Bilateral cyclic high frequency deep brain stimulation (DBS) of the anterior nucleus of the thalamus (ANT) reduces the seizure count in a subset of patients with epilepsy. Detecting stimulation-induced alterations of pathological brain networks may help to unravel the underlying physiological mechanisms related to effective stimulation delivery and optimize target engagement. We acquired 64-channel electroencephalography during ten ANT-DBS cycles (145 ​Hz, 90 ​μs, 3-5 ​V) of 1-min ON followed by 5-min OFF stimulation to detect changes in cortical activity related to seizure reduction. The study included 14 subjects (three responders, four non-responders, and seven healthy controls). Mixed-model ANOVA tests were used to compare differences in cortical activity between subgroups both ON and OFF stimulation, while investigating frequency-specific effects for the seizure onset zones. ANT-DBS had a widespread desynchronization effect on cortical theta and alpha band activity in responders, but not in non-responders. Time domain analysis showed that the stimulation induced reduction in theta-band activity was temporally linked to the stimulation period. Moreover, stimulation induced theta-band desynchronization in the temporal lobe channels correlated significantly with the therapeutic response. Responders to ANT-DBS and healthy-controls had an overall lower level of theta-band activity compared to non-responders. This study demonstrated that temporal lobe channel theta-band desynchronization may be a predictive physiological hallmark of therapeutic response to ANT-DBS and may be used to improve the functional precision of this intervention by verifying implantation sites, calibrating stimulation contacts, and possibly identifying treatment responders prior to implantation.

Background: For the segmentation of medical imaging data, a multitude of precise but very specific algorithms exist. In previous studies, we investigated the possibility of segmenting MRI data to determine cerebrospinal fluid and brain volume using a classical machine learning algorithm. It demonstrated good clinical usability and a very accurate correlation of the volumes to the single area determination in a reproducible axial layer. This study aims to investigate whether these established segmentation algorithms can be transferred to new, more generalizable deep learning algorithms employing an extended transfer learning procedure and whether medically meaningful segmentation is possible.Methods: Ninety-five routinely performed true FISP MRI sequences were retrospectively analyzed in 43 patients with pediatric hydrocephalus. Using a freely available and clinically established segmentation algorithm based on a hidden Markov random field model, four classes of segmentation (brain, cerebrospinal fluid (CSF), background, and tissue) were generated. Fifty-nine randomly selected data sets (10,432 slices) were used as a training data set. Images were augmented for contrast, brightness, and random left/right and X/Y translation. A convolutional neural network (CNN) for semantic image segmentation composed of an encoder and corresponding decoder subnetwork was set up. The network was pre-initialized with layers and weights from a pre-trained VGG 16 model. Following the network was trained with the labeled image data set. A validation data set of 18 scans (3289 slices) was used to monitor the performance as the deep CNN trained. The classification results were tested on 18 randomly allocated labeled data sets (3319 slices) and on a T2-weighted BrainWeb data set with known ground truth.Results: The segmentation of clinical test data provided reliable results (global accuracy 0.90, Dice coefficient 0.86), while the CNN segmentation of data from the BrainWeb data set showed comparable results (global accuracy 0.89, Dice coefficient 0.84). The segmentation of the BrainWeb data set with the classical FAST algorithm produced consistent findings (global accuracy 0.90, Dice coefficient 0.87). Likewise, the area development of brain and CSF in the long-term clinical course of three patients was presented.Conclusion: Using the presented methods, we showed that conventional segmentation algorithms can be transferred to new advances in deep learning with comparable accuracy, generating a large number of training data sets with relatively little effort. A clinically meaningful segmentation possibility was demonstrated.

Background: The clinical evaluation of the upper limb of severely impaired stroke patient is challenging. Sensor-based assessments may allow for an objective evaluation of this patient population. This study investigated the validity of a device-assisted approach in comparison to the clinical outcome that it is supposed to reflect.Methods: In nineteen severely impaired chronic stroke patients, we applied a gravity-compensating, multi-joint arm exoskeleton (Armeo Spring) and compared this sensor-based assessment with the clinical outcome measure Upper Extremity Fugl-Meyer Assessment (UE-FMA) scale. Specifically, we assessed separately and subsequently the range of motion in joint space for four single joints (i.e., wrist, elbow and shoulder flexion/extension (FE), and shoulder internal/external rotation (IER)), and the closing and opening of the hand with a pressure sensor placed in the handle.Results: Within the kinematic parameters, a strong correlation was observed between wrist and elbow FE (r > 0.7, p < 0.003; Bonferroni corrected). The UE-FMA was significantly predicted by a multiple regression model (F (5, 13) = 12.22, p < 0.0005, adj.R² = 0.83). Both shoulder IER and grip pressure added significantly (p < 0.05) to the prediction with the standardized coefficients β of 0.55 and 0.38, respectively.Conclusions: By applying an exoskeleton-based self-contained evaluation of single-joint movements, a clinically valid assessment of the upper limb range of motion in severely impaired stroke patients is feasible. Shoulder IER contributed most relevantly to the prediction of the clinical status. These findings need to be confirmed in a large, independent patient cohort.

Hearing-preserving partial resection of neurofibromatosis type 2 (NF2) associated vestibular schwannomas (VS) is a preferred treatment strategy, particularly for children and adolescents. However, the residual tumors do grow and lead at some point to continued hearing deterioration. An adjuvant bevacizumab treatment may provide an option for slowing down this process. In this retrospective study, we reviewed tumor volume and hearing data of 16 operated VS in nine patients younger than 30 years over a period of 63 to 142 months. All these patients had one or more bevacizumab treatment periods and most of them had a non-treatment period after surgery. Four different patterns of growth were observed for the residual tumors: (1) growth in the non-treatment periods, which slowed down in the treatment periods; (2) growth slowed down in one but not in another on-period; (3) unaffected growth; (4) no or minimal growth regardless of the treatment. Neither radiological regression of tumor volume nor hearing improvement were observed in the treatment periods. Accelerated hearing deterioration was observed in several non-treatment periods, but also in some treatment periods. No straightforward correlation can be drawn between tumor growth and hearing scores. Tumor growth and worsening of hearing between two measurement points were slightly less in the treatment periods; however, the differences were not significant, because variations were large. In conclusion, our comprehensive follow-up on 16 VS in nine NF2 patients did show heterogenous effects of bevacizumab on small residual vestibular schwannomas after surgery both regarding tumor size and hearing preservation. Thus, smaller and slower growing tumor residuals seem to behave differently to bevacizumab than reported for not-operated faster growing VS.