Convolutional neural network (CNN)-based approaches have shown promising results in the pansharpening of the satellite images in recent years. However, they still exhibit limitations in producing ...high-quality pansharpening outputs. To that end, we propose a new self-supervised learning framework, where we treat pansharpening as a colorization problem, which brings an entirely novel perspective and solution to the problem compared with the existing methods that base their solution solely on producing a super-resolution version of the multispectral image. Whereas the CNN-based methods provide a reduced-resolution panchromatic image as the input to their model along with the reduced-resolution multispectral images and, hence, learn to increase their resolution together, we instead provide the grayscale transformed multispectral image as the input and train our model to learn the colorization of the grayscale input. We further address the fixed downscale ratio assumption during training, which does not generalize well to the full-resolution scenario. We introduce a noise injection into the training by randomly varying the downsampling ratios. Those two critical changes, along with the addition of adversarial training in the proposed PanColorization generative adversarial network (PanColorGAN) framework, help overcome the spatial-detail loss and blur problems that are observed in CNN-based pansharpening. The proposed approach outperforms the previous CNN-based and traditional methods, as demonstrated in our experiments.
Transcranial direct current stimulation (tDCS) has been used aiming to boost exercise performance and inconsistent findings have been reported. One possible explanation is related to the limitations ...of the so-called "conventional" tDCS, which uses large rectangular electrodes, resulting in a diffuse electric field. A new tDCS technique called high-definition tDCS (HD-tDCS) has been recently developed. HD-tDCS uses small ring electrodes and produces improved focality and greater magnitude of its aftereffects. This study tested whether HD-tDCS would improve exercise performance to a greater extent than conventional tDCS. Twelve endurance athletes (29.4 ± 7.3 years; 60.15 ± 5.09 ml kg
min
) were enrolled in this single-center, randomized, crossover, and sham-controlled trial. To test reliability, participants performed two time to exhaustion (TTE) tests (control conditions) on a cycle simulator with 80% of peak power until volitional exhaustion. Next, they randomly received HD-tDCS (2.4 mA), conventional (2.0 mA), or active sham tDCS (2.0 mA) over the motor cortex for 20-min before performing the TTE test. TTE, heart rate (HR), associative thoughts, peripheral (lower limbs), and whole-body ratings of perceived exertion (RPE) were recorded every minute. Outcome measures were reliable. There was no difference in TTE between HD-tDCS (853.1 ± 288.6 s), simulated conventional (827.8 ± 278.7 s), sham (794.3 ± 271.2 s), or control conditions (TTE1 = 751.1 ± 261.6 s or TTE2 = 770.8 ± 250.6 s) F
= 1.537; P = 0.24; η
p = 0.123. There was no effect on peripheral or whole-body RPE and associative thoughts (P > 0.05). No serious adverse effect was reported. A single session of neither HD-tDCS nor conventional tDCS changed exercise performance and psychophysiological responses in athletes, suggesting that a ceiling effect may exist.
Although single or multiple sessions of transcranial direct current stimulation (tDCS) on the prefrontal cortex over a few weeks improved cognition in patients with Alzheimer's disease (AD), effects ...of repeated tDCS over longer period and underlying neural correlates remain to be elucidated.
This study investigated changes in cognitive performances and regional cerebral metabolic rate for glucose (rCMRglc) after administration of prefrontal tDCS over 6 months in early AD patients.
Patients with early AD were randomized to receive either active (n = 11) or sham tDCS (n = 7) over the dorsolateral prefrontal cortex (DLPFC) at home every day for 6 months (anode F3/cathode F4, 2 mA for 30 min). All patients underwent neuropsychological tests and brain 18F-fluoro-2-deoxyglucose positron emission tomography (FDG-PET) scans at baseline and 6-month follow-up. Changes in cognitive performances and rCMRglc were compared between the two groups.
Compared to sham tDCS, active tDCS improved global cognition measured with Mini-Mental State Examination (p for interaction = 0.02) and language function assessed by Boston Naming Test (p for interaction = 0.04), but not delayed recall performance. In addition, active tDCS prevented decreases in executive function at a marginal level (p for interaction < 0.10). rCMRglc in the left middle/inferior temporal gyrus was preserved in the active group, but decreased in the sham group (p for interaction < 0.001).
Daily tDCS over the DLPFC for 6 months may improve or stabilize cognition and rCMRglc in AD patients, suggesting the therapeutic potential of repeated at-home tDCS.
•We examined effects of 6-month at-home tDCS over the DLPFC in AD patients.•tDCS improved global cognition and language function.•tDCS prevented decreases in glucose metabolism in the middle/inferior temporal gyrus.•Repeated at-home tDCS may be a promising treatment option for AD.
To compare the accuracy and repeatability of emerging machine learning based (i.e. deep) automatic segmentation algorithms with those of well-established semi-automatic (interactive) methods for ...determining liver volume in living liver transplant donors at computerized tomography (CT) imaging.
A total of 12 (6 semi-, 6 full-automatic) methods are evaluated. The semi-automatic segmentation algorithms are based on both traditional iterative models including watershed, fast marching, region growing, active contours and modern techniques including robust statistical segmenter and super-pixels. These methods entail some sort of interaction mechanism such as placing initialization seeds on images or determining a parameter range. The automatic methods are based on deep learning and they include three framework templates (DeepMedic, NiftyNet and U-Net) the first two of which are applied with default parameter sets and the last two involve adapted novel model designs. For 20 living donors (6 training and 12 test datasets), a group of imaging scientists and radiologists created ground truths by performing manual segmentations on contrast material-enhanced CT images. Each segmentation is evaluated using five metrics (i.e. volume overlap and relative volume errors, average/RMS/maximum symmetrical surface distances). The results are mapped to a scoring system and a final grade is calculated by taking their average. Accuracy and repeatability were evaluated using slice by slice comparisons and volumetric analysis. Diversity and complementarity are observed through heatmaps. Majority voting and Simultaneous Truth and Performance Level Estimation (STAPLE) algorithms are utilized to obtain the fusion of the individual results.
The top four methods are determined to be automatic deep models having 79.63, 79.46 and 77.15 and 74.50 scores. Intra-user score is determined as 95.14. Overall, deep automatic segmentation outperformed interactive techniques on all metrics. The mean volume of liver of ground truth is found to be 1409.93 mL ± 271.28 mL, while it is calculated as 1342.21 mL ± 231.24 mL using automatic and 1201.26 mL ± 258.13 mL using interactive methods, showing higher accuracy and less variation on behalf of automatic methods. The qualitative analysis of segmentation results showed significant diversity and complementarity enabling the idea of using ensembles to obtain superior results. The fusion of automatic methods reached 83.87 with majority voting and 86.20 using STAPLE that are only slightly less than fusion of all methods that achieved 86.70 (majority voting) and 88.74 (STAPLE).
Use of the new deep learning based automatic segmentation algorithms substantially increases the accuracy and repeatability for segmentation and volumetric measurements of liver. Fusion of automatic methods based on ensemble approaches exhibits best results almost without any additional time cost due to potential parallel execution of multiple models.
Transcranial direct current stimulation (tDCS) has been used to improve exercise performance, though the protocols used, and results found are mixed.
We aimed to analyze the effect of tDCS on ...improving exercise performance.
A systematic search was performed on the following databases, until December 2017: PubMed/MEDLINE, Embase, Web of Science, SCOPUS, and SportDiscus. Full-text articles that used tDCS for exercise performance improvement in adults were included. We compared the effect of anodal (anode near nominal target) and cathodal (cathode near nominal target) tDCS to a sham/control condition on the outcome measure (performance in isometric, isokinetic or dynamic strength exercise and whole-body exercise).
22 studies (393 participants) were included in the qualitative synthesis and 11 studies (236 participants) in the meta-analysis. The primary motor cortex (M1) was the main nominal tDCS target (n = 16; 72.5%). A significant effect favoring anodal tDCS (a-tDCS) applied before exercise over M1 was found on cycling time to exhaustion (mean difference = 93.41 s; 95%CI = 27.39 s–159.43 s) but this result was strongly influenced by one study (weight = 84%), no effect was found for cathodal tDCS (c-tDCS). No significant effect was found for a-tDCS applied on M1 before or during exercise on isometric muscle strength of the upper or lower limbs. Studies regarding a-tDCS over M1 on isokinetic muscle strength presented mixed results. Individual results of studies using a-tDCS applied over the prefrontal and motor cortices either before or during dynamic muscle strength testing showed positive results, but performing meta-analysis was not possible.
For the protocols tested, a-tDCS but not c-tDCS vs. sham over M1 improved exercise performance in cycling only. However, this result was driven by a single study, which when removed was no longer significant. Further well-controlled studies with larger sample sizes and broader exploration of the tDCS montages and doses are warranted.
•This systematic review and meta-analysis assessed the effect of tDCS on exercise performance enhancement in healthy adults.•Weak evidence for an exercise performance enhancement effect favoring anodal tDCS during whole-body dynamic cyclic exercise.•The result of a-tDCS on improving whole-body dynamic cyclic exercise performance was strongly influenced by a single study.•However, there is no evidence that tDCS improves measures of isometric, isokinetic and dynamic strength.
Computational models of current flow during Electroconvulsive Therapy (ECT) rely on the quasi-static assumption, yet tissue impedance during ECT may be frequency specific and change adaptively to ...local electric field intensity.
We systematically consider the application of the quasi-static pipeline to ECT under conditions where 1) static impedance is measured before ECT and 2) during ECT when dynamic impedance is measured. We propose an update to ECT modeling accounting for frequency-dependent impedance.
The frequency content on an ECT device output is analyzed. The ECT electrode-body impedance under low-current conditions is measured with an impedance analyzer. A framework for ECT modeling under quasi-static conditions based on a single device-specific frequency (e.g., 1 kHz) is proposed.
Impedance using ECT electrodes under low-current is frequency dependent and subject specific, and can be approximated at >100 Hz with a subject-specific lumped parameter circuit model but at <100 Hz increased non-linearly. The ECT device uses a 2 μA 800 Hz test signal and reports a static impedance that approximate 1 kHz impedance. Combined with prior evidence suggesting that conductivity does not vary significantly across ECT output frequencies at high-currents (800–900 mA), we update the adaptive pipeline for ECT modeling centered at 1 kHz frequency. Based on individual MRI and adaptive skin properties, models match static impedance (at 2 μA) and dynamic impedance (at 900 mA) of four ECT subjects.
By considering ECT modeling at a single representative frequency, ECT adaptive and non-adaptive modeling can be rationalized under a quasi-static pipeline.
•ECT device outputs and method of calculating static impedance and dynamic impedance are characterized.•ECT outputs contain multiple frequency and tissue impedance varies across these frequencies, as well as with intensity.•We develop an updated to ECT modeling accounting for frequency-dependent impedance.•Considering at a single representative frequency, ECT adaptive and non-adaptive modeling can be rationalized under a quasi-static pipeline.•The pipeline allows computationally-tractable prediction of subject-specific static and dynamic impedance, despite complex tissue impedance.
Background
During transcranial direct current stimulation (tDCS), the amount and distribution of current that reaches the brain depends on individual anatomy. Many progressive neurodegenerative ...diseases are associated with cortical atrophy, but the importance of individual brain atrophy during tDCS in patients with progressive atrophy, including primary progressive aphasia (PPA), remains unclear.
Objective
In the present study, we addressed the question whether brain anatomy in patients with distinct cortical atrophy patterns would impact brain current intensity and distribution during tDCS over the left IFG.
Method
We developed state-of-the-art, gyri-precise models of three subjects, each representing a variant of primary progressive aphasia: non-fluent variant PPA (nfvPPA), semantic variant PPA (svPPA), and logopenic variant PPA (lvPPA). We considered two exemplary montages over the left inferior frontal gyrus (IFG): a conventional pad montage (anode over F7, cathode over the right cheek) and a 4 × 1 high-definition tDCS montage. We further considered whether local anatomical features, specifically distance of the cortex to skull, can directly predict local electric field intensity.
Results
We found that the differences in brain current flow across the three PPA variants fall within the distribution of anatomically typical adults. While clustering of electric fields was often around individual gyri or sulci, the minimal distance from the gyri/sulci to skull was not correlated with electric field intensity.
Conclusion
Limited to the conditions and assumptions considered here, this argues against a specific need to adjust the tDCS montage for these patients any more than might be considered useful in anatomically typical adults. Therefore, local atrophy does not, in isolation, reliably predict local electric field. Rather, our results are consistent with holistic head anatomy influencing brain current flow, with tDCS producing diffuse and individualized brain current flow patterns and HD-tDCS producing targeted brain current flow across individuals.
After the success of deep learning in point cloud segmentation and classification tasks, it has also been adopted as common practice in point cloud registration applications. State-of-the-art point ...cloud registration methods generally deal with this problem as a regression task to find the underlying rotation and translation between two point clouds. However, given two point clouds, the transformation between them could be calculated using only definitive point subsets from each cloud. Furthermore, training time is still a major problem among the current registration networks, whereas using a selective approach to define the informative point subsets can lead to reduced network training times. To that end, we developed ALReg, an active learning procedure to select a limited subset of point clouds to train the network. Each of the point clouds in the training set is divided into superpoints (small pieces of each cloud) and the training process is started with a small amount of them. By actively selecting new superpoints and including them in the training process, only a prescribed amount of data is used, hence the time needed to converge drastically decreases. We used DeepBBS, FMR, and DCP methods as our baselines to prove our proposed ALReg method. We trained DeepBBS and DCP on the ModelNet40 dataset and FMR on the 7Scenes dataset. Using 25% of the training data for ModelNet and 4% for the 7Scenes, better or similar accuracy scores are obtained in less than 20% of their original training times. The trained models are also tested on the 3DMatch dataset and better results are obtained than the original FMR training procedure.
Improvements in electroconvulsive therapy (ECT) outcomes have followed refinement in device electrical output and electrode montage. The physical properties of the ECT stimulus, together with those ...of the patient's head, determine the impedances measured by the device and govern current delivery to the brain and ECT outcomes.
However, the precise relations among physical properties of the stimulus, patient head anatomy, and patient-specific impedance to the passage of current are long-standing questions in ECT research and practice. To this end, we develop a computational framework based on diverse clinical data sets.
We developed anatomical MRI-derived models of transcranial electrical stimulation (tES) that included changes in tissue conductivity due to local electrical current flow. These “adaptive” models simulate ECT both during therapeutic stimulation using high current (∼1 A) and when dynamic impedance is measured, as well as prior to stimulation when low current (∼1 mA) is used to measure static impedance. We modeled two scalp layers: a superficial scalp layer with adaptive conductivity that increases with electric field up to a subject-specific maximum (σSS¯), and a deep scalp layer with a subject-specific fixed conductivity (σDS).
We demonstrated that variation in these scalp parameters may explain clinical data on subject-specific static impedance and dynamic impedance, their imperfect correlation across subjects, their relationships to seizure threshold, and the role of head anatomy. Adaptive tES models demonstrated that current flow changes local tissue conductivity which in turn shapes current delivery to the brain in a manner not accounted for in fixed tissue conductivity models.
Our predictions that variation in individual skin properties, rather than other aspects of anatomy, largely govern the relationship between static impedance, dynamic impedance, and ECT current delivery to the brain, themselves depend on assumptions about tissue properties. Broadly, our novel modeling pipeline opens the door to explore how adaptive-scalp conductivity may impact transcutaneous electrical stimulation (tES).