sclerosis (HS) is a common cause of temporal lobe epilepsy. Neuroradiological practice relies on visual assessment, but quantification of HS imaging biomarkers-hippocampal volume loss and T2 elevation-could improve detection. We tested whether quantitative measures, contextualised with normative data, improve rater accuracy and confidence.

METHODS

Quantitative reports (QReports) were generated for 43 individuals with epilepsy (mean age ± SD 40.0 ± 14.8 years, 22 men; 15 histologically unilateral HS; 5 bilateral; 23 MR-negative). Normative data was generated from 111 healthy individuals (age 40.0 ± 12.8 years, 52 men). Nine raters with different experience (neuroradiologists, trainees, and image analysts) assessed subjects' imaging with and without QReports. Raters assigned imaging normal, right, left, or bilateral HS. Confidence was rated on a 5-point scale.

RESULTS

Correct designation (normal/abnormal) was high and showed further trend-level improvement with QReports, from 87.5 to 92.5% (p = 0.07, effect size d = 0.69). Largest magnitude improvement (84.5 to 93.8%) was for image analysts (d = 0.87). For bilateral HS, QReports significantly improved overall accuracy, from 74.4 to 91.1% (p = 0.042, d = 0.7). Agreement with the correct diagnosis (kappa) tended to increase from 0.74 ('fair') to 0.86 ('excellent') with the report (p = 0.06, d = 0.81). Confidence increased when correctly assessing scans with the QReport (p < 0.001, η²_p = 0.945).

CONCLUSIONS

QReports of HS imaging biomarkers can improve rater accuracy and confidence, particularly in challenging bilateral cases. Improvements were seen across all raters, with large effect sizes, greatest for image analysts. These findings may have positive implications for clinical radiology services and justify further validation in larger groups.

KEY POINTS

• Quantification of imaging biomarkers for hippocampal sclerosis-volume loss and raised T2 signal-could improve clinical radiological detection in challenging cases. • Quantitative reports for individual patients, contextualised with normative reference data, improved diagnostic accuracy and confidence in a group of nine raters, in particular for bilateral HS cases. • We present a pre-use clinical validation of an automated imaging assessment tool to assist clinical radiology reporting of hippocampal sclerosis, which improves detection accuracy.

Lactate as clinical tumour biomarker: Optimization of lactate detection and quantification in MR spectroscopic imaging of glioblastomas

PURPOSE

Increased lactate (Lac) level in brain tumours is in vivo detectable by ¹H MR spectroscopy (MRS) but is frequently overlapped by strong lipid signals, which either leads to a weak quality of the Lac signal or even inhibit its detection. We sought to optimize the separation of Lac from lipid signals in intermediate-echo time MRS acquisitions thus allowing its applicability as clinical biomarker in glioblastomas.

METHODS

Data of 27 patients with glioblastoma multiforme (GBM) were analysed using standard post-processing software as well as in-house modified technique based on the same commercially available software. The patients' Lac concentration values provided by the MRS post-processing technique were converted to realistic concentrations by using an appropriately calibrated phantom. The Cramér-Rao lower bound (%CR) was the principal criterion for estimating the quality of the MRS post-processing results.

RESULTS

Based on the ex vivo calibration, the analysis of the in vivo MR spectroscopy measurements led to an improvement of the %CR(Lac) value from 18 % to 8 %. In a single case, the detection of Lac was achievable only by the modified technique, as Lac signal was contaminated with lipids using the standard analysis. The resulting in vivo Lac values from the modified analysis (median: 4.77 mmol/l, range: 1.5-9.2) were considered as a realistic order of magnitude for the metabolite concentrations, whereas no Lac was identified in the normal appearing white matter. This qualified also Lac mapping as a biomarker for regional heterogeneity in GBM.

CONCLUSIONS

The proposed methodology is a promising first step for more reliable analysis of Lac signal, decontaminating it from lipid peaks in MRS, and may help to establish Lac as a biomarker for brain tumours in clinical routine.

Manual segmentation versus semi-automated segmentation for quantifying vestibular schwannoma volume on MRI

Purpose

Management of vestibular schwannoma (VS) is based on tumour size as observed on T1 MRI scans with contrast agent injection. The current clinical practice is to measure the diameter of the tumour in its largest dimension. It has been shown that volumetric measurement is more accurate and more reliable as a measure of VS size. The reference approach to achieve such volumetry is to manually segment the tumour, which is a time intensive task. We suggest that semi-automated segmentation may be a clinically applicable solution to this problem and that it could replace linear measurements as the clinical standard.

Methods

Using high-quality software available for academic purposes, we ran a comparative study of manual versus semi-automated segmentation of VS on MRI with 5 clinicians and scientists. We gathered both quantitative and qualitative data to compare the two approaches; including segmentation time, segmentation effort and segmentation accuracy.

Results

We found that the selected semi-automated segmentation approach is significantly faster (167 s vs 479 s, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$p<0.001$$\end{document}p<0.001), less temporally and physically demanding and has approximately equal performance when compared with manual segmentation, with some improvements in accuracy. There were some limitations, including algorithmic unpredictability and error, which produced more frustration and increased mental effort in comparison with manual segmentation.

Conclusion

We suggest that semi-automated segmentation could be applied clinically for volumetric measurement of VS on MRI. In future, the generic software could be refined for use specifically for VS segmentation, thereby improving accuracy.

Electronic supplementary material

The online version of this article (10.1007/s11548-020-02222-y) contains supplementary material, which is available to authorized users.

Machine learning assisted DSC-MRI radiomics as a tool for glioma classification by grade and mutation status

Background

Combining MRI techniques with machine learning methodology is rapidly gaining attention as a promising method for staging of brain gliomas. This study assesses the diagnostic value of such a framework applied to dynamic susceptibility contrast (DSC)-MRI in classifying treatment-naïve gliomas from a multi-center patients into WHO grades II-IV and across their isocitrate dehydrogenase (IDH) mutation status.

Methods

Three hundred thirty-three patients from 6 tertiary centres, diagnosed histologically and molecularly with primary gliomas (IDH-mutant = 151 or IDH-wildtype = 182) were retrospectively identified. Raw DSC-MRI data was post-processed for normalised leakage-corrected relative cerebral blood volume (rCBV) maps. Shape, intensity distribution (histogram) and rotational invariant Haralick texture features over the tumour mask were extracted. Differences in extracted features across glioma grades and mutation status were tested using the Wilcoxon two-sample test. A random-forest algorithm was employed (2-fold cross-validation, 250 repeats) to predict grades or mutation status using the extracted features.

Results

Shape, distribution and texture features showed significant differences across mutation status. WHO grade II-III differentiation was mostly driven by shape features while texture and intensity feature were more relevant for the III-IV separation. Increased number of features became significant when differentiating grades further apart from one another. Gliomas were correctly stratified by mutation status in 71% and by grade in 53% of the cases (87% of the gliomas grades predicted with distance less than 1).

Conclusions

Despite large heterogeneity in the multi-center dataset, machine learning assisted DSC-MRI radiomics hold potential to address the inherent variability and presents a promising approach for non-invasive glioma molecular subtyping and grading.

Artificial Intelligence-Based Clinical Decision Support Systems Using Advanced Medical Imaging and Radiomics

Artificial intelligence (AI) is poised to make a veritable impact in medicine. Clinical decision support (CDS) is an important area where AI can augment the clinician's capability to collect, understand and make inferences on an overwhelming volume of patient data to reach the optimal clinical decision. Advancements in medical image analysis, such as Radiomics, and data computation, such as machine learning, have expanded our understanding of disease processes and their management. In this article, we review the most relevant concepts of AI as applicable to advanced imaging-based clinical decision support systems.

Current Landscape of Imaging and the Potential Role for Artificial Intelligence in the Management of COVID-19

The clinical management of COVID-19 is challenging. Medical imaging plays a critical role in the early detection, clinical monitoring and outcomes assessment of this disease. Chest x-ray radiography and computed tomography) are the standard imaging modalities used for the structural assessment of the disease status, while functional imaging (namely, positron emission tomography) has had limited application. Artificial intelligence can enhance the predictive power and utilization of these imaging approaches and new approaches focusing on detection, stratification and prognostication are showing encouraging results. We review the current landscape of these imaging modalities and artificial intelligence approaches as applied in COVID-19 management.

Performance of machine learning-augmented analysis of radiomics for the head and neck cancer histopathological diagnosis: A systematic review and meta-analysis

e18526

Background: The imaging criteria used for head and neck cancers (HNC) staging are mostly anatomical with basic quantitative measures, such as size, and admittedly radiologists’ reading of images is dependent on their expertise level. Radiomics, a term referring to extracting and investigating higher dimensional data from images, has been suggested to address these shortcomings. Assisted by machine learning (ML), highly efficient prediction models could revolutionise our diagnostic practices. Our goal was to study the role of ML in the histopathological diagnosis of HNC based on radiomics. Methods: A systematic review and meta-analysis was conducted using electronic databases (PubMed, Scopus, EMBASE, Google Scholar) and including MRI, PET, and CT studies in patients with HNC. Our study was aimed only at diagnosis utilising radiomics and artificial intelligence (ML). A PRISMA diagram retracing the steps of this search process was completed. QUADAS-2 and EQUATOR checklists were completed. A weighted mean, a mean and a median of the performance indicators were recorded. Results: 7 studies were found eligible for meta-analysis. Patient sample sizes ranged between 2-107 patients (median: 18). CT was the most common modality used (4/7 studies). All but one studies were retrospective. Support vector machine and random forest techniques were the main ML techniques used but how the model was built was rarely described. Furthermore, studies did not make clear the exact number of patients in the testing set. Other issues included the reporting of the final model performance with few studies reporting confidence intervals and 2 studies not reporting the exact performance metrics. The accuracy values for the testing set ranged from 58% -94.1%. The meta-analysis showed an overall weighted-mean accuracy of 78.53%, a mean of 82.9% and a median of 84.4%. The weighted mean of the sensitivity was 76.5%, the mean was 83.3%, and for specificity was 83.9% and 88.5%., respectively. The AUC was 0.8. The neuroradiologists’ overall accuracy was 50.4% if weighted, and 54.5% if not, and the corresponding accuracy of the ML classifiers were 78.4% and 79.6%. The ML scored an accuracy of 20% higher than the radiologists. Conclusions: The results are overall encouraging, keeping in perspective the possible calculation biases and small number of studies. There is need for better documentation and standardisation of the applied ML models, which show initially superior performance compared to radiologists.

Amide proton transfer MRI can accurately stratify gliomas according to their IDH mutation and 1p/19q co-deletion status

2561

Background: Amide proton transfer (APT) MRI provides sensitive metrics at the amides and amines offsets from the water resonance and has been shown in small cohorts to differentiate low from high grade gliomas with better diagnostic performance than diffusion- and perfusion-weighted MRI. The purpose of our study was to assess APT-MRI performance to stratify gliomas according to their IDH mutation and 1p/19q status. Methods: Forty-five patients with primary gliomas and diffuse astrocytomas (26 WHO grade II, 11 WHO grade III, 8 WHO grade IV) underwent prospectively multi-parametric MRI with APT imaging at 3T scanner. The molecular classification identified 9 patients with IDH-wildtype, 1p/19q retained and 36 with IDH-mutant (22 had 1p/19q-retained, 14 had 1p/19q-codeleted). Tumour segmentations were manually created and the masks were superimposed on the calculated magnetisation transfer ratio asymmetry (MTRasym) spectra and proton transfer ratio APT maps. Individual and group analysis was conducted to analyse the statistical differences between quantitative imaging parameters for the IDH mutation and 1p/19q codeletion statuses. Results: The MTRasym spectra showed a clear difference between IDH-wildtype and IDH-mutant gliomas, with the IDH-mutant gliomas presenting a stronger contribution in the amines (p < 0.001). In IDH-mutant 1p/19q-retained and IDH-mutant 1p/19q-codeleted, the MTRasym spectra showed similarities in shape with higher intensity (approx. 60%) for the IDH-mutant 1p/19q-retained gliomas over the entire spectrum indicating an increased content in amines and amides in IDH-mutant 1p/19q-retained (p < 0.01). Notably, the latter entities showed higher amides levels than the IDH-wildtype gliomas (p < 0.03). Conclusions: APT-MRI shows a remarkable potential to disentangle the protein metabolism in gliomas, to link metabolic patterns to the IDH and 1p/19q status and hence provide robust surrogate biomarkers for non-invasive histomolecular classification with potential use as treatment monitoring tools.

Combined structural and perfusion MRI enhanced by machine learning may outperform standalone modalities and radiological expertise in high-grade glioma surveillance: A proof-of-concept study

e14528

Background: Treatment monitoring in patients with High-Grade Glioma (HGG) and identification of disease progression, remains a major challenge in clinical neurooncology. We aimed to develop a support vector machine (SVM) classifier utilising combined longitudinal conventional and Dynamic Susceptibility Contrast (DSC) perfusion MRI to classify between Stable Disease (SD), Pseudoprogression (PsP) and Progressive Disease (PD) in glioma patients under surveillance. Methods: Conventional (269) and perfusion (62) MRI studies of HGG patients acquired between 2012 and 2018 were prospectively analysed. Study participants were separated into two groups: Group I with a single DSC time point (64 participants) and Group II with multiple DSC time points (19 participants). The SVM classifier was trained using all available MRI for each group. Classification accuracy was assessed for the use of features extracted from different feature dataset and time point combinations and compared to the experienced radiologists’ predictions. Results: The study included 64 participants (mean age: 48.5 ± 12.8 yrs [standard deviation], 24 female). SVM classification based on combined perfusion and structural features outperformed standalone datasets across all groups. For the clinically relevant classification step (SD/PSP vs PD), both feature combination as well as the addition of multiple DSC time points, improved classification performance (lowest median error rate: 0.016). The SVM algorithm outperformed radiologists in predicting lesion destiny in both groups. Optimal performance was observed in Group II, in which SVM sensitivity/specificity/accuracy was 100/91.67/94.7% for analysis based on the first time point and 85.71/100/ 94.7% based on multiple time points compared to 60/78/68% and 70/90/84.2% for the respective radiologist classifications. In Group I, the SVM also exceeded radiologist predictions, albeit by a smaller margin and resulted in sensitivity/specificity of 86.49/75.00/81.53% (SVM) compared to 75.7/68.9/73.84% (radiologists). Conclusions: Our results indicate that the addition of multiple longitudinal perfusion time points as well as the combination of structural and perfusion features significantly enhance classification outcome in treatment monitoring of HGGs and machine-learning-assisted diagnosis has potentially superior accuracy than the radiologist's visual evaluation and expertise.