Visual Spatial Information Research Articles

Tokengrid: Toward More Efficient Data Extraction From Unstructured Documents

Key information extraction from unstructured documents is a practical problem in many industries. Machine learning models aimed at solving this problem should efficiently utilize textual, visual, and 2D spatial layout information of the document. Grid based approaches achieve this by representing the document as a 2D grid and feeding it to a fully convolutional encoder-decoder network that solves a semantic instance segmentation problem. We propose a new method for the instance detection branch of that network for the task of automatic information extraction from invoices. Our approach reduces this problem to 1D region detection. The proposed network has fewer parameters and a shorter inference times. Additionally, we suggest a new metric for evaluating the results.

Dyslexia and Working Memory: Understanding Reading Comprehension and High Level Language Skills in Students with Dyslexia

Dyslexia is defined as difficulties determining speech sounds and learning the relationships of speech sounds with letters and words. Children with learning disabilities may experience problems in attention, memory, perception, motor processing, information processing speed, planning and problem-solving skills. Although lack of phonological processing is considered one of the leading causes of dyslexia, the level of influence of other underlying factors such as cognitive deficits on reading gain is still unclear. Numerous studies have shown that dyslexia is associated with poor working memory, a critical component of reading skill acquisition because the temporary processing of newly introduced and previously stored information involves critical thinking, use of cognitive executive skills, comprehension, and learning tasks. Working memory is often used synonymously with short-term memory, but some theorists consider their functions distinct in that working memory allows for the manipulation of the information temporarily stored in short-term memory. Working memory, which provides preservation, integration and processing of verbal and visual-spatial information, works together with short-term memory to help the mind manipulate and determine important information while temporary. This study discussed the effects of working memory on reading, reading comprehension, and high-level language skills.

RDFNet: A Fast Caries Detection Method Incorporating Transformer Mechanism.

Dental caries is a prevalent disease of the human oral cavity. Given the lack of research on digital images for caries detection, we construct a caries detection dataset based on the caries images annotated by professional dentists and propose RDFNet, a fast caries detection method for the requirement of detecting caries on portable devices. The method incorporates the transformer mechanism in the backbone network for feature extraction, which improves the accuracy of caries detection and uses the FReLU activation function for activating visual-spatial information to improve the speed of caries detection. The experimental results on the image dataset constructed in this study show that the accuracy and speed of the method for caries detection are improved compared with the existing methods, achieving a good balance in accuracy and speed of caries detection, which can be applied to smart portable devices to facilitate human dental health management.

A multiscale double-branch residual attention network for anatomical–functional medical image fusion

Medical image fusion technology synthesizes complementary information from multimodal medical images. This technology is playing an increasingly important role in clinical applications. In this paper, we propose a new convolutional neural network, which is called the multiscale double-branch residual attention (MSDRA) network, for fusing anatomical–functional medical images. Our network contains a feature extraction module, a feature fusion module and an image reconstruction module. In the feature extraction module, we use three identical MSDRA blocks in series to extract image features. The MSDRA block has two branches. The first branch uses a multiscale mechanism to extract features of different scales with three convolution kernels of different sizes, while the second branch uses six 3 × 3 convolutional kernels. In addition, we propose the Feature L1-Norm fusion strategy to fuse the features obtained from the input images. Compared with the reference image fusion algorithms, MSDRA consumes less fusion time and achieves better results in visual quality and the objective metrics of Spatial Frequency (SF), Average Gradient (AG), Edge Intensity (EI), Quality-Aware Clustering (QAC), Variance (VAR), and Visual Information Fidelity for Fusion (VIFF).

Preliminary Findings on Visual Event-Related Potential P3 in Asymptomatic Patients with Cerebral Small Vessel Disease

BackgroundCerebral small vessel disease is the primary cause of cognitive impairment. Therefore, early recognition is of great significance. Some studies have shown that asymptomatic cerebral small vessel disease (aCSVD) patients have abnormal neurocognitive function, but this is not readily apparent at the initial stage. The objective of this paper was to assess visual spatial attention by event-related potential (ERP) examination and to analyze the relationship between ERP data and clinical characteristics in patients with aCSVD.MethodsWe selected 25 aCSVD patients and enrolled 23 age-matched normal subjects as the control group. We measured the latency and amplitude of original/corresponding differential ERP components using the modified visual oddball paradigm, which included a standard stimulus, target stimulus, and new stimulus. Additionally, we selected aberrant ERP components to study the correlations between the ERP data and clinical characteristics of the patients with aCSVD.ResultsWe found not only lower amplitude but also significantly longer P3 latency in the aCSVD patients. The above results were further verified by analyzing the different components (target minus standard and novel minus standard) of P3. Furthermore, abnormal ERPs in the aCSVD patients were closely related to the changes observed with imaging.ConclusionIt was demonstrated that the speed and capability of processing visual spatial information was impaired in aCSVD patients compared with healthy controls. Thus, ERP examination could detect the presence of attentional deficits and might become a rapid and sensitive method for the early diagnosis of aCSVD. However, its availability needs further investigation.

Evaluating the Fitness-to-Drive Using Evoked Visual Responses in Alzheimer's Disease.

Alzheimer's Disease (AD) is the sixth leading cause of death in the US. AD causes significant disability due to the devastating impact on the patients' day-to-day living activities and their loss of independence. One such day-to-day activity is driving, a complex task that requires attention, concentration, the ability to follow particular steps, react to stimuli promptly, and the ability to perceive and interpret visual-spatial information, all of which can be impaired in AD. Therefore, to ensure the safety of AD patients and other drivers, it is important to develop accurate and low-cost diagnostic tools to assess patients' fitness-to-drive. In this study, we develop machine learning (ML) models to predict fitness-to-drive using the electroencephalogram (EEG) technique of event-related potential (ERP). Specifically, we develop random forest (RF) models using EEG signals in early-stage AD patients and age-matched controls and conduct numerical experiments to predict fitness-to-drive and other driving performance metrics, collected from driving simulator data. Our results show that RF models predict patients' fitness-to-drive with AUC=0.83 and provide accurate measures of other driving performance metrics. Therefore, ML and ERP offer a valuable approach to assess driving safety for patients with early AD symptoms in the laboratory setting.

Partial visual loss disrupts the relationship between judged room size and sound source distance

Visual spatial information plays an important role in calibrating auditory space. Blindness results in deficits in a number of auditory abilities, which have been explained in terms of the hypothesis that visual information is needed to calibrate audition. When judging the size of a novel room when only auditory cues are available, normally sighted participants may use the location of the farthest sound source to infer the nearest possible distance of the far wall. However, for people with partial visual loss (distinct from blindness in that some vision is present), such a strategy may not be reliable if vision is needed to calibrate auditory cues for distance. In the current study, participants were presented with sounds at different distances (ranging from 1.2 to 13.8 m) in a simulated reverberant (T60 = 700 ms) or anechoic room. Farthest distance judgments and room size judgments (volume and area) were obtained from blindfolded participants (18 normally sighted, 38 partially sighted) for speech, music, and noise stimuli. With sighted participants, the judged room volume and farthest sound source distance estimates were positively correlated (p < 0.05) for all conditions. Participants with visual losses showed no significant correlations for any of the conditions tested. A similar pattern of results was observed for the correlations between farthest distance and room floor area estimates. Results demonstrate that partial visual loss disrupts the relationship between judged room size and sound source distance that is shown by sighted participants.

Effects of Chronic Alcohol Use Disorder on the Visual Tilt Illusion.

Rationale: Among the serious consequences of alcohol use disorder (AUD) is the reduced ability to process visual information. It is also generally agreed that AUD tends to occur with disturbed excitation–inhibition (EI) balance in the central nervous system. Thus, a specific visual behavioral probe could directly qualify the EI dysfunction in patients with AUD. The tilt illusion (TI) is a paradigmatic example of contextual influences on perception of central target. The phenomenon shows a characteristic dependence on the angle between the inducing surround stimulus and the central target test. For small angles, there is a repulsion effect; for larger angles, there is a smaller attraction effect. The center-surround inhibition in tilt repulsion is considered to come from spatial orientational interactions between orientation-tuned neurons in the primary visual cortex (V1), and tilt attraction is from higher-level effects of orientation processing in the visual information processing.Objectives: The present study focuses on visual spatial information processing and explores whether chronic AUD patients in abstinence period exhibited abnormal TI compared with healthy controls.Methods: The participants are 30 male volunteers (20–46 years old) divided into two groups: the study group consists of 15 clinically diagnosed AUD patients undergoing abstinence from alcohol, and the control group consists of 15 healthy volunteers. The TI consists of a center target surround with an annulus (both target and annulus are sinusoidal grating with spatial frequency = 2 cycles per degree). The visual angle between center and surround is a variable restricted to 0°, ±15°, ±30°, or ±75°. For measuring the TI, participants have to report whether the center target grating orientation tilted clockwise or counterclockwise from the internal vertical orientation by pressing corresponding keys on the computer keyboard. No feedback is provided regarding response correctness.Results: The results reveal significantly weaker tilt repulsion effect under surround orientation ±15° (p < 0.05) and higher lapse rate (attention limitation index) under all tested surround orientations (all ps < 0.05) in patients with chronic AUD compared with health controls.Conclusions: These results provide psychophysical evidence that visual perception of center-contextual stimuli is different between AUD and healthy control groups.

Kea (Nestor notabilis) show flexibility and individuality in within-session reversal learning tasks

The midsession reversal paradigm confronts an animal with a two-choice discrimination task where the reward contingencies are reversed at the midpoint of the session. Species react to the reversal with either win-stay/lose-shift, using local information of reinforcement, or reversal estimation, using global information, e.g. time, to estimate the point of reversal. Besides pigeons, only mammalian species were tested in this paradigm so far and analyses were conducted on pooled data, not considering possible individually different responses. We tested twelve kea parrots with a 40-trial midsession reversal test and additional shifted reversal tests with a variable point of reversal. Birds were tested in two groups on a touchscreen, with the discrimination task having either only visual or additional spatial information. We used Generalized Linear Mixed Models to control for individual differences when analysing the data. Our results demonstrate that kea can use win-stay/lose-shift independently of local information. The predictors group, session, and trial number as well as their interactions had a significant influence on the response. Furthermore, we discovered notable individual differences not only between birds but also between sessions of individual birds, including the ability to quite accurately estimate the reversal position in alternation to win-stay/lose-shift. Our findings of the kea’s quick and flexible responses contribute to the knowledge of diversity in avian cognitive abilities and emphasize the need to consider individuality as well as the limitation of pooling the data when analysing midsession reversal data.

Selective memory disrupted in intra-modal dual-task encoding conditions.

Given natural memory limitations, people can generally attend to and remember high-value over low-value information even when cognitive resources are depleted in older age and under divided attention during encoding, representing an important form of cognitive control. In the current study, we examined whether tasks requiring overlapping processing resources may impair the ability to selectively encode information in dual-task conditions. Participants in the divided-attention conditions of Experiment 1 completed auditory tone-distractor tasks that required them to discriminate between tones of different pitches (audio-nonspatial) or auditory channels (audio-spatial), while studying items in different locations in a grid (visual-spatial) differing in reward value. Results indicated that, while reducing overall memory accuracy, neither cross-modal auditory distractor task influenced participants' ability to selectively encode high-value items relative to a full attention condition, suggesting maintained cognitive control. Participants in Experiment 2 studied the same important visual-spatial information while completing demanding color (visual-nonspatial) or pattern (visual-spatial) discrimination tasks during study. While the cross-modal visual-nonspatial task did not influence memory selectivity, the intra-modal visual-spatial secondary task eliminated participants' sensitivity to item value. These results add novel evidence of conditions of impaired cognitive control, suggesting that the effectiveness of top-down, selective encoding processes is attenuated when concurrent tasks rely on overlapping processing resources.

Holistic processing of Chinese characters in college students with dyslexia

Expert face recognition has long been marked by holistic processing. Hence, due to the many visual properties shared between face perception and Chinese characters, it has been suggested that Chinese character recognition may induce stronger holistic processing in expert readers than in novices. However, there have been different viewpoints presented about Chinese character recognition, one of which suggests that expertise in this skill involved reduced holistic processing which may be modulated by writing experiences/performances. In this study we examined holistic processing in Chinese character recognition in adults with and without dyslexia, using the complete composite paradigm. Our results showed that the adults with dyslexia recognized Chinese characters with a stronger holistic processing effect than the typical controls. It seems that those with dyslexia relied overly on the visual spatial information of characters and showed deficits in attending selectively to their components when processing Chinese characters, which hindered the development of expert reading and writing skills. This effect was in contrast to previous perceptual expertise studies in which reduced holistic processing marked deficits in face/visual object recognition. This study is also the first to show that Chinese adults with dyslexia had persistent below average performances in Chinese literacy.

A Spatially-Coded Visual Brain-Computer Interface for Flexible Visual Spatial Information Decoding.

Conventional visual BCIs, in which control channels are tagged with stimulation patterns to elicit distinguishable brain patterns, has made impressive progress in terms of the information transfer rates (ITRs). However, less development has been seen with respect to user experience and complexity of the technical setup. The requirement to tag each of targets by a unique stimulus substantially limits the flexibility of conventional visual BCI systems. A method for decoding the targets in the environment flexibly was therefore proposed in the present study. A BCI speller with thirteen symbols drawn on paper was developed. The symbols were interspersed with four flickers with distinct frequencies, but the user did not have to gaze at flickers. Rather, subjects could spell a sequence by looking at the symbols on the paper. In a cue-guided spelling task, the average offline and online accuracies reached 89.3± 7.3% and 90.3± 6.9% for 13 subjects, corresponding to ITRs of 43.0± 7.4 bit/min and 43.8± 6.8 bit/min. In an additional free-spelling task for seven out of thirteen subjects, an accuracy of 92.3± 3.1% and an ITR of 45.6± 3.3 bit/min were achieved. Analysis of a simulated online system showed the possibility to reach an average ITR of 105.8 bit/min by reducing the epoch duration from 4 to 1 second. Reliable BCI control is possible by gazing at targets in the environment instead of dedicated stimuli which encode control channels. The proposed method can drastically reduce the technical effort for visual BCIs and thereby advance their applications outside the laboratory.

Elaboration of Neuropsychological Evaluation of Children: Structural Analysis of Test Results.

Modern neuropsychology is discussing the possibility of combining qualitative and quantitative approaches in the evaluation of cognitive functions. In Russia a battery of tests called "Methods of neuropsychological assessment for children 6-9 years old" (Akhutina et al., 2016) has been proposed; it is based on the Lurian approach to diagnosis and combines qualitative and quantitative approaches to testing. The present paper describes the development of this combined qualitative and quantitative assessment of various groups of cognitive functions in preschool and primary school children. Structural modeling enables us to analyze a possible combination of integral indices of functions that includes the results of both a face-to-face neuropsychological assessment and computerized testing. To develop a combined qualitative and quantitative neuropsychological assessment of children, in order to 1) check the structural reliability of integral indicators of various cognitive functions; and 2) confirm the correctness of combining the results of face-to-face and computerized tests. A sample of 299 children between the ages of 6 and 9 years old (111 preschoolers, 82 first graders, and 106 second graders) underwent a Lurian face-to-face neuropsychological examination adapted for 6-to-9 year-old children, and five tests from the Computerized Neuropsychological Assessment for 6-9 Year-old Children. The five were the "Dots" test, the Schulte Tables, the Cancellation test, the Corsi Tapping Block test, and the Understanding of Similar Sounding Words test. In each of the tests (face-to-face and computerized), key parameters were identified to evaluate various cognitive functions. A confirmatory factor analysis verified the composition of the neuropsychological indices that were based on the results of the face-to-face neuropsychological assessment. At the same time, when the computer test data were added to the model, the fit indices of the model considerably improved. The confirmatory factor analysis confirmed the validity of the identification of eight neuropsychological indices that indicate the component processes underlying complex cognitive functions in children: 1) programming and control of voluntary actions (executive functions); 2) serial organization of movements and speech; 3) the processing of kinesthetic information; 4) the processing of auditory information; 5) the processing of visual information; 6) the processing of visual-spatial information; 7) hyperactivity/impulsivity; and 8) fatigue/slowness.

Working, Declarative, and Procedural Memory in Children With Developmental Language Disorder.

Purpose Previous research into the working, declarative, and procedural memory systems in children with developmental language disorder (DLD) has yielded inconsistent results. The purpose of this research was to profile these memory systems in children with DLD and their typically developing peers. Method One hundred four 5- to 8-year-old children participated in the study. Fifty had DLD, and 54 were typically developing. Aspects of the working memory system (verbal short-term memory, verbal working memory, and visual-spatial short-term memory) were assessed using a nonword repetition test and subtests from the Working Memory Test Battery for Children. Verbal and visual-spatial declarative memory were measured using the Children's Memory Scale, and an audiovisual serial reaction time task was used to evaluate procedural memory. Results The children with DLD demonstrated significant impairments in verbal short-term and working memory, visual-spatial short-term memory, verbal declarative memory, and procedural memory. However, verbal declarative memory and procedural memory were no longer impaired after controlling for working memory and nonverbal IQ. Declarative memory for visual-spatial information was unimpaired. Conclusions These findings indicate that children with DLD have deficits in the working memory system. While verbal declarative memory and procedural memory also appear to be impaired, these deficits could largely be accounted for by working memory skills. The results have implications for our understanding of the cognitive processes underlying language impairment in the DLD population; however, further investigation of the relationships between the memory systems is required using tasks that measure learning over long-term intervals. Supplemental Material https://doi.org/10.23641/asha.13250180.

Functional Features of the Brain Blocks and Prevailing Strategies of Information Processing in Adults

The statutory meanings of higher mental functions in adults remained understudied for a long time. The present research was based on a neuropsychological analysis of the functional capacities of three brain blocks and favored strategies of information processing in adults. The analysis made it possible to distinguish five main types: 1 – with relatively weak functions of the dynamic praxis; 2 – with a moderate deficit of the left hemisphere functions; 3 – with relatively weak visualspatial functions; 4 – with a moderate deficit of brain block I resulting in the general deterioration of higher mental functions; 5 – with no functional deficit. The neuropsychological indexes showed that subgroup 1 had the lowest values of the functions of brain block III and a high level of processing of visual, auditory, and visual-spatial information. In subgroup 2, the index of the state of left hemisphere functions was law, which resulted in weak analytical strategy for processing auditory information, verbal-perceptual problems, and a slight decrease in brain block I. This meant a decrease in the voluntary regulation of activity. Subgroup 3 had the lowest index of right hemisphere functions and a deficiency of right hemispheric holistic information processing strategies. Subgroup 4 had low functional capacities of brain block I and other indicators, especially those associated with the processing of auditory, visual-spatial information, planning, and control. The adults in subgroup 5 showed good for all the indicators.

BIOELECTRIC ACTIVITY OF THE BRAIN IN PUPILS WITH DIFFERENT ACADEMIC SUCCESS. REPORT 2. THE RELATIONSHIP BETWEEN THE ACADEMIC SUCCESS AND PATTERNS OF BIOELECTRIC ACTIVITY OF THE CORTEX

Background. The paper presents the study of the factors of academic success and failure in students, which is a relevant and socially significant problem. Academic failure often results in behavior deviations, drug abuse and other types of dangerous behavior. Therefore, it is of utmost importance to establish the reasons of academic failure, as well as the ways to improve academic performance. Aim. The paper aims to establish the features of responses of academically successful and unsuccessful 12-year-old children to cognitive load based on EEG data. Materials and methods. Twelve-year-old schoolchildren (n = 51) from Ekaterinburg participated in the study. Participants were divided into two groups depending on their academic performance. EEG recording was performed using the CONAN-m equipment (Informatics and Computers, Russia). Monopolar EEG recordings were obtained from 10 symmetrical leads in different conditions: during resting wakefulness (for eyes open and closed) and when solving experimental tasks of three types (simple arithmetic task, verbal and logical task, spatial thinking task). The differences between the samples of academically successful and unsuccessful children were assessed using the Wilcoxon nonparametric test. Results. Differences were revealed in a number of EEG indicators both at rest and when solving experimental tasks between academically successful and unsuccessful schoolchildren. The responses of academically successful children to cognitive load were characterized by predominant ctivation of the anterior cortical areas with a focus of activity in the left frontal area. Academically unsuccessful children were characterized by a more generalized type of EEG response with a focus of activity in the caudal areas of the cortex, especially when solving a figure rotation task. It was also found that successful children made mistakes when solving a figure rotation task and unsuccessful ones-when solving verbal and logical tasks. Conclusion. Children with low academic success were found to have characteristic features of the functioning of the cerebral cortex, which hampered the perception of educational material presented in the form of a logically structured message. However, these children were able to brilliantly operate with visual-spatial information. This must be taken into account when working with such children to improve their academic performance.

Interindividual differences in brain dynamics of early visual processes: Impact on score accuracy in the mental rotation task.

Interindividual variations in the ability to perform visuospatial mental transformations have been investigated extensively, in particular through mental rotation tasks. However, the impact of early visual processes on performance has been largely ignored. To clarify this issue, we explored the time-course of early visual processing (from 0 to 450ms poststimulus) using event-related potentials topographic analyses. The main findings demonstrated a significant link between early attentional processes and accuracy scores occurring more than five seconds later, as well as a strong association between spatial covariance and microstate topographies exhibiting substantial gender differences. More specifically, the results indicated that, in a classical mental rotation task, the male brain expends more time processing visual-spatial information resulting in a longer bilateral positive potential at posterior-occipital sites. In comparison, the female brain initiates earlier processing of non-spatial information resulting in a faster transition from a bilateral positive potential of posterior-occipital sites to a negative potential at central-frontal sites. These findings illustrate how a more complete utilization of the spatiotemporal information contained in EEG recordings can provide important insights about the impact of early visual processes on interindividual differences, particularly across gender, and thus shed new light on alternate cognitive strategies.

Enhanced Gray Matter Volume Compensates for Decreased Brain Activity in the Ocular Motor Area in Children with Anisometropic Amblyopia

PurposeAnisometropic amblyopia usually occurs during early childhood and results in monocular visual deficit. Recent neuroimaging studies have demonstrated structural and functional alterations in pediatric anisometropic amblyopia (PAA) patients. However, the relationship between structural and functional alterations remains largely unknown. The aim of this study was to investigate the relationship between structural and functional alterations in PAA patients. Materials and Methods Eighteen PAA patients and 14 healthy children underwent a multimodal MRI scanning including T1WI and functional MRI (fMRI). Voxel-based morphometry was used to assess structural alterations between PAA patients and healthy children. Regional homogeneity (ReHo) was used to investigate changes in local spontaneous brain activity in the enrolled subjects. Correlations between structural, functional alterations, and clinical information were analyzed in the PAA group. Results Compared with healthy children, PAA patients exhibited significantly reduced ReHo of spontaneous brain activity in the right superior temporal gyrus (STG) and right middle frontal gyrus (MFG) and increased gray matter volume in the right lobules 4 and 5 of the cerebellum. The gray matter volume of the right lobules 4 and 5 of the cerebellum was negatively correlated with the ReHo values of the right MFG. Conclusions Our findings may suggest that PAA patients experience structural and functional abnormalities in brain regions related to oculomotor and visual-spatial information. In addition, the increased gray matter volume may compensate the decreased brain activity in the oculomotor regions, which reflects compensatory or neural plasticity in PAA patients.

The Comparison of Mental Rotation Performances and Academic Achievements in Students of the Faculty of Sport Sciences and Education Sciences

Mental rotation (MR) is the ability to rotate three-dimensional objects using the imagination. As the ability to generate, retrieve, maintain, and manipulate visual-spatial information, MR plays a critical role throughout education and sport. In this aspect our objective in this study was to compare MR performances and academic achievement between students of art, foreign languages (FL), computer-instructional technologies (CIT) and physical education (PE).&#x0D; &#x0D; In the computer-based MR test, the image files of the library &amp;ldquo;Mental Rotation Stimulus Library&amp;rdquo; were used with the written permission of the author. A total of 144 participants composed of 79 males and 65 females volunteered in the study.&#x0D; &#x0D; Comparing the academic achievements (AA) with the PE group and art, CIT, FL and control groups; it was observed that all the groups are better than PE (p &amp;lt; 0.05). However, no significant difference was found between PE and all groups in MR performances (p &amp;gt; 0.05).&#x0D; &#x0D; In conclusion, PE students performed similarly with other students in terms of MR, although their AA were low. MR ability is clearly related not only to problem solving, acquiring mathematical knowledge and academic thinking but also to motor abilties.

Long-Term Loop Closure Detection through Visual-Spatial Information Preserving Multi-Order Graph Matching

Loop closure detection is a fundamental problem for simultaneous localization and mapping (SLAM) in robotics. Most of the previous methods only consider one type of information, based on either visual appearances or spatial relationships of landmarks. In this paper, we introduce a novel visual-spatial information preserving multi-order graph matching approach for long-term loop closure detection. Our approach constructs a graph representation of a place from an input image to integrate visual-spatial information, including visual appearances of the landmarks and the background environment, as well as the second and third-order spatial relationships between two and three landmarks, respectively. Furthermore, we introduce a new formulation that formulates loop closure detection as a multi-order graph matching problem to compute a similarity score directly from the graph representations of the query and template images, instead of performing conventional vector-based image matching. We evaluate the proposed multi-order graph matching approach based on two public long-term loop closure detection benchmark datasets, including the St. Lucia and CMU-VL datasets. Experimental results have shown that our approach is effective for long-term loop closure detection and it outperforms the previous state-of-the-art methods.