A deep learning anthropomorphic model observer for a detection task in PET.

BACKGROUND: Lesion detection is one of the most important clinical tasks in positron emission tomography (PET) for oncology. An anthropomorphic model observer (MO) designed to replicate human observers (HOs) in a detection task is an important tool for assessing task-based image quality. The channelized Hotelling observer (CHO) has been the most popular anthropomorphic MO. Recently, deep learning MOs (DLMOs), mostly based on convolutional neural networks (CNNs), have been investigated for various imaging modalities. However, there have been few studies on DLMOs for PET. PURPOSE: The goal of the study is to investigate whether DLMOs can predict HOs better than conventional MOs such as CHO in a two-alternative forced-choice (2AFC) detection task using PET images with real anatomical variability. METHODS: Two types of DLMOs were implemented: (1) CNN DLMO, and (2) CNN-SwinT DLMO that combines CNN and Swin Transformer (SwinT) encoders. Lesion-absent PET images were reconstructed from clinical data, and lesion-present images were reconstructed with adding simulated lesion sinogram data. Lesion-present and lesion-absent PET image pairs were labeled by eight HOs consisting of four radiologists and four image scientists in a 2AFC detection task. In total, 2268 pairs of lesion-present and lesion-absent images were used for training, 324 pairs for validation, and 324 pairs for test. CNN DLMO, CNN-SwinT DLMO, CHO with internal noise, and non-prewhitening matched filter (NPWMF) were compared in the same train-test paradigm. For comparison, six quantitative metrics including prediction accuracy, mean squared errors (MSEs) and correlation coefficients, which measure how well a MO predicts HOs, were calculated in a 9-fold cross-validation experiment. RESULTS: In terms of the accuracy and MSE metrics, CNN DLMO and CNN-SwinT DLMO showed better performance than CHO and NPWMF, and CNN-SwinT DLMO showed the best performance among the MOs evaluated. CONCLUSIONS: DLMO can predict HOs more accurately than conventional MOs such as CHO in PET lesion detection. Combining SwinT and CNN encoders can improve the DLMO prediction performance compared to using CNN only.

TSANet: multibranch attention deep neural network for classifying tactile selective attention in brain-computer interfaces.

Brain-computer interfaces (BCIs) enable communication between the brain and a computer and electroencephalography (EEG) has been widely used to implement BCIs because of its high temporal resolution and noninvasiveness. Recently, a tactile-based EEG task was introduced to overcome the current limitations of visual-based tasks, such as visual fatigue from sustained attention. However, the classification performance of tactile-based BCIs as control signals is unsatisfactory. Therefore, a novel classification approach is required for this purpose. Here, we propose TSANet, a deep neural network, that uses multibranch convolutional neural networks and a feature-attention mechanism to classify tactile selective attention (TSA) in a tactile-based BCI system. We tested TSANet under three evaluation conditions, namely, within-subject, leave-one-out, and cross-subject. We found that TSANet achieved the highest classification performance compared with conventional deep neural network models under all evaluation conditions. Additionally, we show that TSANet extracts reasonable features for TSA by investigating the weights of spatial filters. Our results demonstrate that TSANet has the potential to be used as an efficient end-to-end learning approach in tactile-based BCIs. Supplementary Information: The online version contains supplementary material available at 10.1007/s13534-023-00309-4.

An Investigation of Stochastic Variance Reduction Algorithms for Relative Difference Penalized 3D PET Image Reconstruction.

Penalised PET image reconstruction algorithms are often accelerated during early iterations with the use of subsets. However, these methods may exhibit limit cycle behaviour at later iterations due to variations between subsets. Desirable converged images can be achieved for a subclass of these algorithms via the implementation of a relaxed step size sequence, but the heuristic selection of parameters will impact the quality of the image sequence and algorithm convergence rates. In this work, we demonstrate the adaption and application of a class of stochastic variance reduction gradient algorithms for PET image reconstruction using the relative difference penalty and numerically compare convergence performance to BSREM. The two investigated algorithms are: SAGA and SVRG. These algorithms require the retention in memory of recently computed subset gradients, which are utilised in subsequent updates. We present several numerical studies based on Monte Carlo simulated data and a patient data set for fully 3D PET acquisitions. The impact of the number of subsets, different preconditioners and step size methods on the convergence of regions of interest values within the reconstructed images is explored. We observe that when using constant preconditioning, SAGA and SVRG demonstrate reduced variations in voxel values between subsequent updates and are less reliant on step size hyper-parameter selection than BSREM reconstructions. Furthermore, SAGA and SVRG can converge significantly faster to the penalised maximum likelihood solution than BSREM, particularly in low count data.

Alpha transcranial alternating current stimulation reduces depressive symptoms in people with schizophrenia and auditory hallucinations: a double-blind, randomized pilot clinical trial.

People with schizophrenia exhibit reduced alpha oscillations and frontotemporal coordination of brain activity. Alpha oscillations are associated with top-down inhibition. Reduced alpha oscillations may fail to censor spurious endogenous activity, leading to auditory hallucinations. Transcranial alternating current stimulation (tACS) at the alpha frequency was shown to enhance alpha oscillations in people with schizophrenia and may thus be a network-based treatment for auditory hallucinations. We conducted a double-blind, randomized, placebo-controlled pilot clinical trial to examine the efficacy of 10-Hz tACS in treating auditory hallucinations in people with schizophrenia. 10-Hz tACS was administered in phase at the dorsolateral prefrontal cortex and the temporoparietal junction with a return current at Cz. Patients were randomized to receive tACS or sham for five consecutive days during the treatment week (40 min/day), followed by a maintenance period, during which participants received weekly tACS (40 min/visit) or sham. tACS treatment reduced general psychopathology (p < 0.05, Cohen's d = -0.690), especially depression (p < 0.005, Cohen's d = -0.806), but not auditory hallucinations. tACS treatment increased alpha power in the target region (p < 0.05), increased the frequency of peak global functional connectivity towards 10 Hz (p < 0.05), and reduced left-right frontal functional connectivity (p < 0.005). Importantly, changes in brain functional connectivity significantly correlated with symptom improvement (p < 0.05). Daily 10 Hz-tACS increased alpha power and altered alpha-band functional connectivity. Successful target engagement reduced depression and other general psychopathology symptoms, but not auditory hallucinations. Considering existing research of 10Hz tACS as a treatment for major depressive disorder, our study demonstrates its transdiagnostic potential for treating depression.

Inflammation and tissue remodeling contribute to fibrogenesis in stricturing Crohn's disease: image processing and analysis study.

BACKGROUND: Inflammation and structural remodeling may contribute to fibrogenesis in Crohn's disease (CD). We quantified the immunoexpression of calretinin, CD34, and calprotectin as a surrogate for mucosal innervation, telocytes (interstitial cells playing a role in networking), and inflammation, respectively, and correlated them with bowel alterations in stricturing CD. METHODS: Primary resection specimens for ileal CD (n = 44, 31 stricturing CD, 13 inflammatory CD) were identified. Left-sided ulcerative colitis and trauma cases were used as controls. Proximal and distal margin and middle (diseased) sections were stained for calretinin, CD34, and calprotectin. Microscopic images were captured from the mucosa (calretinin), submucosa (calprotectin), and myenteric plexus (CD34), and the immunostaining was quantified using image processing and analysis. Bowel thickness at the corresponding sections were measured and correlated with the amount of immunoexpression. RESULTS: A total of 2,037 images were analyzed. In stricturing CD, submucosal alteration/thickening at the stricture site correlated with calprotectin staining and inversely correlated with calretinin staining at the proximal margin. Muscularis propria alteration/thickening at the stricture site correlated with mucosal calretinin staining at the proximal margin. Submucosal alteration/thickening at the proximal margin correlated with calretinin and CD34 staining at the proximal margin and inversely correlated with CD34 staining at the stricture site. Calretinin immunostaining at the distal margin was significantly higher in stricturing CD than the controls. CONCLUSIONS: Inflammation and tissue remodeling appear to contribute to fibrogenesis in stricturing CD. Increased mucosal calretinin immunostaining distal to the diseased segment could be helpful in diagnosing CD in the right clinical context.

A Fast Convergent Ordered-Subsets Algorithm With Subiteration-Dependent Preconditioners for PET Image Reconstruction.

We investigated the imaging performance of a fast convergent ordered-subsets algorithm with subiteration-dependent preconditioners (SDPs) for positron emission tomography (PET) image reconstruction. In particular, we considered the use of SDP with the block sequential regularized expectation maximization (BSREM) approach with the relative difference prior (RDP) regularizer due to its prior clinical adaptation by vendors. Because the RDP regularization promotes smoothness in the reconstructed image, the directions of the gradients in smooth areas more accurately point toward the objective function's minimizer than those in variable areas. Motivated by this observation, two SDPs have been designed to increase iteration step-sizes in the smooth areas and reduce iteration step-sizes in the variable areas relative to a conventional expectation maximization preconditioner. The momentum technique used for convergence acceleration can be viewed as a special case of SDP. We have proved the global convergence of SDP-BSREM algorithms by assuming certain characteristics of the preconditioner. By means of numerical experiments using both simulated and clinical PET data, we have shown that the SDP-BSREM algorithms substantially improve the convergence rate, as compared to conventional BSREM and a vendor's implementation as Q.Clear. Specifically, SDP-BSREM algorithms converge 35%-50% faster in reaching the same objective function value than conventional BSREM and commercial Q.Clear algorithms. Moreover, we showed in phantoms with hot, cold and background regions that the SDP-BSREM algorithms approached the values of a highly converged reference image faster than conventional BSREM and commercial Q.Clear algorithms.

Increasing Trend of Calretinin-Positive Mucosal Innervation from Aganglionic Zone toward Transition Zone in Hirschsprung's Disease.

INTRODUCTION: Diagnosis of very short-segment Hirschsprung's disease (vsHD) by rectal suction biopsy is challenging as its aganglionic zone (AZ) overlaps with physiologic hypoganglionic zone and calretinin-positive mucosal nerves may extend from the transition zone (TZ) into AZ. We studied whether an increasing trend/gradient of calretinin-positive mucosal nerves along the distance from AZ toward TZ aids in diagnosis of HD. MATERIALS AND METHODS: In this study, 46 rectal suction biopsies from non-HD and HD, and 15 pull-through specimens from short-segment HD were evaluated by mucosal calretinin immunostain (CI) and image processing and analysis (IPA) to measure pixel count (PC, the percentage of calretinin stained pixels in the mucosa). Consecutive longitudinal sections of proximal AZ toward distal TZ in HD pull-through specimens were utilized as a vsHD surrogate model. First, we studied variability of mucosal CI in non-HD biopsies along the distance from dentate line. Second, we determined a cutoff point of mucosal CI by IPA that separated non-HD versus HD and applied this cutoff to longitudinal sections from proximal AZ to distal TZ segments in HD pull-through specimens. Third, we studied whether an increasing trend of mucosal CI was universally observed in HD pull-through. RESULTS: Our findings included a significant variability in PC along the biopsy distance in non-HD cases. Positive mucosal CI was found in proximal AZ in 6 (43%) of 14 HD pull-through, among which 1 case lacked submucosal nerve hypertrophy in the proximal AZ. All 14 HD pull-through cases showed an increasing trend/gradient of PC from AZ toward TZ. CONCLUSION: Based on our findings, the presence or absence of mucosal CI positivity and submucosal nerve hypertrophy may not reliably diagnose vsHD in rectal suction biopsy. While we acknowledge that the density of mucosal innervation in variable contexts and anatomical locations is unknown and yet to be explored, our study suggests that an increasing trend of positive mucosal CI from AZ toward TZ by IPA might prove to be a useful tool for the diagnosis of vsHD in the future.

Attenuation Coefficient Estimation for PET/MRI With Bayesian Deep Learning Pseudo-CT and Maximum-Likelihood Estimation of Activity and Attenuation.

A major remaining challenge for magnetic resonance-based attenuation correction methods (MRAC) is their susceptibility to sources of magnetic resonance imaging (MRI) artifacts (e.g., implants and motion) and uncertainties due to the limitations of MRI contrast (e.g., accurate bone delineation and density, and separation of air/bone). We propose using a Bayesian deep convolutional neural network that in addition to generating an initial pseudo-CT from MR data, it also produces uncertainty estimates of the pseudo-CT to quantify the limitations of the MR data. These outputs are combined with the maximum-likelihood estimation of activity and attenuation (MLAA) reconstruction that uses the PET emission data to improve the attenuation maps. With the proposed approach uncertainty estimation and pseudo-CT prior for robust MLAA (UpCT-MLAA), we demonstrate accurate estimation of PET uptake in pelvic lesions and show recovery of metal implants. In patients without implants, UpCT-MLAA had acceptable but slightly higher root-mean-squared-error (RMSE) than Zero-echotime and Dixon Deep pseudo-CT when compared to CTAC. In patients with metal implants, MLAA recovered the metal implant; however, anatomy outside the implant region was obscured by noise and crosstalk artifacts. Attenuation coefficients from the pseudo-CT from Dixon MRI were accurate in normal anatomy; however, the metal implant region was estimated to have attenuation coefficients of air. UpCT-MLAA estimated attenuation coefficients of metal implants alongside accurate anatomic depiction outside of implant regions.

Artificial Intelligence in PET: An Industry Perspective.

Artificial intelligence (AI) has significant potential to positively impact and advance medical imaging, including positron emission tomography (PET) imaging applications. AI has the ability to enhance and optimize all aspects of the PET imaging chain from patient scheduling, patient setup, protocoling, data acquisition, detector signal processing, reconstruction, image processing, and interpretation. AI poses industry-specific challenges which will need to be addressed and overcome to maximize the future potentials of AI in PET. This article provides an overview of these industry-specific challenges for the development, standardization, commercialization, and clinical adoption of AI and explores the potential enhancements to PET imaging brought on by AI in the near future. In particular, the combination of on-demand image reconstruction, AI, and custom-designed data-processing workflows may open new possibilities for innovation which would positively impact the industry and ultimately patients.

Transcranial alternating current stimulation entrains alpha oscillations by preferential phase synchronization of fast-spiking cortical neurons to stimulation waveform.

Computational modeling and human studies suggest that transcranial alternating current stimulation (tACS) modulates alpha oscillations by entrainment. Yet, a direct examination of how tACS interacts with neuronal spiking activity that gives rise to the alpha oscillation in the thalamo-cortical system has been lacking. Here, we demonstrate how tACS entrains endogenous alpha oscillations in head-fixed awake ferrets. We first show that endogenous alpha oscillations in the posterior parietal cortex drive the primary visual cortex and the higher-order visual thalamus. Spike-field coherence is largest for the alpha frequency band, and presumed fast-spiking inhibitory interneurons exhibit strongest coupling to this oscillation. We then apply alpha-tACS that results in a field strength comparable to what is commonly used in humans (<0.5 mV/mm). Both in these ferret experiments and in a computational model of the thalamo-cortical system, tACS entrains alpha oscillations by following the theoretically predicted Arnold tongue. Intriguingly, the fast-spiking inhibitory interneurons exhibit a stronger entrainment response to tACS in both the ferret experiments and the computational model, likely due to their stronger endogenous coupling to the alpha oscillation. Our findings demonstrate the in vivo mechanism of action for the modulation of the alpha oscillation by tACS.

Transcranial Alternating Current Stimulation Reduces Network Hypersynchrony and Persistent Vertigo.

OBJECTIVES: Persistent oscillating vertigo that occurs after entrainment to periodic motion is known as Mal de Débarquement Syndrome (MdDS). Down-modulation of this oscillating vertigo is associated with reduction in long-range resting-state functional connectivity between fronto-parieto-occipital regions. In order to determine the association between this oscillating vertigo and hypersynchrony as measured by the auditory steady-state response (ASSR), we investigated the differences in ASSR between individuals with MdDS and healthy controls as well as the change in ASSR in individuals with MdDS before and after treatment with transcranial alternating current stimulation (tACS). MATERIALS AND METHODS: Individuals with treatment refractory MdDS lasting at least six months received single administrations of fronto-parieto-occipital tACS in an "n-of-1" double-blind randomized design: alpha-frequency in-phase, alpha-frequency antiphase, and gamma frequency antiphase control. The treatment protocol that led to the most acute reduction in symptoms and improved balance was administered for 10-12 sessions given over three days (each session 20-min at 2-4 mA). RESULTS: Twenty-four individuals with MdDS participated (mean age 53.0 ± 11.8 years [range: 22-66 years, median: 57.0 years]; mean duration of illness 38.6 ± 53.4 months [range: 6-240 months, median: 18.0 months]). Individuals with MdDS had elevated ASSR compared to healthy controls at baseline (t11 = 5.95, p < 0.001). There was a significant decrease in the 40 Hz-ASSR response between responders compared to nonresponders to tACS (t-test, t15 = -2.26, p = 0.04). Both in-phase and anti-phase alpha tACS lead to symptom improvement but only antiphase alpha-tACS led to a significant decrease of 40 Hz-ASSR (t-test, t12 = -9.6, p < 0.001). CONCLUSIONS: Our findings suggest that tACS has the potential to reduce network-level hypersynchrony and pathological susceptibility to entrainment by sensory input. To the best of our knowledge, this is the first successful demonstration of desynchronization by noninvasive brain stimulation leading to reduced vertigo. Other disease states associated with pathological functional coupling of neuronal networks may similarly benefit from this novel approach.

Corrigendum: Multi-Modal Integration of EEG-fNIRS for Brain-Computer Interfaces - Current Limitations and Future Directions.

[This corrects the article DOI: 10.3389/fnhum.2017.00503.].

Pinging the brain with transcranial magnetic stimulation reveals cortical reactivity in time and space.

BACKGROUND: Single-pulse transcranial magnetic stimulation (TMS) elicits an evoked electroencephalography (EEG) potential (TMS-evoked potential, TEP), which is interpreted as direct evidence of cortical reactivity to TMS. Thus, combining TMS with EEG can be used to investigate the mechanism underlying brain network engagement in TMS treatment paradigms. However, controversy remains regarding whether TEP is a genuine marker of TMS-induced cortical reactivity or if it is confounded by responses to peripheral somatosensory and auditory inputs. Resolving this controversy is of great significance for the field and will validate TMS as a tool to probe networks of interest in cognitive and clinical neuroscience. OBJECTIVE: Here, we delineated the cortical origin of TEP by spatially and temporally localizing successive TEP components, and modulating them with transcranial direct current stimulation (tDCS) to investigate cortical reactivity elicited by single-pulse TMS and its causal relationship with cortical excitability. METHODS: We recruited 18 healthy participants in a double-blind, cross-over, sham-controlled design. We collected motor-evoked potentials (MEPs) and TEPs elicited by suprathreshold single-pulse TMS targeting the left primary motor cortex (M1). To causally test cortical and corticospinal excitability, we applied tDCS to the left M1. RESULTS: We found that the earliest TEP component (P25) was localized to the left M1. The following TEP components (N45 and P60) were largely localized to the primary somatosensory cortex, which may reflect afferent input by hand-muscle twitches. The later TEP components (N100, P180, and N280) were largely localized to the auditory cortex. As hypothesized, tDCS selectively modulated cortical and corticospinal excitability by modulating the pre-stimulus mu-rhythm oscillatory power. CONCLUSION: Together, our findings provide causal evidence that the early TEP components reflect cortical reactivity to TMS.

Exploring the relationship between geomagnetic activity and human heart rate variability.

PURPOSE: Both geomagnetic and solar activity fluctuate over time and have been proposed to affect human physiology. Heart rate variability (HRV) has substantial health implications regarding the ability to adapt to stressors and has been shown to be altered in many cardiovascular and neurological disorders. Intriguingly, previous work found significant, strong correlations between HRV and geomagnetic/solar activity. The purpose of this study to replicate these findings. We simultaneously measured HRV during a 30-day period. METHODS: We recruited 20 healthy participants and measured their HRV for a 30-day period. We also collected geomagnetic and solar activity during this period for investigating their relationship with the HRV data. RESULTS: In agreement with previous work, we found several significant correlations between short-term HRV and geophysical time-series. However, after correction for autocorrelation, which is inherent in time-series, the only significant results were an increase in very low frequency during higher local geomagnetic activity and a geomagnetic anticipatory decrease in heart rate a day before the higher global geomagnetic activity. Both correlations were very low. The loss of most significant effects after this correction suggests that previous findings may be a result of autocorrelation. A further note of caution is required since our and the previous studies in the field do not correct for multiple comparisons given the exploratory analysis strategy. CONCLUSION: We thus conclude that the effects of geomagnetic and solar activity are (if present) most likely of very small effect size and we question the validity of the previous studies given the methodological concerns we have uncovered with our work.

Progesterone modulates theta oscillations in the frontal-parietal network.

The neuroactive metabolites of the steroid hormones progesterone (P4) and testosterone (T) are GABAergic modulators that influence cognition, yet, the specific effect of P4 and T on brain network activity remains poorly understood. Here, we investigated if a fundamental oscillatory network activity pattern, often related to cognitive control, frontal midline theta (FMT) oscillations, are modulated by steroids hormones, P4 and T. We measured the concentration of P4 and T using salivary enzyme immunoassay and FMT oscillations using high-density electroencephalography (EEG) during eyes-open resting-state in 55 healthy women and men. Electrical brain activity was analyzed using Fourier analysis, aperiodic signal fitting, and beamformer source localization. Steroid hormone concentrations and biological sex were used as predictors for scalp and source-estimated amplitude of theta oscillations. Elevated concentrations of P4 predicted increased amplitude of FMT oscillations across both sexes, and no relationship was found with T. The positive correlation with P4 was specific to the frontal midline electrodes and survived correction for the background aperiodic signal of the brain. Using source localization, FMT oscillations were localized to the frontal-parietal network (FPN). Additionally, theta amplitude within the FPN, but not the default mode network, positively correlated with P4 concentration. Our results suggest that P4 concentration modulates brain activity via upregulation of theta oscillations in the FPN.

Benefits of Using a Spatially-Variant Penalty Strength With Anatomical Priors in PET Reconstruction.

In this study, we explore the use of a spatially-variant penalty strength in penalized image reconstruction using anatomical priors to reduce the dependence of lesion contrast on surrounding activity and lesion location. This work builds on a previous method to make the local perturbation response (LPR) approximately spatially invariant. While the dependence of lesion contrast on the local properties introduced by the anatomical penalty is intentional, the method aims to reduce the influence from surroundings lying along the lines of response (LORs) but not in the penalty neighborhood structure. The method is evaluated using simulated data, assuming that the anatomical information is absent or well-aligned with the corresponding activity images. Since the parallel level sets (PLS) penalty is convex and has shown promising results in the literature, it is chosen as the representative anatomical penalty and incorporated into the previously proposed preconditioned algorithm (L-BFGS-B-PC) for achieving good image quality and fast convergence rate. A 2D disc phantom with a feature at the center and a 3D XCAT thorax phantom with lesions inserted in different slices are used to study how surrounding activity and lesion location affect the visual appearance and quantitative consistency. A bias and noise analysis is also performed with the 2D disc phantom. The consistency of the algorithm convergence rate with respect to different data noise and background levels is also investigated using the XCAT phantom. Finally, an example of reconstruction for a patient dataset with inserted pseudo lesions is used as a demonstration in a clinical context. We show that applying the spatially-variant penalization with PLS can reduce the dependence of the lesion contrast on the surrounding activity and lesion location. It does not affect the bias and noise trade-off curves for matched local resolution. Moreover, when using the proposed penalization, significant improvement in algorithm convergence rate and convergence consistency is observed.

Transition Zone in Total Colonic Aganglionosis and Colorectal Hirschsprung's Disease Shows a Similar Trend of Mucosal Innervation: Image Processing and Analysis Study.

The aganglionic segment of bowel in Hirschsprung's disease (HD) varies in length. It is not clear whether total colonic aganglionosis (TCA) merely represents a long form of HD or a different phenotype of the disease. Animal model studies suggest that TCA may have a longer transition zone (TZ) than conventional colorectal HD. We compared mucosal innervation of TZ in 2 TCA cases and 10 conventional colorectal HD cases by quantifying calretinin-positive mucosal nerve fibers using image processing and analysis. One TCA was associated with esophageal atresia-tracheoesophageal fistula, the other with trisomy 21. The gradients of calretinin-stained pixel count increase per distance from the beginning of TZ (slope) for TCA were not significantly different from those for the conventional HD group. Given this observation, it is speculated that the length of TZ in TCA may fall within the range of and may not be much longer than conventional colorectal HD.