External Validation and Retraining of DeepBleed: The First Open-Source 3D Deep Learning Network for the Segmentation of Spontaneous Intracerebral and Intraventricular Hemorrhage.

BACKGROUND: The objective of this study was to assess the performance of the first publicly available automated 3D segmentation for spontaneous intracerebral hemorrhage (ICH) based on a 3D neural network before and after retraining. METHODS: We performed an independent validation of this model using a multicenter retrospective cohort. Performance metrics were evaluated using the dice score (DSC), sensitivity, and positive predictive values (PPV). We retrained the original model (OM) and assessed the performance via an external validation design. A multivariate linear regression model was used to identify independent variables associated with the model's performance. Agreements in volumetric measurements and segmentation were evaluated using Pearson's correlation coefficients (r) and intraclass correlation coefficients (ICC), respectively. With 1040 patients, the OM had a median DSC, sensitivity, and PPV of 0.84, 0.79, and 0.93, compared to thoseo f 0.83, 0.80, and 0.91 in the retrained model (RM). However, the median DSC for infratentorial ICH was relatively low and improved significantly after retraining, at p < 0.001. ICH volume and location were significantly associated with the DSC, at p < 0.05. The agreement between volumetric measurements (r > 0.90, p > 0.05) and segmentations (ICC ≥ 0.9, p < 0.001) was excellent. CONCLUSION: The model demonstrated good generalization in an external validation cohort. Location-specific variances improved significantly after retraining. External validation and retraining are important steps to consider before applying deep learning models in new clinical settings.

Tablets as an Option for Telemedicine-Evaluation of Diagnostic Performance and Efficiency in Intracranial Arterial Aneurysm Detection.

Purpose: To evaluate a commercially available mobile device for the highly specialized task of detection of intracranial arterial aneurysm in telemedicine. Methods: Six radiologists with three different levels of experience retrospectively interpreted 60 computed tomography (CT) angiographies for the presence of intracranial arterial aneurysm, among them 30 cases with confirmed positive findings. Each radiologist reviewed the angiography datasets twice: once on a dedicated medical-grade workstation and on a commercially available mobile consumer-grade tablet with an interval of 3 months. Diagnostic performance, reading efficiency and subjective scorings including diagnostic confidence were analyzed and compared. Results: Diagnostic performance was comparable on both devices regardless of readers' experience, and no significant differences in sensitivity (66-87.5%) and specificity (79.4-87%) were found. Results obtained with tablets and medical workstations were also comparable in terms of subjective assessment across all reader groups. Conclusions: There was no significant difference between tablet and workstation readings of angiography datasets for the presence of intracranial arterial aneurysm. Sensitivity, specificity, efficiency and subjective scorings were similar with the two devices for all three reader groups. While medical workstations are 10 times more expensive, tablets allow higher mobility especially for radiologists on call.

Interpretable surface-based detection of focal cortical dysplasias: a Multi-centre Epilepsy Lesion Detection study.

One outstanding challenge for machine learning in diagnostic biomedical imaging is algorithm interpretability. A key application is the identification of subtle epileptogenic focal cortical dysplasias (FCDs) from structural MRI. FCDs are difficult to visualize on structural MRI but are often amenable to surgical resection. We aimed to develop an open-source, interpretable, surface-based machine-learning algorithm to automatically identify FCDs on heterogeneous structural MRI data from epilepsy surgery centres worldwide. The Multi-centre Epilepsy Lesion Detection (MELD) Project collated and harmonized a retrospective MRI cohort of 1015 participants, 618 patients with focal FCD-related epilepsy and 397 controls, from 22 epilepsy centres worldwide. We created a neural network for FCD detection based on 33 surface-based features. The network was trained and cross-validated on 50% of the total cohort and tested on the remaining 50% as well as on 2 independent test sites. Multidimensional feature analysis and integrated gradient saliencies were used to interrogate network performance. Our pipeline outputs individual patient reports, which identify the location of predicted lesions, alongside their imaging features and relative saliency to the classifier. On a restricted 'gold-standard' subcohort of seizure-free patients with FCD type IIB who had T1 and fluid-attenuated inversion recovery MRI data, the MELD FCD surface-based algorithm had a sensitivity of 85%. Across the entire withheld test cohort the sensitivity was 59% and specificity was 54%. After including a border zone around lesions, to account for uncertainty around the borders of manually delineated lesion masks, the sensitivity was 67%. This multicentre, multinational study with open access protocols and code has developed a robust and interpretable machine-learning algorithm for automated detection of focal cortical dysplasias, giving physicians greater confidence in the identification of subtle MRI lesions in individuals with epilepsy.

A translational perspective on pathophysiological changes of oscillatory activity in dystonia and parkinsonism.

Intracerebral recordings from movement disorders patients undergoing deep brain stimulation have allowed the identification of pathophysiological patterns in oscillatory activity that correlate with symptom severity. Changes in oscillatory synchrony occur within and across brain areas, matching the classification of movement disorders as network disorders. However, the underlying mechanisms of oscillatory changes are difficult to assess in patients, as experimental interventions are technically limited and ethically problematic. This is why animal models play an important role in neurophysiological research of movement disorders. In this review, we highlight the contributions of translational research to the mechanistic understanding of pathological changes in oscillatory activity, with a focus on parkinsonism and dystonia, while addressing the limitations of current findings and proposing possible future directions.

Lead-OR: A multimodal platform for deep brain stimulation surgery.

Background: Deep brain stimulation (DBS) electrode implant trajectories are stereotactically defined using preoperative neuroimaging. To validate the correct trajectory, microelectrode recordings (MERs) or local field potential recordings can be used to extend neuroanatomical information (defined by MRI) with neurophysiological activity patterns recorded from micro- and macroelectrodes probing the surgical target site. Currently, these two sources of information (imaging vs. electrophysiology) are analyzed separately, while means to fuse both data streams have not been introduced. Methods: Here, we present a tool that integrates resources from stereotactic planning, neuroimaging, MER, and high-resolution atlas data to create a real-time visualization of the implant trajectory. We validate the tool based on a retrospective cohort of DBS patients (N = 52) offline and present single-use cases of the real-time platform. Results: We establish an open-source software tool for multimodal data visualization and analysis during DBS surgery. We show a general correspondence between features derived from neuroimaging and electrophysiological recordings and present examples that demonstrate the functionality of the tool. Conclusions: This novel software platform for multimodal data visualization and analysis bears translational potential to improve accuracy of DBS surgery. The toolbox is made openly available and is extendable to integrate with additional software packages. Funding: Deutsche Forschungsgesellschaft (410169619, 424778381), Deutsches Zentrum für Luft- und Raumfahrt (DynaSti), National Institutes of Health (2R01 MH113929), and Foundation for OCD Research (FFOR).

Deep brain stimulation is an established therapy for patients with Parkinson's disease and an emerging option for other neurological conditions. Electrodes are implanted deep in the brain to stimulate precise brain regions and control abnormal brain activity in those areas. The most common target for Parkinson's disease, for instance, is a structure called the subthalamic nucleus, which sits at the base of the brain, just above the brain stem. To ensure electrodes are placed correctly, surgeons use various sources of information to characterize the patient's brain anatomy and decide on an implant site. These data include brain scans taken before surgery and recordings of brain activity taken during surgery to confirm the intended implant site. Sometimes, the brain activity signals from this last confirmation step may slightly alter surgical plans. It represents one of many challenges for clinical teams: to analyse, assimilate, and communicate data as it is collected during the procedure. Oxenford et al. developed a software pipeline to aggregate the data surgeons use to implant electrodes. The open-source platform, dubbed Lead-OR, visualises imaging data and brain activity recordings (termed electrophysiology data) in real time. The current set-up integrates with commercial tools and existing software for surgical planning. Oxenford et al. tested Lead-OR on data gathered retrospectively from 32 patients with Parkinson's who had electrodes implanted in their subthalamic nucleus. The platform showed good agreement between imaging and electrophysiology data, although there were some unavoidable discrepancies, arising from limitations in the imaging pipeline and from the surgical procedure. Lead-OR was also able to correct for brain shift, which is where the brain moves ever so slightly in the skull. With further validation, this proof-of-concept software could serve as a useful decision-making tool for surgical teams implanting electrodes for deep brain stimulation. In time, if implemented, its use could improve the accuracy of electrode placement, translating into better surgical outcomes for patients. It also has the potential to integrate forthcoming ultra-high-resolution data from current brain mapping projects, and other commercial surgical planning tools.

Levodopa-Induced Dyskinesia Are Mediated by Cortical Gamma Oscillations in Experimental Parkinsonism.

BACKGROUND: Levodopa is the most efficacious drug in the symptomatic therapy of motor symptoms in Parkinson's disease (PD); however, long-term treatment is often complicated by troublesome levodopa-induced dyskinesia (LID). Recent evidence suggests that LID might be related to increased cortical gamma oscillations. OBJECTIVE: The objective of this study was to test the hypothesis that cortical high-gamma network activity relates to LID in the 6-hydroxydopamine model and to identify new biomarkers for adaptive deep brain stimulation (DBS) therapy in PD. METHODS: We recorded and analyzed primary motor cortex (M1) electrocorticogram data and motor behavior in freely moving 6-OHDA lesioned rats before and during a daily treatment with levodopa for 3 weeks. The results were correlated with the abnormal involuntary movement score (AIMS) and used for generalized linear modeling (GLM). RESULTS: Levodopa reverted motor impairment, suppressed beta activity, and, with repeated administration, led to a progressive enhancement of LID. Concurrently, we observed a highly significant stepwise amplitude increase in finely tuned gamma (FTG) activity and gamma centroid frequency. Whereas AIMS and FTG reached their maximum after the 4th injection and remained on a stable plateau thereafter, the centroid frequency of the FTG power continued to increase thereafter. Among the analyzed gamma activity parameters, the fraction of longest gamma bursts showed the strongest correlation with AIMS. Using a GLM, it was possible to accurately predict AIMS from cortical recordings. CONCLUSIONS: FTG activity is tightly linked to LID and should be studied as a biomarker for adaptive DBS. © 2020 The Authors. Movement Disorders published by Wiley Periodicals LLC on behalf of International Parkinson and Movement Disorder Society.

Acute In Vivo Electrophysiological Recordings of Local Field Potentials and Multi-unit Activity from the Hyperdirect Pathway in Anesthetized Rats.

Converging evidence shows that many neuropsychiatric diseases should be understood as disorders of large-scale neuronal networks. To better understand the pathophysiological basis of these diseases, it is necessary to precisely characterize in which way the processing of information is disturbed between the different neuronal parts of the circuit. Using extracellular in vivo electrophysiological recordings, it is possible to accurately delineate neuronal activity within a neuronal network. The application of this method has several advantages over alternative techniques, e.g., functional magnetic resonance imaging and calcium imaging, as it allows a unique temporal and spatial resolution and does not rely on genetically engineered organisms. However, the use of extracellular in vivo recordings is limited since it is an invasive technique that cannot be universally applied. In this article, a simple and easy to use method is presented with which it is possible to simultaneously record extracellular potentials such as local field potentials and multiunit activity at multiple sites of a network. It is detailed how a precise targeting of subcortical nuclei can be achieved using a combination of stereotactic surgery and online analysis of multi-unit recordings. Thus, it is demonstrated, how a complete network such as the hyperdirect cortico-basal ganglia loop can be studied in anesthetized animals in vivo.