AI-Based Measurement of Lumbar Spinal Stenosis on MRI: External Evaluation of a Fully Automated Model.

OBJECTIVES: The aim of this study was to clinically validate a fully automated AI model for magnetic resonance imaging (MRI)-based quantifications of lumbar spinal canal stenosis. MATERIALS AND METHODS: This retrospective study included lumbar spine MRI of 100 consecutive clinical patients (56 ± 17 years; 43 females, 57 males) performed on clinical 1.5 (51 examinations) and 3 T MRI scanners (49 examinations) with heterogeneous clinical imaging protocols. The AI model performed segmentations of the thecal sac on axial T2-weighted sequences. Based on these segmentations, the anteroposterior (AP) and mediolateral (ML) distance, and the area of the thecal sac were measured in a fully automated manner. For comparison, 2 fellowship-trained musculoskeletal radiologists performed the same segmentations and measurements independently. Statistics included 1-sample t tests, the intraclass correlation coefficient (ICC), Bland-Altman plots, and Dice coefficients. A P value of <0.05 was considered statistically significant. RESULTS: The average measurements of the AI model, reader 1, and reader 2 were 194 ± 72 mm2, 181 ± 71 mm2, and 179 ± 70 mm2 for thecal sac area, 13 ± 3.3 mm, 12.6 ± 3.3 mm, and 12.6 ± 3.2 mm for AP distance, and 19.5 ± 3.9 mm, 20 ± 4.3 mm, and 19.4 ± 4 mm for ML distance, respectively. Significant differences existed for all pairwise comparisons, besides reader 1 versus AI model for the ML distance and reader 1 versus reader 2 for the AP distance (P = 0.1 and P = 0.21, respectively). The pairwise mean absolute errors among reader 1, reader 2, and the AI model ranged from 0.59 mm and 0.75 mm for the AP distance, from 1.16 mm to 1.37 mm for the ML distance, and from 7.9 mm2 to 15.54 mm2 for the thecal sac area. Pairwise ICCs among reader 1, reader 2, and the AI model ranged from 0.91 and 0.94 for the AP distance and from 0.86 to 0.9 for the ML distance without significant differences. For the thecal sac area, the pairwise ICC between both readers and the AI model of 0.97 each was slightly, but significantly lower than the ICC between reader 1 and reader 2 of 0.99. Similarly, the Dice coefficient and Hausdorff distance between both readers and the AI model were significantly lower than the values between reader 1 and reader 2, overall ranging from 0.93 to 0.95 for the Dice coefficients and 1.1 to 1.44 for the Hausdorff distances. CONCLUSIONS: The investigated AI model is reliable for assessing the AP and the ML thecal sac diameters with human level accuracies. The small differences for measurement and segmentation of the thecal sac area between the AI model and the radiologists are likely within a clinically acceptable range.

Direct subthalamic nucleus stimulation influences speech and voice quality in Parkinson's disease patients.

BACKGROUND: DBS of the subthalamic nucleus (STN) considerably ameliorates cardinal motor symptoms in PD. Reported STN-DBS effects on secondary dysarthric (speech) and dysphonic symptoms (voice), as originating from vocal tract motor dysfunctions, are however inconsistent with rather deleterious outcomes based on post-surgical assessments. OBJECTIVE: To parametrically and intra-operatively investigate the effects of deep brain stimulation (DBS) on perceptual and acoustic speech and voice quality in Parkinson's disease (PD) patients. METHODS: We performed an assessment of instantaneous intra-operative speech and voice quality changes in PD patients (n = 38) elicited by direct STN stimulations with variations of central stimulation features (depth, laterality, and intensity), separately for each hemisphere. RESULTS: First, perceptual assessments across several raters revealed that certain speech and voice symptoms could be improved with STN-DBS, but this seems largely restricted to right STN-DBS. Second, computer-based acoustic analyses of speech and voice features revealed that both left and right STN-DBS could improve dysarthric speech symptoms, but only right STN-DBS can considerably improve dysphonic symptoms, with left STN-DBS being restricted to only affect voice intensity features. Third, several subareas according to stimulation depth and laterality could be identified in the motoric STN proper and close to the associative STN with optimal (and partly suboptimal) stimulation outcomes. Fourth, low-to-medium stimulation intensities showed the most optimal and balanced effects compared to high intensities. CONCLUSIONS: STN-DBS can considerably improve both speech and voice quality based on a carefully arranged stimulation regimen along central stimulation features.

Distinct functional levels of human voice processing in the auditory cortex.

Voice signaling is integral to human communication, and a cortical voice area seemed to support the discrimination of voices from other auditory objects. This large cortical voice area in the auditory cortex (AC) was suggested to process voices selectively, but its functional differentiation remained elusive. We used neuroimaging while humans processed voices and nonvoice sounds, and artificial sounds that mimicked certain voice sound features. First and surprisingly, specific auditory cortical voice processing beyond basic acoustic sound analyses is only supported by a very small portion of the originally described voice area in higher-order AC located centrally in superior Te3. Second, besides this core voice processing area, large parts of the remaining voice area in low- and higher-order AC only accessorily process voices and might primarily pick up nonspecific psychoacoustic differences between voices and nonvoices. Third, a specific subfield of low-order AC seems to specifically decode acoustic sound features that are relevant but not exclusive for voice detection. Taken together, the previously defined voice area might have been overestimated since cortical support for human voice processing seems rather restricted. Cortical voice processing also seems to be functionally more diverse and embedded in broader functional principles of the human auditory system.

Performance of a deep convolutional neural network for MRI-based vertebral body measurements and insufficiency fracture detection.

OBJECTIVES: The aim is to validate the performance of a deep convolutional neural network (DCNN) for vertebral body measurements and insufficiency fracture detection on lumbar spine MRI. METHODS: This retrospective analysis included 1000 vertebral bodies in 200 patients (age 75.2 ± 9.8 years) who underwent lumbar spine MRI at multiple institutions. 160/200 patients had ≥ one vertebral body insufficiency fracture, 40/200 had no fracture. The performance of the DCNN and that of two fellowship-trained musculoskeletal radiologists in vertebral body measurements (anterior/posterior height, extent of endplate concavity, vertebral angle) and evaluation for insufficiency fractures were compared. Statistics included (a) interobserver reliability metrics using intraclass correlation coefficient (ICC), kappa statistics, and Bland-Altman analysis, and (b) diagnostic performance metrics (sensitivity, specificity, accuracy). A statistically significant difference was accepted if the 95% confidence intervals did not overlap. RESULTS: The inter-reader agreement between radiologists and the DCNN was excellent for vertebral body measurements, with ICC values of > 0.94 for anterior and posterior vertebral height and vertebral angle, and good to excellent for superior and inferior endplate concavity with ICC values of 0.79-0.85. The performance of the DCNN in fracture detection yielded a sensitivity of 0.941 (0.903-0.968), specificity of 0.969 (0.954-0.980), and accuracy of 0.962 (0.948-0.973). The diagnostic performance of the DCNN was independent of the radiological institution (accuracy 0.964 vs. 0.960), type of MRI scanner (accuracy 0.957 vs. 0.964), and magnetic field strength (accuracy 0.966 vs. 0.957). CONCLUSIONS: A DCNN can achieve high diagnostic performance in vertebral body measurements and insufficiency fracture detection on heterogeneous lumbar spine MRI. KEY POINTS: • A DCNN has the potential for high diagnostic performance in measuring vertebral bodies and detecting insufficiency fractures of the lumbar spine.

Inhibiting Human Aversive Memory by Transcranial Theta-Burst Stimulation to the Primary Sensory Cortex.

BACKGROUND: Predicting adverse events from past experience is fundamental for many biological organisms. However, some individuals suffer from maladaptive memories that impair behavioral control and well-being, e.g., after psychological trauma. Inhibiting the formation and maintenance of such memories would have high clinical relevance. Previous preclinical research has focused on systemically administered pharmacological interventions, which cannot be targeted to specific neural circuits in humans. Here, we investigated the potential of noninvasive neural stimulation on the human sensory cortex in inhibiting aversive memory in a laboratory threat conditioning model. METHODS: We build on an emerging nonhuman literature suggesting that primary sensory cortices may be crucially required for threat memory formation and consolidation. Immediately before conditioning innocuous somatosensory stimuli (conditioned stimuli [CS]) to aversive electric stimulation, healthy human participants received continuous theta-burst transcranial magnetic stimulation (cTBS) to individually localized primary somatosensory cortex in either the CS-contralateral (experimental) or CS-ipsilateral (control) hemisphere. We measured fear-potentiated startle to infer threat memory retention on the next day, as well as skin conductance and pupil size during learning. RESULTS: After overnight consolidation, threat memory was attenuated in the experimental group compared with the control cTBS group. There was no evidence that this differed between simple and complex CS or that CS identification or initial learning were affected by cTBS. CONCLUSIONS: Our results suggest that cTBS to the primary sensory cortex inhibits threat memory, likely by an impact on postlearning consolidation. We propose that noninvasive targeted stimulation of the sensory cortex may provide a new avenue for interfering with aversive memories in humans.

Neurocognitive processing efficiency for discriminating human non-alarm rather than alarm scream calls.

Across many species, scream calls signal the affective significance of events to other agents. Scream calls were often thought to be of generic alarming and fearful nature, to signal potential threats, with instantaneous, involuntary, and accurate recognition by perceivers. However, scream calls are more diverse in their affective signaling nature than being limited to fearfully alarming a threat, and thus the broader sociobiological relevance of various scream types is unclear. Here we used 4 different psychoacoustic, perceptual decision-making, and neuroimaging experiments in humans to demonstrate the existence of at least 6 psychoacoustically distinctive types of scream calls of both alarming and non-alarming nature, rather than there being only screams caused by fear or aggression. Second, based on perceptual and processing sensitivity measures for decision-making during scream recognition, we found that alarm screams (with some exceptions) were overall discriminated the worst, were responded to the slowest, and were associated with a lower perceptual sensitivity for their recognition compared with non-alarm screams. Third, the neural processing of alarm compared with non-alarm screams during an implicit processing task elicited only minimal neural signal and connectivity in perceivers, contrary to the frequent assumption of a threat processing bias of the primate neural system. These findings show that scream calls are more diverse in their signaling and communicative nature in humans than previously assumed, and, in contrast to a commonly observed threat processing bias in perceptual discriminations and neural processes, we found that especially non-alarm screams, and positive screams in particular, seem to have higher efficiency in speeded discriminations and the implicit neural processing of various scream types in humans.

Neural competition between concurrent speech production and other speech perception.

Understanding others' speech while individuals simultaneously produce speech utterances implies neural competition and requires specific mechanisms for a neural resolution given that previous studies proposed opposing signal dynamics for both processes in the auditory cortex (AC). We here used neuroimaging in humans to investigate this neural competition by lateralized stimulations with other speech samples and ipsilateral or contralateral lateralized feedback of actively produced self speech utterances in the form of various speech vowels. In experiment 1, we show, first, that others' speech classifications during active self speech lead to activity in the planum temporale (PTe) when both self and other speech samples were presented together to only the left or right ear. The contralateral PTe also seemed to indifferently respond to single self and other speech samples. Second, specific activity in the left anterior superior temporal cortex (STC) was found during dichotic stimulations (i.e. self and other speech presented to separate ears). Unlike previous studies, this left anterior STC activity supported self speech rather than other speech processing. Furthermore, right mid and anterior STC was more involved in other speech processing. These results signify specific mechanisms for self and other speech processing in the left and right STC beyond a more general speech processing in PTe. Third, other speech recognition in the context of listening to recorded self speech in experiment 2 led to largely symmetric activity in STC and additionally in inferior frontal subregions. The latter was previously reported to be generally relevant for other speech perception and classification, but we found frontal activity only when other speech classification was challenged by recorded but not by active self speech samples. Altogether, unlike formerly established brain networks for uncompetitive other speech perception, active self speech during other speech perception seemingly leads to a neural reordering, functional reassignment, and unusual lateralization of AC and frontal brain activations.

Cortical voice processing is grounded in elementary sound analyses for vocalization relevant sound patterns.

A subregion of the auditory cortex (AC) was proposed to selectively process voices. This selectivity of the temporal voice area (TVA) and its role in processing non-voice sounds however have remained elusive. For a better functional description of the TVA, we investigated its neural responses both to voice and non-voice sounds, and critically also to textural sound patterns (TSPs) that share basic features with natural sounds but that are perceptually very distant from voices. Listening to these TSPs, first, elicited activity in large subregions of the TVA, which was mainly driven by perpetual ratings of TSPs along a voice similarity scale. This similar TVA activity in response to TSPs might partially explain activation patterns typically observed during voice processing. Second, we reconstructed the TVA activity that is usually observed in voice processing with a linear combination of activation patterns from TSPs. An analysis of the reconstruction model weights demonstrated that the TVA similarly processes both natural voice and non-voice sounds as well as TSPs along their acoustic and perceptual features. The predominant factor in reconstructing the TVA pattern by TSPs were the perceptual voice similarity ratings. Third, a multi-voxel pattern analysis confirms that the TSPs contain sufficient sound information to explain TVA activity for voice processing. Altogether, rather than being restricted to higher-order voice processing only, the human "voice area" uses mechanisms to evaluate the perceptual and acoustic quality of non-voice sounds, and responds to the latter with a "voice-like" processing pattern when detecting some rudimentary perceptual similarity with voices.

Neurocognitive dynamics of near-threshold voice signal detection and affective voice evaluation.

Communication and voice signal detection in noisy environments are universal tasks for many species. The fundamental problem of detecting voice signals in noise (VIN) is underinvestigated especially in its temporal dynamic properties. We investigated VIN as a dynamic signal-to-noise ratio (SNR) problem to determine the neurocognitive dynamics of subthreshold evidence accrual and near-threshold voice signal detection. Experiment 1 showed that dynamic VIN, including a varying SNR and subthreshold sensory evidence accrual, is superior to similar conditions with nondynamic SNRs or with acoustically matched sounds. Furthermore, voice signals with affective meaning have a detection advantage during VIN. Experiment 2 demonstrated that VIN is driven by an effective neural integration in an auditory cortical-limbic network at and beyond the near-threshold detection point, which is preceded by activity in subcortical auditory nuclei. This demonstrates the superior recognition advantage of communication signals in dynamic noise contexts, especially when carrying socio-affective meaning.

Primary auditory cortex representation of fear-conditioned musical sounds.

Auditory cortex is required for discriminative fear conditioning beyond the classical amygdala microcircuit, but its precise role is unknown. It has previously been suggested that Heschl's gyrus, which includes primary auditory cortex (A1), but also other auditory areas, encodes threat predictions during presentation of conditioned stimuli (CS) consisting of monophones, or frequency sweeps. The latter resemble natural prosody and contain discriminative spectro-temporal information. Here, we use functional magnetic resonance imaging (fMRI) in humans to address CS encoding in A1 for stimuli that contain only spectral but no temporal discriminative information. Two musical chords (complex) or two monophone tones (simple) were presented in a signaled reinforcement context (reinforced CS+ and nonreinforced CS-), or in a different context without reinforcement (neutral sounds, NS1 and NS2), with an incidental sound detection task. CS/US association encoding was quantified by the increased discriminability of BOLD patterns evoked by CS+/CS-, compared to NS pairs with similar physical stimulus differences and task demands. A1 was defined on a single-participant level and based on individual anatomy. We find that in A1, discriminability of CS+/CS- was higher than for NS1/NS2. This representation of unconditioned stimulus (US) prediction was of comparable magnitude for both types of sounds. We did not observe such encoding outside A1. Different from frequency sweeps investigated previously, musical chords did not share representations of US prediction with monophone sounds. To summarize, our findings suggest decodable representation of US predictions in A1, for various types of CS, including musical chords that contain no temporal discriminative information.

Non-medical prescription opioid users exhibit dysfunctional physiological stress responses to social rejection.

Non-medical prescription opioid use (NMPOU) recently increased dramatically, especially in the U.S. Although chronic opioid use is commonly accompanied by deficits in social functioning and dysregulation of the hypothalamic-pituitary adrenergic (HPA) stress axis, little is known about the impact of NMPOU on psychosocial stress responses. Therefore, we measured physiological responses of the autonomic nervous system and the HPA axis to social rejection using the Cyberball paradigm. We compared 23 individuals with NMPOU, objectively confirmed by hair and urine analyses, with 29 opioid-naïve, healthy controls. As expected, heart rate variability (HRV), an index of parasympathetic activity, increased significantly during exclusion within controls, while in the NMPOU group only a trend in the same direction was found. However, increased HRV was robustly moderated by opioid craving indicating worse emotion regulation to social exclusion specifically in individuals with high opioid craving. Greater levels of the adrenocorticotropic hormone and cortisol responses to social rejection were found in the NMPOU group indicating hyperreactivity of the HPA axis to social exclusion. Self-ratings suggest that opioid users were aware of rejection, but less emotionally affected by exclusion. Furthermore, controls showed greater negative mood after the Cyberball confirming the task's validity. Moreover, NMPOU individuals reported a smaller social network size compared to controls. Present findings suggest that chronic NMPOU is associated with dysfunctional physiological responses to psychosocial stressors such as social rejection. In sum, NMPOU was associated with poorer regulation of the parasympathetic nervous system, especially under opioid craving highlighting its potential importance in relapse prevention.

Stimulus-invariant auditory cortex threat encoding during fear conditioning with simple and complex sounds.

Learning to predict threat depends on amygdala plasticity and does not require auditory cortex (ACX) when threat predictors (conditioned stimuli, CS) are simple sine tones. However, ACX is required in rodents to learn from some naturally occurring CS. Yet, the precise function of ACX, and whether it differs for different CS types, is unknown. Here, we address how ACX encodes threat predictions during human fear conditioning using functional magnetic resonance imaging (fMRI) with multivariate pattern analysis. As in previous rodent work, CS+ and CS- were defined either by direction of frequency modulation (complex) or by frequency of pure tones (simple). In an instructed non-reinforcement context, different sets of simple and complex sounds were always presented without reinforcement (neutral sounds, NS). Threat encoding was measured by separation of fMRI response patterns induced by CS+/CS-, or similar NS1/NS2 pairs. We found that fMRI patterns in Heschl's gyrus encoded threat prediction over and above encoding the physical stimulus features also present in NS, i.e. CS+/CS- could be separated better than NS1/NS2. This was the case both for simple and complex CS. Furthermore, cross-prediction demonstrated that threat representations were similar for simple and complex CS, and thus unlikely to emerge from stimulus-specific top-down, or learning-induced, receptive field plasticity. Searchlight analysis across the entire ACX demonstrated further threat representations in a region including BA22 and BA42. However, in this region, patterns were distinct for simple and complex sounds, and could thus potentially arise from receptive field plasticity. Strikingly, across participants, individual size of Heschl's gyrus predicted strength of fear learning for complex sounds. Overall, our findings suggest that ACX represents threat predictions, and that Heschl's gyrus contains a threat representation that is invariant across physical stimulus categories.

Neurocircuitry of impaired affective sound processing: A clinical disorders perspective.

Decoding affective meaning from sensory information is central to accurate and adaptive behavior in many natural and social contexts. Human vocalizations (speech and non-speech), environmental sounds (e.g. thunder, noise, or animal sounds) and human-produced sounds (e.g. technical sounds or music) can carry a wealth of important aversive, threatening, appealing, or pleasurable affective information that sometimes implicitly influences and guides our behavior. A deficit in processing such affective information is detrimental to adaptive environmental behavior, psychological well-being, and social interactive abilities. These deficits can originate from a diversity of psychiatric and neurological disorders, and are associated with neural dysfunctions across largely distributed brain networks. Recent neuroimaging studies in psychiatric and neurological patients outline the cortical and subcortical neurocircuitry of the complimentary and differential functional roles for affective sound processing. This points to and confirms a recently proposed distributed network rather than a single brain region underlying affective sound processing, and highlights the notion of a multi-functional process that can be differentially impaired in clinical disorders.

Assessing fear learning via conditioned respiratory amplitude responses.

Respiratory physiology is influenced by cognitive processes. It has been suggested that some cognitive states may be inferred from respiration amplitude responses (RAR) after external events. Here, we investigate whether RAR allow assessment of fear memory in cued fear conditioning, an experimental model of aversive learning. To this end, we built on a previously developed psychophysiological model (PsPM) of RAR, which regards interpolated RAR time series as the output of a linear time invariant system. We first establish that average RAR after CS+ and CS- are different. We then develop the response function of fear-conditioned RAR, to be used in our PsPM. This PsPM is inverted to yield estimates of cognitive input into the respiratory system. We analyze five validation experiments involving fear acquisition and retention, delay and trace conditioning, short and medium CS-US intervals, and data acquired with bellows and MRI-compatible pressure chest belts. In all experiments, CS+ and CS- are distinguished by their estimated cognitive inputs, and the sensitivity of this distinction is higher for model-based estimates than for peak scoring of RAR. Comparing these data with skin conductance responses (SCR) and heart period responses (HPR), we find that, on average, RAR performs similar to SCR in distinguishing CS+ and CS-, but is less sensitive than HPR. Overall, our work provides a novel and robust tool to investigate fear memory in humans that may allow wide and straightforward application to diverse experimental contexts.

A pupil size response model to assess fear learning.

During fear conditioning, pupil size responses dissociate between conditioned stimuli that are contingently paired (CS+) with an aversive unconditioned stimulus, and those that are unpaired (CS-). Current approaches to assess fear learning from pupil responses rely on ad hoc specifications. Here, we sought to develop a psychophysiological model (PsPM) in which pupil responses are characterized by response functions within the framework of a linear time-invariant system. This PsPM can be written as a general linear model, which is inverted to yield amplitude estimates of the eliciting process in the central nervous system. We first characterized fear-conditioned pupil size responses based on an experiment with auditory CS. PsPM-based parameter estimates distinguished CS+/CS- better than, or on par with, two commonly used methods (peak scoring, area under the curve). We validated this PsPM in four independent experiments with auditory, visual, and somatosensory CS, as well as short (3.5 s) and medium (6 s) CS/US intervals. Overall, the new PsPM provided equal or decisively better differentiation of CS+/CS- than the two alternative methods and was never decisively worse. We further compared pupil responses with concurrently measured skin conductance and heart period responses. Finally, we used our previously developed luminance-related pupil responses to infer the timing of the likely neural input into the pupillary system. Overall, we establish a new PsPM to assess fear conditioning based on pupil responses. The model has a potential to provide higher statistical sensitivity, can be applied to other conditioning paradigms in humans, and may be easily extended to nonhuman mammals.

Modeling fear-conditioned bradycardia in humans.

Across species, cued fear conditioning is a common experimental paradigm to investigate aversive Pavlovian learning. While fear-conditioned stimuli (CS+) elicit overt behavior in many mammals, this is not the case in humans. Typically, autonomic nervous system activity is used to quantify fear memory in humans, measured by skin conductance responses (SCR). Here, we investigate whether heart period responses (HPR) evoked by the CS, often observed in humans and small mammals, are suitable to complement SCR as an index of fear memory in humans. We analyze four datasets involving delay and trace conditioning, in which heart beats are identified via electrocardiogram or pulse oximetry, to show that fear-conditioned heart rate deceleration (bradycardia) is elicited and robustly distinguishes CS+ from CS-. We then develop a psychophysiological model (PsPM) of fear-conditioned HPR. This PsPM is inverted to yield estimates of autonomic input into the heart. We show that the sensitivity to distinguish CS+ and CS- (predictive validity) is higher for model-based estimates than peak-scoring analysis, and compare this with SCR. Our work provides a novel tool to investigate fear memory in humans that allows direct comparison between species.

Optimising a model-based approach to inferring fear learning from skin conductance responses.

Anticipatory sympathetic arousal is often inferred from skin conductance responses (SCR) and used to quantify fear learning. We have previously provided a model-based approach for this inference, based on a quantitative Psychophysiological Model (PsPM) formulated in non-linear dynamic equations. Here we seek to optimise the inversion of this PsPM. Using two independent fear conditioning datasets, we benchmark predictive validity as the sensitivity to separate the likely presence or absence of the unconditioned stimulus. Predictive validity is optimised across both datasets by (a) using a canonical form of the SCR shape (b) filtering the signal with a bi-directional band-pass filter with cut off frequencies 0.0159 and 5 Hz, (c) simultaneously inverting two trials (d) explicitly modelling skin conductance level changes between trials (e) the choice of the inversion algorithm (f) z-scoring estimates of anticipatory sympathetic arousal from each participant across trials. The original model-based method has higher predictive validity than conventional peak-scoring or an alternative model-based method (Ledalab), and benefits from constraining the model, optimised data preconditioning, and post-processing of ensuing parameters.

A matching pursuit algorithm for inferring tonic sympathetic arousal from spontaneous skin conductance fluctuations.

Tonic sympathetic arousal is often inferred from spontaneous fluctuations in skin conductance, and this relies on assumptions about the shape of these fluctuations and how they are generated. We have previously furnished a psychophysiological model for this relation, and an efficient and reliable inversion method to estimate tonic arousal from given data in the framework of dynamic causal modeling (DCM). Here, we provide a fast alternative inversion method in the form of a matching pursuit (MP) algorithm. Analyzing simulated data, this algorithm approximates the true underlying arousal up to about 10 spontaneous fluctuations per minute of data. For empirical data, we assess predictive validity as the ability to differentiate two known psychological arousal states. Predictive validity is comparable between the methods for three datasets, and also comparable to visual peak scoring. Computation time of the MP algorithm is 2-3 orders of magnitude faster for the MP than the DCM algorithm. In summary, the new MP algorithm provides a fast and reliable alternative to DCM inversion for SF data, in particular when the expected number of fluctuations is lower than 10 per minute, as in typical experimental situations.