Masked Targets Research Articles

Masked image modeling-based boundary reconstruction for 3D medical image segmentation

Accurate segmentation of 3D medical images is vital for computer-aided diagnosis. However, the complexity of target morphological variations and a scarcity of labeled data make segmentation more challenging. Furthermore, existing models make it difficult to fully and efficiently integrate global and local information, which hinders structured knowledge acquisition. To overcome these challenges, we introduce the TNT Masking Network (TNT-MNet), a groundbreaking transformer-based 3D model that utilizes a transformer-in-transformer (TNT) encoder. For the first time, we present masked image modeling (MIM) in supervised learning, utilizing target boundary regions as masked prediction targets to enhance structured knowledge acquisition. We execute multiscale random masking on inner and outer tokens in online branch to tackle the challenge of segmenting organs and lesion regions with varying structures at multiple scales and to enhance modeling capabilities. In contrast, the target branch utilizes all tokens to guide the online branch to reconstruct the masked tokens. Our experiments suggest that TNT-MNet’s performance is comparable, or even better, than state-of-the-art models in three medical image datasets (BTCV, LiTS2017, and BraTS2020) and effectively reduces the dependence on labeled data. The code and models are publicly available at https://github.com/changliu-work/TNT_MNet.

Leveraging impression management motives to increase the use of face masks

ABSTRACT Three pilot studies (Ntotal = 832) revealed that people held more positive attitudes toward targets wearing protective face masks. Therefore, we examined whether knowledge of this self-presentational benefit would increase people’s intentions to wear face masks. Participants (N = 997) were randomly assigned to read a passage about the COVID-19 pandemic, the safety benefit of mask-wearing, the self-presentational benefit of mask-wearing, or a combination of the latter two. Although this manipulation failed, findings revealed that preexisting beliefs about masked targets being more likable were positively associated with mask-wearing intentions, particularly among participants less concerned with disease or more politically conservative.

Empathising with masked targets: limited side effects of face masks on empathy for dynamic, context-rich stimuli

ABSTRACT Multiple studies revealed detrimental effects of face masks on communication, including reduced empathic accuracy and enhanced listening effort. Yet, extant research relied on artificial, decontextualised stimuli, which prevented assessing empathy under more ecologically valid conditions. In this preregistered online experiment (N = 272), we used film clips featuring targets reporting autobiographical events to address motivational mechanisms underlying face mask effects on cognitive (empathic accuracy) and emotional facets (emotional congruence, sympathy) of empathy. Surprisingly, targets whose faces were covered by a mask (or a black bar) elicited the same level of empathy motives (affiliation, cognitive effort), and accordingly, the same level of cognitive and emotional empathy compared to targets with uncovered faces. We only found a negative direct effect of face coverings on sympathy. Additional analyses revealed that older (compared to young) adults showed higher empathy, but age did not moderate face mask effects. Our findings speak against strong negative face mask effects on empathy when using dynamic, context-rich stimuli, yet support motivational mechanisms of empathy.

Effects of rhythmic auditory stimulation on vision: Oscillations in performance can be enhanced, but not induced.

A variety of reports suggest that rhythmic auditory stimulation can entrain visual perception, inducing perceptual oscillations as a function of time relative to the auditory rhythm. These effects have, to date, been reported only for stimulation frequencies at and below 3 Hz. Here we investigate the effects of rhythmic auditory stimulation on the detection of masked visual targets when this stimulation occurs at frequencies to which the visual system has been shown to entrain (8-12 Hz). Across four experiments, we found no consistent evidence of poststimulation modulations in performance induced by 8.5, 10.6, or 12.1 Hz auditory stimulation. This absence of a consistent auditory-to-visual effect was paralleled by an absence of unimodal effects (oscillations in auditory performance) following 12.1 Hz auditory stimulation. In a fifth experiment, we found that although auditory stimulation alone did not induce oscillations in visual performance, auditory stimulation did enhance the effects of concurrent visual stimulation. Notably, this enhancement did not require synchronous presentation of stimuli. These observations are consistent with recent reports that passive auditory stimulation beyond 5 Hz induces neither auditory perceptual oscillations nor visual cortical oscillations and suggest limits to the extent of cross-modal entrainment. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Detection of Sea-Surface Small Targets Masked by Range Sidelobes of Large Objects

Many high-resolution maritime radars for sea-surface small target detection transmit pulses with large time-bandwidth-products to increase transmitter power and to use the pulse compression technique for high range resolution. On the sea surface, besides small targets of interest, there exist many uninteresting large objects such as big ships, reefs, and offshore platforms for oil production. Large objects have too strong radar returns and their range sidelobes often mask wide range intervals. In these range intervals, masked small targets are difficult to be detected directly. In this article, a complete scheme is proposed to find small targets masked by range sidelobes of large objects in high-resolution maritime radars. In this scheme, high-precision reconstruction of radar returns of large objects is crucial because tiny reconstruction errors are enough to mask weak returns of small targets of interest. An algorithm using the linear programming and oversampled high-resolution range profile model of a large object is given to obtain high-precision reconstruction of the strong radar returns of a large object. From residual radar returns removed the reconstructed strong returns, a near-optimal coherent detector under sea clutter modeled by the spherically invariant random vector with inverse Gamma texture plus colored Gaussian noise is constructed to find masked small targets. Simulated experiments and real measured radar data are used to verify the effectiveness of the proposed scheme and method and reveal their limitations. The results show that the proposed scheme and method are all effective unless small targets fall into the range mainlobe regions of large objects.

Modeling of intense pulsed ion beam heated masked targets for extreme materials characterization

Intense, pulsed ion beams locally heat materials and deliver dense electronic excitations that can induce material modifications and phase transitions. Material properties can potentially be stabilized by rapid quenching. Pulsed ion beams with pulse lengths of order ns have recently become available for materials processing. Here, we optimize mask geometries for local modification of materials by intense ion pulses. The goal is to rapidly excite targets volumetrically to the point where a phase transition or local lattice reconstruction is induced followed by rapid cooling that stabilizes desired material's properties fast enough before the target is altered or damaged by, e.g., hydrodynamic expansion. By using a mask, the longitudinal dimension can be large compared to the transverse dimension, allowing the possibility of rapid transverse cooling. We performed HYDRA simulations that calculate peak temperatures for a series of excitation conditions and cooling rates of silicon targets with micro-structured masks and compare these to a simple analytical model. The model gives scaling laws that can guide the design of targets over a wide range of pulsed ion beam parameters.

The magical number one-on-square-root-two: The double-target detection deficit in brief visual displays.

How limited representational capacity is divided when multiple items need to be processed simultaneously is a fundamental question in cognitive psychology. The double-target deficit is the finding that, when monitoring multiple locations or information streams for targets, identification of 2 simultaneous targets is substantially worse than is predicted from the cost of divided attention alone. This finding suggests that targets and nontargets are treated differently by the cognitive system. We investigated the double-target deficit in 4 different visual decision tasks using noisy, backwardly masked targets presented for a range of exposure durations to test the theory that the deficit reflects a capacity limitation of visual short-term memory (VSTM). We quantified the deficit using a sample-size model of VSTM and 2 different models of the decision process: a signal detection MAX model and an optimum likelihood ratio model. We found a double-target deficit in all 4 tasks which increased in magnitude for briefer displays, consistent with the capacity limits of VSTM. We explained the exposure dependency using a competitive interaction model in which nontargets compete for access to VSTM at a slower rate than targets. Our findings support 2-stage models of visual processing in which the most target-like stimuli gain priority access into VSTM before the decision process begins. (PsycINFO Database Record

Early visual processing allows for selective behavior, shifts of attention, and conscious visual experience in spite of masking.

Object-substitution masking (OSM) occurs when a briefly displayed target in a search array is surrounded by a mask, which remains onscreen after the target has disappeared. It has been suggested that OSM results from a specific interference with reentrant visual processing, while the initial feedforward processing is left intact. Here, we tested the prediction that the fastest saccadic responses towards a masked target, supposedly triggered before the onset of reentrant processing, are not impaired by OSM. Indeed, saccades faster than 350ms "escaped" the influence of the mask. Notably, participants' judgements of subjective awareness indicated that stimulus processing during this early stage is not entirely devoid of conscious awareness. Furthermore, the N2pc event-related potential component indicated shifts of spatial attention towards the masked targets on trials with correct fast saccades, suggesting that both target detection and spatial attention can be based on the computations accomplished during the initial feedforward sweep.

Ongoing Slow Fluctuations in V1 Impact on Visual Perception.

The human brain’s ongoing activity is characterized by intrinsic networks of coherent fluctuations, measured for example with correlated functional magnetic resonance imaging signals. So far, however, the brain processes underlying this ongoing blood oxygenation level dependent (BOLD) signal orchestration and their direct relevance for human behavior are not sufficiently understood. In this study, we address the question of whether and how ongoing BOLD activity within intrinsic occipital networks impacts on conscious visual perception. To this end, backwardly masked targets were presented in participants’ left visual field only, leaving the ipsi-lateral occipital areas entirely free from direct effects of task throughout the experiment. Signal time courses of ipsi-lateral BOLD fluctuations in visual areas V1 and V2 were then used as proxies for the ongoing contra-lateral BOLD activity within the bilateral networks. Magnitude and phase of these fluctuations were compared in trials with and without conscious visual perception, operationalized by means of subjective confidence ratings. Our results show that ipsi-lateral BOLD magnitudes in V1 were significantly higher at times of peak response when the target was perceived consciously. A significant difference between conscious and non-conscious perception with regard to the pre-target phase of an intrinsic-frequency regime suggests that ongoing V1 fluctuations exert a decisive impact on the access to consciousness already before stimulation. Both effects were absent in V2. These results thus support the notion that ongoing slow BOLD activity within intrinsic networks covering V1 represents localized processes that modulate the degree of readiness for the emergence of visual consciousness.

Reward-associated features capture attention in the absence of awareness: Evidence from object-substitution masking

Reward-associated visual features have been shown to capture visual attention, evidenced in faster and more accurate behavioral performance, as well as in neural responses reflecting lateralized shifts of visual attention to those features. Specifically, the contralateral N2pc event-related-potential (ERP) component that reflects attentional shifting exhibits increased amplitude in response to task-relevant targets containing a reward-associated feature. In the present study, we examined the automaticity of such reward-association effects using object-substitution masking (OSM) in conjunction with MEG measures of visual attentional shifts. In OSM, a visual-search array is presented, with the target item to be detected indicated by a surrounding mask (here, four surrounding squares). Delaying the offset of the target-surrounding four-dot mask relative to the offset of the rest of the target/distracter array disrupts the viewer's awareness of the target (masked condition), whereas simultaneous offsets do not (unmasked condition). Here we manipulated whether the color of the OSM target was or was not of a previously reward-associated color. By tracking reward-associated enhancements of behavior and the N2pc in response to masked targets containing a previously rewarded or unrewarded feature, the automaticity of attentional capture by reward could be probed. We found an enhanced N2pc response to targets containing a previously reward-associated color feature. Moreover, this enhancement of the N2pc by reward did not differ between masking conditions, nor did it differ as a function of the apparent visibility of the target within the masked condition. Overall, these results underscore the automaticity of attentional capture by reward-associated features, and demonstrate the ability of feature-based reward associations to shape attentional capture and allocation outside of perceptual awareness.

Degraded precision of consciously perceived targets in the attentional blink.

A debate regarding the discrete or continuous nature of visual awareness has revolved around the attentional blink (AB). The AB is a transient limitation in the ability to perceive the second (T2) of two masked targets 200-500ms after the attentional processing of the first target (T1). The AB is considered to represent both a central limitation of attention and limitations within earlier visual stages of information processing (Dux & Marois, 2009). Central to this debate is whether the failure to report T2 is due to the discrete failure of information reaching post-perceptual stages of processing, or whether the graded quality of information produces a weak conscious representation. Asplund et al. (2014) applied a mixture-modeling analysis (Zhang & Luck, 2008) to errors in T2 responses to estimate 1) the width of the error distribution to measure the precision of the T2 percept and 2) the probability of successful T2 encoding, assuming an all-or-none representation of T2. They found evidence in favor of discrete failures within central stages of information processing. Here, we tested whether the AB necessarily leads to discrete failures of perception, or whether weak representations can occur once attention is taxed within the same early visual processing channel (Awh et al. 2004), or feature dimension of orientation. Observers were presented with two oriented gratings, which appeared in a rapid serial visual presentation of random noise. Participants performed two task types: report the orientation of both targets, or of T2 alone. Mixture-modeling analysis revealed an AB, defined by the interaction between task and time, in precision but not in the probability of successful encoding. These results suggest that the quality of T2 representations gradually becomes more precise as attentional resources are made available over time, and that discrete or continuous processing depends on the level in which attentional capacity is taxed. Meeting abstract presented at VSS 2015

Retinotopy of visual masking and non-retinotopic perception during masking.

Due to the movements of the observer and those of objects in the environment, retinotopic representations are highly unstable during ecological viewing conditions. The phenomenal stability of our perception suggests that retinotopic representations are transformed into non-retinotopic representations. It remains to show, however, which visual processes operate under retinotopic representations and which ones operate under non-retinotopic representations. Visual masking refers to the reduced visibility of one stimulus, called the target, due to the presence of a second stimulus, called the mask. Masking has been used extensively to study the dynamic aspects of visual perception. Previous studies using Saccadic Stimulus Presentation Paradigm (SSPP) suggested both retinotopic and non-retinotopic bases for visual masking. In order to understand how the visual system deals with retinotopic changes induced by moving targets, we investigated the retinotopy of visual masking and the fate of masked targets under conditions that do not involve eye movements. We have developed a series of experiments based on a radial Ternus-Pikler display. In this paradigm, the perceived Ternus-Pikler motion is used as a non-retinotopic reference frame to pit retinotopic against non-retinotopic visual masking hypothesis. Our results indicate that both metacontrast and structure masking are retinotopic. We also show that, under conditions that allow observers to read-out effectively non-retinotopic feature attribution, the target becomes visible at a destination different from its retinotopic/spatiotopic location. We discuss the implications of our findings within the context of ecological vision and dynamic form perception.

Repetition in visual word identification: Benefits and costs

University students performed lexical tasks with visually presented target words after the presentation of an identical or unrelated prime, at short (80-120 ms) or longer (410-710 ms) prime-target stimulus onset asynchronies (SOAs). Experiment 1 showed perceptual identification benefits in vocal responding at a short SOA that were reduced (accuracy) or reversed (latency) at a longer SOA. Experiment 2 showed a transition from a repetition benefit to a cost over 3 SOAs in a target-masked version of the lexical decision task (LDT; target displayed for only 141 ms). In Experiment 3 the repetition cost was replicated at a 530-ms SOA in the LDT with masked targets, but a repetition benefit was observed in the conventional LDT (target displayed until response). The dependence of repetition costs on target masking is more consistent with biases based on episodic confusions than refractoriness of lexical representations.

Contrast and strength of visual memory and imagery differentially affect visual perception.

Visual short-term memory (VSTM) and visual imagery have been shown to modulate visual perception. However, how the subjective experience of VSTM/imagery and its contrast modulate this process has not been investigated. We addressed this issue by asking participants to detect brief masked targets while they were engaged either in VSTM or visual imagery. Subjective experience of memory/imagery (strength scale), and the visual contrast of the memory/mental image (contrast scale) were assessed on a trial-by-trial basis. For both VSTM and imagery, contrast of the memory/mental image was positively associated with reporting target presence. Consequently, at the sensory level, both VSTM and imagery facilitated visual perception. However, subjective strength of VSTM was positively associated with visual detection whereas the opposite pattern was found for imagery. Thus the relationship between subjective strength of memory/imagery and visual detection are qualitatively different for VSTM and visual imagery, although their impact at the sensory level appears similar. Our results furthermore demonstrate that imagery and VSTM are partly dissociable processes.

Eye Movements and Practice Effects in the Attentional Dwell Time Paradigm

In the attentional dwell time paradigm by Duncan, Ward, and Shapiro (1994), two backward masked targets are presented at different spatial locations and separated by a varying time interval. Results show that report of the second target is severely impaired when the time interval is less than 500 ms which has been taken as a direct measure of attentional dwell time in human vision. However, we show that eye movements may have confounded the estimate of the dwell time and that the measure may not be robust as previously suggested. The latter is supported by evidence suggesting that intensive training strongly attenuates the dwell time because of habituation to the masks. Thus, this article points to eye movements and masking as two potential methodological pitfalls that should be considered when using the attentional dwell time paradigm to investigate the temporal dynamics of attention.

Overlapping activity periods in early visual cortex and posterior intraparietal area in conscious visual shape perception: A TMS study

Parietal cortex is often activated in brain imaging studies on conscious visual processing, but its causal role and timing in conscious and nonconscious perception are poorly understood. We studied the role of posterior intraparietal sulcus (IPS) and early visual areas (V1/V2) in conscious and nonconscious vision by interfering with their functioning with MRI-guided transcranial magnetic stimulation (TMS). The observers made binary forced-choice decisions concerning the shape or color of the metacontrast masked targets and rated the quality of their conscious perception. TMS was applied 30, 60, 90, or 120ms after stimulus-onset. In the shape discrimination task, TMS of V1/V2 impaired conscious perception at 60, 90, and 120ms and nonconscious perception at 90ms. TMS of IPS impaired only conscious shape perception, also around 90ms. Conscious color perception was facilitated or suppressed depending on the strength of the TMS-induced electric field in V1/V2 at 90ms. The results suggest that simultaneous activity in V1/V2 and IPS around 90ms is necessary for visual awareness of shape but not for nonconscious perception. The overlapping activity periods of IPS and V1/V2 may reflect recurrent interaction between parietal cortex and V1 in conscious shape perception.

Is the Diminished Attentional Blink for Salient T2 Stimuli Driven by a Response Bias?

The attentional blink (AB) effect refers to the finding that the second of two quickly presented masked targets (T1 and T2) is less likely to be identified when the temporal lag between the targets is short than when it is long (Raymond, Shapiro, & Arnell, 1992). Interestingly, the AB effect is diminished for salient T2 stimuli. The common explanation is that salient stimuli are processed more efficiently than other stimuli. A contrasting hypothesis is that it is due to a bias to report such stimuli. Recently, Tibboel, Van Bockstaele, and De Houwer (2011) introduced a signal detection method that allowed them to reject the response bias explanation for the effect of arousing words on the AB. Presently, we used the same method to examine the diminished AB for participants’ own names (Experiment 1) and stimuli that form a coherent category (Experiment 2). Both experiments confirmed that the effect of these stimuli on the AB reflects more efficient processing rather than a response bias.

On the source and scope of priming effects of masked stimuli on endogenous shifts of spatial attention

Unconscious stimuli can influence participants' motor behavior as well as more complex mental processes. Previous cue-priming experiments demonstrated that masked cues can modulate endogenous shifts of spatial attention as measured by choice reaction time tasks. Here, we applied a signal detection task with masked luminance targets to determine the source and the scope of effects of masked stimuli. Target-detection performance was modulated by prime-cue congruency, indicating that prime-cue congruency modulates signal enhancement at early levels of target processing. These effects, however, were only found when the prime was perceptually similar to the cue indicting that primes influence early target processing in an indirect way by facilitating cue processing. Together with previous research we conclude that masked stimuli can modulate perceptual and post-central levels of processing. Findings mark a new limit of the effects of unconscious stimuli which seem to have a smaller scope than conscious stimuli.

Distractor Detection and Suppression Have a Beneficial Effect on Attentional Blink

BackgroundAttentional blink (AB) is a phenomenon that describes the difficulty individuals have in reporting the second of two masked targets if the second target (T2) arrives 200–500 ms after the first target (T1). Recent studies explain the AB from cognitive resources limitation to distractors interference. For example, the temporary loss of control (TLC) hypothesis suggests that the AB is conduced by distractors disrupting the input filter for target processing. The inhibition models suggest that the T1+1 distractor triggers a suppression mechanism which could be beneficial for T1 processing but would suppress T2 at short T1–T2 lags. These models consider that the AB is caused by the appearance of distractors. However, in the present study, two methods were taken to help individuals to detect the distractors more effectively. An attenuated AB deficit was found when the distractors could be excluded or suppressed in time. We consider that under an appropriate condition the distractors detection and suppression have a beneficial effect on attentional blink.Methodology/Principal FindingsTwo methods were employed to help individuals to detect the distractors more effectively: enlarging the low-level-physical characteristic difference between targets and distractors (Experiment 1) and restricting the sets of distractors (Experiment 2). Attenuated AB deficits were found as using the above manipulations.Conclusions/SignificanceThe present study found when the distractors are detected or identified quickly, they could be effectively suppressed, in order to reduce the interference from the targets and result in a smaller AB deficit. We suggest that the suppression mechanism for distractors have a beneficial effect on AB.

Target Cueing Provides Support for Target- and Resource-Based Models of the Attentional Blink

The attentional blink (AB) describes a time-based deficit in processing the second of two masked targets. The AB is attenuated if successive targets appear between the first and final target, or if a cueing target is positioned before the final target. Using various speeds of stimulus presentation, the current study employed successive targets and cueing targets to confirm and extend an understanding of target-target cueing in the AB. In Experiment 1, three targets were presented sequentially at rates of 30 msec/item or 90 msec/item. Successive targets presented at 90 msec improved performance compared with non-successive targets. However, accuracy was equivalently high for successive and non-successive targets presented at 30 msec/item, suggesting that–regardless of whether they occurred consecutively–those items fell within the temporally defined attentional window initiated by the first target. Using four different presentation speeds, Experiment 2 confirmed the time-based definition of the AB and the success of target-cueing at 30 msec/item. This experiment additionally revealed that cueing was most effective when resources were not devoted to the cue, thereby implicating capacity limitations in the AB. Across both experiments, a novel order-error measure suggested that errors tend to decrease with an increasing duration between the targets, but also revealed that certain stimulus conditions result in stable order accuracy. Overall, the results are best encapsulated by target-based and resource-sharing theories of the AB, which collectively value the contributions of capacity limitations and optimizing transient attention in time.