RESUMEN

The scanpath is an important concept in eye tracking. It refers to a person's eye movements over a period of time, commonly represented as a series of alternating fixations and saccades. Machine learning has been increasingly used for the automatic interpretation of scanpaths over the past few years, particularly in research on passive gaze-based interaction, i.e., interfaces that implicitly observe and interpret human eye movements, with the goal of improving the interaction. This literature review investigates research on machine learning applications in scanpath analysis for passive gaze-based interaction between 2012 and 2022, starting from 2,425 publications and focussing on 77 publications. We provide insights on research domains and common learning tasks in passive gaze-based interaction and present common machine learning practices from data collection and preparation to model selection and evaluation. We discuss commonly followed practices and identify gaps and challenges, especially concerning emerging machine learning topics, to guide future research in the field.

RESUMEN

Different tests measure text comprehension, including the cloze gap-filling test, often used for language learning. Different studies hypothesized cognitive strategies in this type of test and their relationship with working memory and performance. However, no study investigated the cloze test, working memory, and possible cognitive strategies, while performing the test. Therefore, this study aimed to identify cognitive visual strategies in the cloze test by applying an unsupervised algorithm and to analyze the relationship between these strategies with working memory and performance in the cloze test. Our sample consisted of 51 university students, the largest sample in studies of cognitive strategies with cloze tests. Participants answered an 11-item cloze test in a computer with eye-tracking, a verbal working memory test, and a visuospatial working memory test. Our analysis of participants' scanpath identified two main strategies: one with fewer toggles between text and word bank and fewer fixations than the other one, indicating the existence of a global strategy. Furthermore, a model predicting the efficiency of participants in the cloze test found that item complexity, using a global strategy, and higher scores of working memory were the most significant predictors. These results confirm the hypothesis of a global strategy being related to successfully achieving higher-order reading processes.

Asunto(s)

RESUMEN

A new algorithm to identify a representative scanpath in a sample is presented and evaluated with eye-tracking data. According to Gestalt theory, each fixation of the scanpath should be on an area of interest (AOI) of the stimuli. As with existing methods, we first identify the AOIs and then extract the fixations of the representative scanpath from the AOIs. In contrast to existing methods, we propose a new concept of short-time AOI and extract the fixations of representative scanpath from the short-time AOIs. Our method outperforms the existing methods on two publicly available datasets. Our method can be applied to arbitrary visual stimuli, including static stimuli without natural segmentation, as well as dynamic stimuli. Our method also provides a solution for issues caused by the selection of scanpath similarity.

Asunto(s)

RESUMEN

OBJECTIVES: We aimed to understand how artificial intelligence (AI) influences dentists by comparing their gaze behavior when using versus not using an AI software to detect primary proximal carious lesions on bitewing radiographs. METHODS: 22 dentists assessed a median of 18 bitewing images resulting in 170 datasets from dentists without AI and 179 datasets from dentists with AI, after excluding data with poor gaze recording quality. We compared time to first fixation, fixation count, average fixation duration, and fixation frequency between both trial groups. Analyses were performed for the entire image and stratified by (1) presence of carious lesions and/or restorations and (2) lesion depth (E1/2: outer/inner enamel; D1-3 outer-inner third of dentin). We also compared the transitional pattern of the dentists' gaze between the trial groups. RESULTS: Median time to first fixation was shorter in all groups of teeth for dentists with AI versus without AI, although p>0.05. Dentists with AI had more fixations (median=68, IQR=31, 116) on teeth with restorations compared to dentists without AI (median=47, IQR=19, 100), p = 0.01. In turn, average fixation duration was longer on teeth with caries for the dentists with AI than those without AI; although p>0.05. The visual search strategy employed by dentists with AI was less systematic with a lower proportion of lateral tooth-wise transitions compared to dentists without AI. CONCLUSIONS: Dentists with AI exhibited more efficient viewing behavior compared to dentists without AI, e.g., lesser time taken to notice caries and/or restorations, more fixations on teeth with restorations, and fixating for shorter durations on teeth without carious lesions and/or restorations. CLINICAL SIGNIFICANCE: Analysis of dentists' gaze patterns while using AI-generated annotations of carious lesions demonstrates how AI influences their data extraction methods for dental images. Such insights can be exploited to improve, and even customize, AI-based diagnostic tools, thus reducing the dentists' extraneous attentional processing and allowing for more thorough examination of other image areas.

Asunto(s)

RESUMEN

Eye movements have been used to examine the cognitive function of pilots and understand how information processing abilities impact performance. Traditional and advanced measures of gaze behaviour effectively reflect changes in cognitive load, situational awareness, and expert-novice differences. However, the extent to which gaze behaviour changes during the early stages of skill development has yet to be addressed. The current study investigated the impact of task difficulty on gaze behaviour in low-time pilots (N=18) while they completed simulated landing scenarios. An increase in task difficulty resulted in longer fixation of the runway, and a reduction in the stationary gaze entropy (gaze dispersion) and gaze transition entropy (sequence complexity). These findings suggest that pilots' gaze became less complex and more focused on fewer areas of interest when task difficulty increased. Additionally, a novel approach to identify and track instances when pilots restrict their attention outside the cockpit (i.e., gaze tunneling) was explored and shown to be sensitive to changes in task difficulty. Altogether, the gaze-related metrics used in the present study provide valuable information for assessing pilots gaze behaviour and help further understand how gaze contributes to better performance in low-time pilots.

RESUMEN

OBJECTIVES: Understanding dentists' gaze patterns on radiographs may allow to unravel sources of their limited accuracy and develop strategies to mitigate them. We conducted an eye tracking experiment to characterize dentists' scanpaths and thus their gaze patterns when assessing bitewing radiographs to detect primary proximal carious lesions. METHODS: 22 dentists assessed a median of nine bitewing images each, resulting in 170 datasets after excluding data with poor quality of gaze recording. Fixation was defined as an area of attentional focus related to visual stimuli. We calculated time to first fixation, fixation count, average fixation duration, and fixation frequency. Analyses were performed for the entire image and stratified by (1) presence of carious lesions and/or restorations and (2) lesion depth (E1/2: outer/inner enamel; D1-3: outer-inner third of dentin). We also examined the transitional nature of the dentists' gaze. RESULTS: Dentists had more fixations on teeth with lesions and/or restorations (median=138 [interquartile range=87, 204]) than teeth without them (32 [15, 66]), p<0.001. Notably, teeth with lesions had longer fixation durations (407 milliseconds [242, 591]) than those with restorations (289 milliseconds [216, 337]), p<0.001. Time to first fixation was longer for teeth with E1 lesions (17,128 milliseconds [8813, 21,540]) than lesions of other depths (p = 0.049). The highest number of fixations were on teeth with D2 lesions (43 [20, 51]) and lowest on teeth with E1 lesions (5 [1, 37]), p<0.001. Generally, a systematic tooth-by-tooth gaze pattern was observed. CONCLUSIONS: As hypothesized, while visually inspecting bitewing radiographic images, dentists employed a heightened focus on certain image features/areas, relevant to the assigned task. Also, they generally examined the entire image in a systematic tooth-by-tooth pattern.

Asunto(s)

RESUMEN

Image quality assessment of 360-degree images is still in its early stages, especially when it comes to solutions that rely on machine learning. There are many challenges to be addressed related to training strategies and model architecture. In this paper, we propose a perceptually weighted multichannel convolutional neural network (CNN) using a weight-sharing strategy for 360-degree IQA (PW-360IQA). Our approach involves extracting visually important viewports based on several visual scan-path predictions, which are then fed to a multichannel CNN using DenseNet-121 as the backbone. In addition, we account for users' exploration behavior and human visual system (HVS) properties by using information regarding visual trajectory and distortion probability maps. The inter-observer variability is integrated by leveraging different visual scan-paths to enrich the training data. PW-360IQA is designed to learn the local quality of each viewport and its contribution to the overall quality. We validate our model on two publicly available datasets, CVIQ and OIQA, and demonstrate that it performs robustly. Furthermore, the adopted strategy considerably decreases the complexity when compared to the state-of-the-art, enabling the model to attain comparable, if not better, results while requiring less computational complexity.

RESUMEN

Eye movements have been examined as an index of attention and comprehension during reading in the literature for over 30 years. Although eye-movement measurements are acknowledged as reliable indicators of readers' comprehension skill, few studies have analyzed eye-movement patterns using network science. In this study, we offer a new approach to analyze eye-movement data. Specifically, we recorded visual scanpaths when participants were reading expository science text, and used these to construct scanpath networks that reflect readers' processing of the text. Results showed that low ability and high ability readers' scanpath networks exhibited distinctive properties, which are reflected in different network metrics including density, centrality, small-worldness, transitivity, and global efficiency. Such patterns provide a new way to show how skilled readers, as compared with less skilled readers, process information more efficiently. Implications of our analyses are discussed in light of current theories of reading comprehension.

Asunto(s)

RESUMEN

Remediation of attentional impairments is an essential component of cognitive rehabilitation after traumatic brain injury (TBI). Evidence from healthy participants has demonstrated attentional improvement following playing an action video game. This exploratory study investigated its application in TBI participants in a multiple baselines single case experimental design (SCED). Saccadic eye movements, recognized as the visible indicators of visual attention, were assessed to evaluate the effectiveness of the game training. Three severe TBI participants were trained in an action game for 10 hours. Saccadic eye movements during a self-paced saccade and an abstract visual search task were investigated during baseline, mid training and post-training. Using Percentage of Non-overlapping Data (PND), analysis showed consistent increase in the rate of the self-paced saccades in participants 1 (PND=80%) and 2 (PND=70%). In abstract search, fixation duration showed a minimally effective decrease for participant 2 (PND= 60%) and a moderately effective reduction in participant 3 (PND= 80%). Search time showed a highly effective reduction in participant 2 (PND = 100%) and moderately effective decrease in participant 3 (PND=70%). Overall, video game training might modify allocation of attention in eye movements. More evidence is required to validate the usefulness of this novel method of the cognitive training.

RESUMEN

Recent studies matching eye gaze patterns with those of others contain research that is heavily reliant on string editing methods borrowed from early work in bioinformatics. Previous studies have shown string editing methods to be susceptible to false negative results when matching mutated genes or unordered regions of interest in scanpaths. Even as new methods have emerged for matching amino acids using novel combinatorial techniques, scanpath matching is still limited by a traditional collinear approach. This approach reduces the ability to discriminate between free viewing scanpaths of two people looking at the same stimulus due to the heavy weight placed on linearity. To overcome this limitation, we here introduce a new method called SoftMatch to compare pairs of scanpaths. SoftMatch diverges from traditional scanpath matching in two different ways: firstly, by preserving locality using fractal curves to reduce dimensionality from 2D Cartesian (x,y) coordinates into 1D (h) Hilbert distances, and secondly by taking a combinatorial approach to fixation matching using discrete Fréchet distance measurements between segments of scanpath fixation sequences. These matching "sequences of fixations over time" are a loose acronym for SoftMatch. Results indicate high degrees of statistical and substantive significance when scoring matches between scanpaths made during free-form viewing of unfamiliar stimuli. Applications of this method can be used to better understand bottom up perceptual processes extending to scanpath outlier detection, expertise analysis, pathological screening, and salience prediction.

Asunto(s)

RESUMEN

Various eye movement abnormalities and impairments in visual information processing have been reported in patients with schizophrenia. Therefore, dysfunction of saccadic eye movements is a potential biological marker for schizophrenia. In the present study, we used a pharmacological model of schizophrenia symptoms in marmosets and compared the eye movement characteristics of marmosets during free-viewing, using an image set identical to those used for human studies. It contains natural and complex images that were randomly presented for 8 s. As a pharmacological model of schizophrenia symptoms, a subanesthetic dose of ketamine was injected intramuscularly for transient and reversible manipulation. Eye movements were recorded and compared under a ketamine condition and a saline condition as a control. The results showed that ketamine affected eye movement characteristics during free-viewing. Saccades amplitude and scanpath length were significantly reduced in the ketamine condition. In addition, the duration of saccades was longer under the ketamine condition than under the saline condition. A similar tendency was observed for the duration of fixations. The number of saccades and fixations tended to decrease in the ketamine condition. The peak saccades velocity also decreased after ketamine injection whereas there was no difference in the main sequence relationship between saccades amplitude and peak velocity. These results suggest that ketamine affected visual exploration but did not affect the oculomotor aspect of saccades in marmosets, consistent with studies in patients with schizophrenia. Therefore, we conclude that the subanesthetic dose of ketamine is a promising pharmacological model of schizophrenia symptoms in common marmosets and can be used in combination with free-viewing paradigms to establish "translatable markers" for schizophrenia in primates.

RESUMEN

OBJECTIVE: Pathophysiology explanations for functional movement disorders often assume a role for emotional hyperarousal. Pupillometry is a validated method for evaluation of emotional arousal by detecting changes in pupil size in response to emotional stimuli. In a case-control study design, we aimed to study objective and subjective emotional arousal using pupillometry and affective ratings. To assess attentional engagement by affective stimuli, we used videooculographic tracking of eye movement patterns (scanpath). METHODS: Twenty-five female patients with functional movement disorders (mean age: 40.9 [SD 12.7] years) and 23 age matched healthy female controls participated in the study. Using infrared high-resolution eye-tracker, both pupil size and eye movement pattern in response to emotionally charged erotic, adventure, threat, victim, and neutral pictures were recorded along with subjective ratings of emotional valence and arousal of the presented pictures. RESULTS: A between-group comparison showed significantly smaller pupil dilation to adventure stimuli compared to neutral stimuli in patients compared to controls (P < 0.004, bootstrap, uncorr., adj. η2 = 0.00). No significant difference in pupillary response to other stimuli and scanpath parameters was found between the groups. Patients rated significantly lower emotional arousal to erotic pictures than controls (P < 0.001, bootstrap, uncorr., adj. η2 = 0.09). CONCLUSION: This study did not find evidence of autonomous or subjective emotional hyperarousal. The mismatch between objective autonomic measures and subjective arousal ratings in patients is of pathophysiological interest and in line with recent findings of impaired interoception in functional movement disorders.

Asunto(s)

RESUMEN

OBJECTIVE: This study was performed to establish a quantitative evaluation and comparison of fixation stability, as measured by an eye tracker, using image-based areas determined by the bivariate contour ellipse area (BCEA), kernel density estimation (KDE), and Scanpath methods. METHODS: This prospective cross-sectional study included 45 and 20 participants with abnormal and normal phoria, respectively. Eye movements were recorded using a remote eye tracker and were plotted using RStudio software. Image-based areas were evaluated using ImageJ software. RESULTS: The image-based areas used to evaluate fixation stability exhibited decreasing stability in the abnormal phoria group in the following order: KDE with ±1 standard deviation (SD), BCEA with ±1 SD, KDE with ±2 SD or Scanpath, and BCEA with ±2 SD. The BCEA tended to be overestimated, and the KDE tended to be underestimated at high density. The Scanpath method had a very high probability area because the area spans all gaze points. CONCLUSIONS: Fixation stability could be quantified as image-based areas by the KDE, BCEA, and Scanpath methods. Our findings suggest that fixation stability may be evaluated using one or more methods.

Asunto(s)

RESUMEN

Our objective is to analyze scanpaths acquired through participants achieving a reading task aiming at answering a binary question: Is the text related or not to some given target topic? We propose a data-driven method based on hidden semi-Markov chains to segment scanpaths into phases deduced from the model states, which are shown to represent different cognitive strategies: normal reading, fast reading, information search, and slow confirmation. These phases were confirmed using different external covariates, among which semantic information extracted from texts. Analyses highlighted some strong preference of specific participants for specific strategies and more globally, large individual variability in eye-movement characteristics, as accounted for by random effects. As a perspective, the possibility of improving reading models by accounting for possible heterogeneity sources during reading is discussed.

RESUMEN

The degree of spatial similarity between the gaze of participants viewing dynamic stimuli such as videos has been previously measured using metrics which are based on the NSS (Normalized Scanpath Saliency). Methods currently used to calculate this metric rely upon a numerical grid, which can be computationally prohibitive for a variety of otherwise useful applications such as Monte Carlo analyses. In the present work we derive a new analytical calculation method for the same metric that yields equal or more accurate results, but with speeds than can be orders of magnitude faster (depending on parameters). Our analytical method scales well with dimensionality, and could also be of use for other applications. The drawback is that it can become very slow if the number of participants in the study is very large or if the gaze sampling rate is high. We provide performance benchmarks for a Fortran implementation of our method, and make available the source code developed.

Asunto(s)

RESUMEN

The prediction of human gaze behavior is important for building human-computer interaction systems that can anticipate the user's attention. Computer vision models have been developed to predict the fixations made by people as they search for target objects. But what about when the target is not in the image? Equally important is to know how people search when they cannot find a target, and when they would stop searching. In this paper, we propose a data-driven computational model that addresses the search-termination problem and predicts the scanpath of search fixations made by people searching for targets that do not appear in images. We model visual search as an imitation learning problem and represent the internal knowledge that the viewer acquires through fixations using a novel state representation that we call Foveated Feature Maps (FFMs). FFMs integrate a simulated foveated retina into a pretrained ConvNet that produces an in-network feature pyramid, all with minimal computational overhead. Our method integrates FFMs as the state representation in inverse reinforcement learning. Experimentally, we improve the state of the art in predicting human target-absent search behavior on the COCO-Search18 dataset. Code is available at: https://github.com/cvlab-stonybrook/Target-absent-Human-Attention.

RESUMEN

The visual scanpath to emotional facial expressions was recorded in BR, a 35-year-old male with chronic severe traumatic brain injury (TBI), both before and after he underwent intervention. The novel intervention paradigm combined visual scanpath training with verbal feedback and was implemented over a 3-month period using a single case design (AB) with one follow up session. At baseline BR's scanpath was restricted, characterised by gaze allocation primarily to salient facial features on the right side of the face stimulus. Following intervention his visual scanpath became more lateralised, although he continued to demonstrate an attentional bias to the right side of the face stimulus. This study is the first to demonstrate change in both the pattern and the position of the visual scanpath to emotional faces following intervention in a person with chronic severe TBI. In addition, these findings extend upon our previous work to suggest that modification of the visual scanpath through targeted facial feature training can support improved facial recognition performance in a person with severe TBI.

RESUMEN

Outlier scanpaths identification is a crucial preliminary step in designing visual software, digital media analysis, radiology training and clustering participants in eye-tracking experiments. However, the task is challenging due to the visual irregularity of the scanpath shapes and the difficulty in dimensionality reduction due to geometric complexity. Conventional approaches have used heat maps to exclude scanpaths that lack a similarity pattern. However, the typically-used packages, such as ScanMatch and MultiMatch often generate discordant results when outlier identification is done empirically. This paper introduces a novel outlier evaluation approach by integrating the fractal dimension (FD), capturing the geometrical complexity of patterns, as an additional parameter with the heat map. This additional parameter is used to evaluate the degree of influence of a scanpath within a dataset. More specifically, the 2D Cartesian coordinates of a scanpath are fitted to a space filling 1D fractal curve to characterise its temporal FD. The FDs of the scanpaths are then compared to match their geometric complexity to one another. The findings indicate that the FD can be a beneficial additional parameter when evaluating the candidacy of poorly matching scanpaths as outliers and performs better at identifying unusual scanpaths than using other methods, including scanpath matching, Jaccard, or bounding box methods alone.

RESUMEN

Vision is an important way for human beings to interact with the outside world and obtain information. In order to research human visual behavior under different conditions, this paper uses a Gaussian mixture-hidden Markov model (GMM-HMM) to model the scanpath, and proposes a new model optimization method, time-shifting segmentation (TSS). The TSS method can highlight the characteristics of the time dimension in the scanpath, improve the pattern recognition results, and enhance the stability of the model. In this paper, a linear discriminant analysis (LDA) method is used for multi-dimensional feature pattern recognition to evaluates the rationality and the accuracy of the proposed model. Four sets of comparative trials were carried out for the model evaluation. The first group applied the GMM-HMM to model the scanpath, and the average accuracy of the classification could reach 0.507, which is greater than the opportunity probability of three classification (0.333). The second set of trial applied TSS method, and the mean accuracy of classification was raised to 0.610. The third group combined GMM-HMM with TSS method, and the mean accuracy of classification reached 0.602, which was more stable than the second model. Finally, comparing the model analysis results with the saccade amplitude (SA) characteristics analysis results, the modeling analysis method is much better than the basic information analysis method. Via analyzing the characteristics of three types of tasks, the results show that the free viewing task have higher specificity value and a higher sensitivity to the cued object search task. In summary, the application of GMM-HMM model has a good performance in scanpath pattern recognition, and the introduction of TSS method can enhance the difference of scanpath characteristics. Especially for the recognition of the scanpath of search-type tasks, the model has better advantages. And it also provides a new solution for a single state eye movement sequence.

Asunto(s)

RESUMEN

Ocular activity is known to be sensitive to variations in mental workload, and recent studies have successfully related the distribution of eye fixations to the mental load. This study aimed to verify the effectiveness of the spatial distribution of fixations as a measure of mental workload and its sensitivity to different types of demands imposed by the task: mental, temporal, and physical. To test the research hypothesis, two experimental studies were run: Experiment 1 evaluated the sensitivity of an index of spatial distribution (Nearest Neighbor Index; NNI) to changes in workload. A sample of 30 participants participated in a within-subject design with different types of task demands (mental, temporal, physical) applied to Tetris game; Experiment 2 investigated the accuracy of the index through the analysis of 1-min epochs during the execution of a visual-spatial task (the "spot the differences" puzzle game). Additionally, NNI was compared to a better-known ocular mental workload index, the entropy rate. The data analysis showed a relation between the NNI and the different workload levels imposed by the tasks. In particular: Experiment 1 demonstrated that increased difficulty, due to higher temporal demand, led to a more dispersed pattern with respect to the baseline, whereas the mental demand led to a more grouped pattern of fixations with respect to the baseline; Experiment 2 indicated that the entropy rate and the NNI show a similar pattern over time, indicating high mental workload after the first minute of activity. That suggests that NNI highlights the greater presence of fixation groups and, accordingly, the entropy indicates a more regular and orderly scanpath. Both indices are sensitive to changes in workload and they seem to anticipate the drop in performance. However, the entropy rate is limited by the use of the areas of interest, making it impossible to apply it in dynamic contexts. Conversely, NNI works with the entire scanpath and it shows sensitivity to different types of task demands. These results confirm the NNI as a measure applicable to different contexts and its potential use as a trigger in adaptive systems implemented in high-risk settings, such as control rooms and transportation systems.