No abstract

Well-established in the field of bioelectronic medicine, Spinal Cord Stimulation (SCS) offers an implantable, nonpharmacologic treatment for patients with intractable chronic pain conditions. Chronic pain is a widely heterogenous syndrome with regard to both pathophysiology and the resultant phenotype. Despite advances in our understanding of SCS-mediated antinociception, there still exists limited evidence clarifying the pathways recruited when patterned electric pulses are applied to the epidural space. The rapid clinical implementation of novel SCS methods including burst, high frequency and dorsal root ganglion SCS has provided the clinician with multiple options to treat refractory chronic pain. While compelling evidence for safety and efficacy exists in support of these novel paradigms, our understanding of their mechanisms of action (MOA) dramatically lags behind clinical data. In this review, we reconstruct the available basic science and clinical literature that offers support for mechanisms of both paresthesia spinal cord stimulation (P-SCS) and paresthesia-free spinal cord stimulation (PF-SCS). While P-SCS has been heavily examined since its inception, PF-SCS paradigms have recently been clinically approved with the support of limited preclinical research. Thus, wide knowledge gaps exist between their clinical efficacy and MOA. To close this gap, many rich investigative avenues for both P-SCS and PF-SCS are underway, which will further open the door for paradigm optimization, adjunctive therapies and new indications for SCS. As our understanding of these mechanisms evolves, clinicians will be empowered with the possibility of improving patient care using SCS to selectively target specific pathophysiological processes in chronic pain.

In recent years, significant progress has been made in developing more accurate and efficient machine learning algorithms for segmentation of medical and natural images. In this review article, we highlight the imperative role of machine learning algorithms in enabling efficient and accurate segmentation in the field of medical imaging. We specifically focus on several key studies pertaining to the application of machine learning methods to biomedical image segmentation. We review classical machine learning algorithms such as Markov random fields, k‐means clustering, random forest, etc. Although such classical learning models are often less accurate compared to the deep‐learning techniques, they are often more sample efficient and have a less complex structure. We also review different deep‐learning architectures, such as the artificial neural networks (ANNs), the convolutional neural networks (CNNs), and the recurrent neural networks (RNNs), and present the segmentation results attained by those learning models that were published in the past 3 yr. We highlight the successes and limitations of each machine learning paradigm. In addition, we discuss several challenges related to the training of different machine learning models, and we present some heuristics to address those challenges.

Traumatic brain injury (TBI) is a leading cause of sustained impairment in military and civilian populations. However, mild TBI (mTBI) can be difficult to detect using conventional MRI or CT. Injured brain tissues in mTBI patients generate abnormal slow-waves (1–4 Hz) that can be measured and localized by resting-state magnetoencephalography (MEG). In this study, we develop a voxel-based whole-brain MEG slow-wave imaging approach for detecting abnormality in patients with mTBI on a single-subject basis. A normative database of resting-state MEG source magnitude images (1–4 Hz) from 79 healthy control subjects was established for all brain voxels. The high-resolution MEG source magnitude images were obtained by our recent Fast-VESTAL method. In 84 mTBI patients with persistent post-concussive symptoms (36 from blasts, and 48 from non-blast causes), our method detected abnormalities at the positive detection rates of 84.5%, 86.1%, and 83.3% for the combined (blast-induced plus with non-blast causes), blast, and non-blast mTBI groups, respectively. We found that prefrontal, posterior parietal, inferior temporal, hippocampus, and cerebella areas were particularly vulnerable to head trauma. The result also showed that MEG slow-wave generation in prefrontal areas positively correlated with personality change, trouble concentrating, affective lability, and depression symptoms. Discussion is provided regarding the neuronal mechanisms of MEG slow-wave generation due to deafferentation caused by axonal injury and/or blockages/limitations of cholinergic transmission in TBI. This study provides an effective way for using MEG slow-wave source imaging to localize affected areas and supports MEG as a tool for assisting the diagnosis of mTBI.

IMPORTANCE Predicting infarct size and location is important for decision-making and prognosis in patients with acute stroke. OBJECTIVES To determine whether a deep learning model can predict final infarct lesions using magnetic resonance images (MRIs) acquired at initial presentation (baseline) and to compare the model with current clinical prediction methods. DESIGN, SETTING, AND PARTICIPANTS In this multicenter prognostic study, a specific type of neural network for image segmentation (U-net) was trained, validated, and tested using patients from the Imaging Collaterals in Acute Stroke (iCAS) study from April 14, 2014, to April 15, 2018, and the Diffusion Weighted Imaging Evaluation for Understanding Stroke Evolution Study-2 (DEFUSE-2) study from July 14, 2008, to September 17, 2011 (reported in October 2012). Patients underwent baseline perfusion-weighted and diffusion-weighted imaging and MRI at 3 to 7 days after baseline. Patients were grouped into unknown, minimal, partial, and major reperfusion status based on 24-hour imaging results. Baseline images acquired at presentation were inputs, and the final true infarct lesion at 3 to 7 days was considered the ground truth for the model. The model calculated the probability of infarction for every voxel, which can be thresholded to produce a prediction. Data were analyzed from July 1, 2018, to March 7, 2019. MAIN OUTCOMES AND MEASURES Area under the curve, Dice score coefficient (DSC) (a metric from 0-1 indicating the extent of overlap between the prediction and the ground truth; a DSC of Ն0.5 represents significant overlap), and volume error. Current clinical methods were compared with model performance in subgroups of patients with minimal or major reperfusion. RESULTS Among the 182 patients included in the model (97 women [53.3%]; mean [SD] age, 65 [16] years), the deep learning model achieved a median area under the curve of 0.92 (interquartile range [IQR], 0.87-0.96), DSC of 0.53 (IQR, 0.31-0.68), and volume error of 9 (IQR, −14 to 29) mL. In subgroups with minimal (DSC, 0.58 [IQR, 0.31-0.67] vs 0.55 [IQR, 0.40-0.65]; P = .37) or major (DSC, 0.48 [IQR, 0.29-0.65] vs 0.45 [IQR, 0.15-0.54]; P = .002) reperfusion for which comparison with existing clinical methods was possible, the deep learning model had comparable or better performance. CONCLUSIONS AND RELEVANCE The deep learning model appears to have successfully predicted infarct lesions from baseline imaging without reperfusion information and achieved comparable performance to existing clinical methods. Predicting the subacute infarct lesion may help clinicians prepare for decompression treatment and aid in patient selection for neuroprotective clinical trials.

The present study developed a fast MEG source imaging technique based on Fast Vector-based Spatio-Temporal Analysis using a L1-minimum-norm (Fast-VESTAL) and then used the method to obtain the source amplitude images of resting-state magnetoencephalography (MEG) signals for different frequency bands. The Fast-VESTAL technique consists of two steps. First, L1-minimum-norm MEG source images were obtained for the dominant spatial modes of sensor-waveform covariance matrix. Next, accurate source time-courses with millisecond temporal resolution were obtained using an inverse operator constructed from the spatial source images of Step 1. Using simulations, Fast-VESTAL’s performance of was assessed for its 1) ability to localize multiple correlated sources; 2) ability to faithfully recover source time-courses; 3) robustness to different SNR conditions including SNR with negative dB levels; 4) capability to handle correlated brain noise; and 5) statistical maps of MEG source images. An objective pre-whitening method was also developed and integrated with Fast-VESTAL to remove correlated brain noise. Fast-VESTAL’s performance was then examined in the analysis of human mediannerve MEG responses. The results demonstrated that this method easily distinguished sources in the entire somatosensory network. Next, Fast-VESTAL was applied to obtain the first whole-head MEG source-amplitude images from resting-state signals in 41 healthy control subjects, for all standard frequency bands. Comparisons between resting-state MEG sources images and known neurophysiology were provided. Additionally, in simulations and cases with MEG human responses, the results obtained from using conventional beamformer technique were compared with those from Fast-VESTAL, which highlighted the beamformer’s problems of signal leaking and distorted source time-courses.

Post-traumatic stress disorder (PTSD) is a leading cause of sustained impairment, distress, and poor quality of life in military personnel, veterans, and civilians. Indirect functional neuroimaging studies using PET or fMRI with fear-related stimuli support a PTSD neurocircuitry model that includes amygdala, hippocampus, and ventromedial prefrontal cortex (vmPFC). However, it is not clear if this model can fully account for PTSD abnormalities detected directly by electromagnetic-based source imaging techniques in resting-state. The present study examined resting-state magnetoencephalography (MEG) signals in 25 active-duty service members and veterans with PTSD and 30 healthy volunteers. In contrast to the healthy volunteers, individuals with PTSD showed: 1) hyperactivity from amygdala, hippocampus, posterolateral orbitofrontal cortex (OFC), dorsomedial prefrontal cortex (dmPFC), and insular cortex in high-frequency (i.e., beta, gamma, and high-gamma) bands; 2) hypoactivity from vmPFC, Frontal Pole (FP), and dorsolateral prefrontal cortex (dlPFC) in high-frequency bands; 3) extensive hypoactivity from dlPFC, FP, anterior temporal lobes, precuneous cortex, and sensorimotor cortex in alpha and low-frequency bands; and 4) in individuals with PTSD, MEG activity in the left amygdala and posterolateral OFC correlated positively with PTSD symptom scores, whereas MEG activity in vmPFC and precuneous correlated negatively with symptom score. The present study showed that MEG source imaging technique revealed new abnormalities in the resting-state electromagnetic signals from the PTSD neurocircuitry. Particularly, posterolateral OFC and precuneous may play important roles in the PTSD neurocircuitry model.

Blast mild traumatic brain injury (mTBI) is a leading cause of sustained impairment in military service members and veterans. However, the mechanism of persistent disability is not fully understood. The present study investigated disturbances in brain functioning in mTBI participants using a source-imaging-based approach to analyze functional connectivity (FC) from resting-state magnetoencephalography (rs-MEG). Study participants included 26 active-duty service members or veterans who had blast mTBI with persistent post-concussive symptoms, and 22 healthy control active-duty service members or veterans. The source time courses from regions of interest (ROIs) were used to compute ROI to whole-brain (ROI-global) FC for different frequency bands using two different measures: 1) time-lagged cross-correlation and 2) phase-lock synchrony. Compared with the controls, blast mTBI participants showed increased ROI-global FC in beta, gamma, and low-frequency bands, but not in the alpha band. Sources of abnormally increased FC included the: 1) prefrontal cortex (right ventromedial prefrontal cortex [vmPFC], right rostral anterior cingulate cortex [rACC]), and left ventrolateral and dorsolateral prefrontal cortex; 2) medial temporal lobe (bilateral parahippocampus, hippocampus, and amygdala); and 3) right putamen and cerebellum. In contrast, the blast mTBI group also showed decreased FC of the right frontal pole. Group differences were highly consistent across the two different FC measures. FC of the left ventrolateral prefrontal cortex correlated with executive functioning and processing speed in mTBI participants. Altogether, our findings of increased and decreased regionalpatterns of FC suggest that disturbances in intrinsic brain connectivity may be the result of multiple mechanisms, and are associated with cognitive sequelae of the injury.