Attention Training Research Articles

Neural correlates of virtual reality-based attention training: An fMRI study

Theoretical backgroundVirtual Reality technology is increasingly used in attention rehabilitation for functional training purposes. However, the neural mechanisms by which Virtual Reality can affect attentional functioning are still unclear. The current study's objective is to examine the effects of stereoscopic vs. monoscopic presentation on neural processing during a visual attention task. MethodThirty-two healthy participants performed a visual attention task in an immersive virtual environment that was displayed via MR-compatible video goggles in an MRI scanner. The paradigm altered between trials that required active engagement with the task and mere observation trials. Furthermore, the form of binocular presentation switched between monoscopic and stereoscopic presentation. ResultsAnalyses yielded evidence for increased activation in stereoscopic compared to monoscopic trials in the tertiary visual cortex area V3A as well as elevated activation in the dorsal attention network when engaging in the attention task. An additional ROI analysis of area V3A revealed significantly lower attentional engagement costs in stereoscopic conditions. DiscussionResults support previous findings suggesting that V3A is involved in binocular depth perception. Furthermore, heightened activation in V3A following stereoscopic presentation seemed to facilitate attentional engagement with the task. Considering that V3A is the origin of the dorso-dorsal, ventro-dorsal, and ventral visual processing pathways, we regard it as a gating area that decides which kind of visual perception is processed.

26 Improvement in Executive Function Following Goal-Oriented Attentional Self-Regulation Training in Veterans with mTBI

Objective:Goal-Oriented Attentional Self-Regulation (GOALS) is a cognitive rehabilitation training that targets difficulties with executive control functions. Prior research has found that GOALS training improves executive function, functional outcomes, and PTSD symptoms in Veterans with comorbid PTSD and TBI. The objective of this study is to assess the effects of GOALS training versus an active time-matched Behavioral Health Education (BHE) intervention in military Veterans with history of mTBI and executive function difficulties.Participants and Methods:Thirty-four veterans with a history of mild TBI (>6 months post-injury) and cognitive difficulties were randomized to GOALS or BHE group intervention. GOALS is a cognitive rehabilitation intervention that incorporates mindfulness-based attention regulation training, and goal management strategies applied to participant-defined goals. It consists of ten 2-hour sessions administered in small group format, and three individual 1-hour sessions. The BHE group is a matched control condition that mirrors GOALS in terms of duration and level of involvement. Participants completed a comprehensive evaluation consisting of neuropsychological and ecologically valid functional performance measures of complex attention and executive function, before and after GOALS or BHE intervention. Overall neuropsychological attention/executive functioning included measures of working memory (Letter Number Sequencing, Auditory Consonant Trigrams), mental flexibility (Design & Verbal Fluency Switching, Trails B, Stroop Inhibition/Switching), inhibition (Stroop Inhibition), and sustained attention (Digit Vigilance). The Goal Processing Scale (GPS) was used to measure ecologically valid complex functional performance.Results:Post GOALS training, but not post BHE, participants significantly improved on measures of attention /executive function including overall performance (p=.02), working memory (p=.04), sustained attention (p=.03)., and inhibition (p=.01). On measures of complex functional task performance, post GOALS participants significantly improved in the domain of planning (p=.02).Conclusions:GOALS training improved performance on neuropsychological and ecologically valid functional measures of complex attention and executive function in Veterans with mild TBI and cognitive difficulties. The current study highlights how mindfulness-based attention regulation and goalmanagement strategies applied to personal goals has positive effects across various outcomes. A percentage of individuals with mTBI continue to report persistent problems with cognition and daily functioning. Cognitive rehabilitation training such as GOALS may be beneficial for those with persistent difficulties following mTBI, even in chronic phase.

Cognitive training for children with ADHD: composite cognitive score outcome in a randomized controlled trial

Purpose of the article Cognitive training for Attention Deficit/Hyperactivity Disorder (ADHD) has shown promising, although mixed results. In post-hoc analyses, we evaluate effects of cognitive training using a novel composite cognition score as the outcome for children attending at least 16 sessions of training, dose-response of training and associations between symptoms and cognitive functioning. Materials and methods Children (age 6–13) with ADHD were randomized to intervention (n = 26) or control (n = 34). For the current analysis, we restricted the intervention group to children, who completed at least 16 sessions of cognitive training (n = 26) and examined a dose response within that group. Results Cognition improved significantly in the intervention, but not control group. Amount of the completed training sessions correlated significantly with the amount of cognitive improvement. Conclusion Variations in dose and frequency of training may be an important source of the variance in previous studies.

A progressive deep learning framework for fine-grained primate behavior recognition

Fine-grained primate behavior recognition is crucial for conservation efforts and longitudinal research, but a lack of suitable datasets and complex network structures has hindered its progress. To address these challenges, we propose a novel approach that employs a region-focused deep convolutional neural network and a labeled fine-grained recognition dataset. Our model leverages a progressive attention training strategy that emphasizes discriminative region attention and encourages complementarity with multiple leading levels. Additionally, we incorporate a region-focused image generator that identifies informative local regions and provides a priori knowledge for the subsequent layer. Our experiments demonstrate that our model surpasses state-of-the-art performance on fine-grained primate behavior recognition benchmark datasets, even outperforming some models by leveraging auxiliary information. Our approach produces the automated fine-grained action recognition of primate behavior, making it a significant step for exploiting large datasets in ethology and conservation. This approach can be readily applied to a more comprehensive range of species and scenarios.

The role of attention problems in predicting applying to college among high school girls with disabilities

Students with disabilities have higher rates of attention problems than those without disabilities. This can impede their academic success and postsecondary transition, but these effects have not been well-studied. Understanding these effects is especially critical among high school girls with disabilities who additionally experience significant other barriers to college enrolment. Using longitudinal data from 366 high school girls with disabilities, we examined whether attention problems predicted a lower likelihood of applying to college, and whether this effect was mediated by academic difficulties. We also tested whether attention problems moderated the effect of students’ future aspirations on the likelihood of applying to college. Consistent with our predictions, attention problems were associated with a lower likelihood of applying to college. The individual paths through academic difficulties were significant, but the bootstrap estimation of the indirect effect was not significant. Attention problems did not moderate the effect of future aspirations. Attention training interventions have the potential to improve postsecondary educational outcomes.

Investigating Learning Trajectories on Digital Attention Training Tasks in Primary School Children

Attention training programs have demonstrated potential for improving select cognitive skills and behaviors in children, but reported benefits are inconsistent. It is unclear whether variability in training benefits can be attributed to differences in children’s learning trajectories on training tasks over the intervention period. This study examined the functional form of learning trajectories on adaptive attention training tasks in primary school children, and potential associations between these learning trajectories and (a) pre-training child characteristics (general cognitive ability, hyperactivity, inattention, age) and (b) outcomes on untrained tasks from pre- to post-training (attention, hyperactivity, inattention). A total of 38 children (5–9 years) completed a 5-week attention training program in class. The training sessions involved four tasks targeting selective attention, sustained attention, inhibition, or interference control. Assessments were conducted pre- and immediately post-training. Based on non-linear mixed-effects models, the learning trajectories were best described by asymptotic regression, where the largest rate of improvement occurred initially, followed by gradual flattening out as task difficulty approached the asymptote. For the sustained attention training task, lower asymptotes predicted larger reductions in hyperactive behaviour from pre- to post-training. For the sustained and selective attention training tasks, age was associated with the asymptotes of the learning trajectories. Collectively, these findings suggest that learning trajectories on certain attention training tasks are associated with select child characteristics and training outcomes. It is of interest for future cognitive training studies to examine the learning trajectories of tasks targeting different skills to contribute understanding of the processes associated with cognitive training outcomes.

Side of motor symptom onset predicts sustained attention deficits and motor improvements after attention training in Parkinson's disease

ObjectiveParkinson's disease (PD) side of motor symptom onset has been associated with distinct cognitive deficits; individuals with left-side onset (LPD) show more visuospatial impairments, whereas those with right-side onset (RPD) show more verbal impairments. Non-spatial attention is a critical cognitive ability associated with motor functioning that is right hemisphere lateralized but has not been characterized with regard to PD side of onset. We compared individuals with LPD and RPD on non-spatial attention tasks and examined differential responses to a 4-week sustained attention training program. MethodParticipants included 9 with LPD and 12 with RPD, who performed both brief and extended go/no-go continuous performance tasks and an attentional blink task. Participants also engaged in an at-home sustained attention training program, Tonic and Phasic Alertness Training (TAPAT), 5 days/week for 4 weeks. We assessed cognitive and motor symptoms before and after training, and after a 4-week no-contact period. ResultsAt baseline, participants with LPD exhibited worse performance than those with RPD on the extended continuous performance task, indicating specific deficits in sustaining attention. Poorer attention was associated with worse clinical motor scores. Notably, side of onset had a significant effect on clinical motor changes after sustained attention training, with only LPD participants improving after training, and 4/9 showing clinically meaningful improvements. ConclusionsCompared to RPD, participants with LPD had poorer sustained attention pre-training and were more likely to improve on clinical motor functioning after sustained attention training. These findings support mechanistic differences between LPD and RPD and suggest potential differential treatment approaches.

The Effects of Quiet Eye Training on Attention in Children with ADHD.

Based on linkage between attention control, gaze and visuomotor control, previous studies suggested that quiet eye training (QET) could improve attention control. The aim of the study was to investigate the effects of QET-based visuomotor intervention on different aspects of attention in children with ADHD. Fifty seven children with ADHD aged 9-12 years, were tested on focused, tonic and phasic attention, as well as attention control during a throwing task analysed via eye tracking, in pre- and post-test sessions, six weeks apart. The QET group that underwent specific 5-week QET-based intervention significantly increased focused attention including its accuracy aspect. Reaction times (RTs) and intraindividual RT variability under non-alert and alert conditions as measures of tonic and phasic attention did not change over time in either group. In contrast with the control (CON) group, an extended total time of eye fixations during the pre-throw and throw phase was found in the QET group after the intervention. The results of this study suggest that focused attention in children with ADHD can be improved by a short-term QET-based visuomotor intervention. Future research needs to examine the effects of a longer QET-based intervention, or the intervention in combination with other cognitive training of attention.

Intermittent theta burst stimulation for poststroke non-spatial attention deficit: a protocol of prospective, double-blinded, single-centre, randomised controlled trial in China

IntroductionAttention deficit is the most common cognitive impairment after stroke, which can significantly hinder the recovery of both other cognitive domains and motor functions. Increasing evidence suggests that the left...

Enhancing attention in autism spectrum disorder: comparative analysis of virtual reality-based training programs using physiological data

BackgroundThe ability to maintain attention is crucial for achieving success in various aspects of life, including academic pursuits, career advancement, and social interactions. Attention deficit disorder (ADD) is a common symptom associated with autism spectrum disorder (ASD), which can pose challenges for individuals affected by it, impacting their social interactions and learning abilities. To address this issue, virtual reality (VR) has emerged as a promising tool for attention training with the ability to create personalized virtual worlds, providing a conducive platform for attention-focused interventions. Furthermore, leveraging physiological data can be instrumental in the development and enhancement of attention-training techniques for individuals.MethodsIn our preliminary study, a functional prototype for attention therapy systems was developed. In the current phase, the objective is to create a framework called VR-PDA (Virtual Reality Physiological Data Analysis) that utilizes physiological data for tracking and improving attention in individuals. Four distinct training strategies such as noise, score, object opacity, and red vignette are implemented in this framework. The primary goal is to leverage virtual reality technology and physiological data analysis to enhance attentional capabilities.ResultsOur data analysis results revealed that reinforcement training strategies are crucial for improving attention in individuals with ASD, while they are not significant for non-autistic individuals. Among all the different strategies employed, the noise strategy demonstrates superior efficacy in training attention among individuals with ASD. On the other hand, for Non-ASD individuals, no specific training proves to be effective in enhancing attention. The total gazing time feature exhibited benefits for participants with and without ASD.DiscussionThe results consistently demonstrated favorable outcomes for both groups, indicating an enhanced level of attentiveness. These findings provide valuable insights into the effectiveness of different strategies for attention training and emphasize the potential of virtual reality (VR) and physiological data in attention training programs for individuals with ASD. The results of this study open up new avenues for further research and inspire future developments.

Attention training technique delivered in groups as treatment for anxiety and depression in patients with coronary heart disease: study protocol for a waiting-list randomized controlled trial.

Clinically significant symptoms of depression and anxiety in coronary heart disease (CHD) patients are common and associated with adverse outcomes. Psychological treatments have shown limited effectiveness and more effective treatments have been requested. Attention training technique (ATT), a component of metacognitive therapy, can potentially be effective as a stand-alone treatment for anxiety and depression. In an open study, ATT delivered face-to-face in a group format was feasible and potentially effective for improving depression and anxiety symptoms in CHD patients. The next progressive step is to test the effectiveness of ATT in a randomized controlled trial. This paper describes the methodology of this trial. ATT-CHD is a randomized wait-list (WL) controlled study. Eligible CHD patients from two hospitals with Hospital Anxiety and Depression Scale (HADS)-Anxiety and/or HADS-Depression subscales scores ≥8 will be randomized into ATT (n = 32) or WL control (n = 32). After 6-8 weeks, WL patients will be allocated to ATT. Participants will be evaluated pre-, mid- and post-treatment, and at 6-months follow-up using changes in HADS as primary outcome. Secondary outcomes will be changes in psychiatric disorders, rumination, worry, type D-personality, metacognitions, insomnia, quality of life, and C-Reactive protein (CRP). To our knowledge, this will be the first WL-controlled randomized study testing the effectiveness of group-based ATT as treatment of symptoms of anxiety and depression in CHD patients. It will also explore correlations between changes in psychological distress and CRP. A qualitative analysis will reveal patients' experience with ATT including processes that may facilitate or serve as barriers to effectiveness. Recruitment into the study commenced in December 2022. The Regional Committees for Medical and Health Research Ethics (REK), Norway has granted approval for the study (ID 52002). The trial results will be published in peer-reviewed journals. According to Norwegian legislation, the Norwegian Data Protection Authority, and the Committee of Ethics, we are not allowed to share original study data publicly. ClinicalTrials.gov, identifier NCT05621408 pre-inclusion. There were no significant changes of methods or outcomes after study start.

Attentional Bias Modification Training Based on Virtual Reality and Eye Tracking in Anorexia Nervosa Patients.

Anorexia nervosa (AN) patients exhibit attentional bias (AB) related to the body, which is the tendency to pay greater attention to weight-related body areas compared to non-weight-related ones. This phenomenon has been linked to elevated levels of body dissatisfaction (BD) and may potentially reduce the effectiveness of body exposure therapy. The purpose of this pilot study is to assess the efficacy of a single session of a new body-related AB modification task (ABMT) that combines virtual reality with eye tracking in patients with AN. The goals of the ABMT are to reduce body-related AB by balancing attention between weight and non-weight-related body areas and to reduce BD levels. Twenty-three adolescent patients with AN were embodied in a virtual avatar and immersed in a virtual environment where they completed the ABMT. Body-related AB measures and BD levels were assessed before and after the training. A paired samples t-test showed statistically significant differences between pre-assessment and post-assessment; the complete fixation time on weight-related body parts was reduced and BD levels decreased. The initial evidence of the efficacy of this ABMT has important clinical implications, since AB and BD are considered risk factors for developing and maintaining eating disorder symptomatology among patients with AN.

Long-term outcomes of an internet-delivered cognitive enhancement intervention targeting residual cognitive deficits after major depressive disorder: a 2-year follow-up of an open trial.

Cognitive deficits are common and disabling residual symptoms following major depressive disorder (MDD) and are related to increased risk of relapse. Residual cognitive deficits should thus be considered an important target for treatment. However, few have reported long-term outcomes of interventions targeting residual cognitive deficits. This study aimed to (1) investigate change between pre-treatment and 2-year follow-up assessments in cognitive deficits, rumination, and symptoms of MDD after an internet-delivered intervention targeting residual cognitive deficits; (2) to investigate stability in outcomes between 6-month and 2-year follow-up assessments; (3) to report the number of participants' experiencing a new episode of MDD in the follow-up period; and (4) to investigate differences in outcomes between those who experienced a new episode of MDD and those who did not. A total of 43 partly remitted adults were included to test a guided internet-delivered intervention, which consisted of 10 modules involving psychoeducation, cognitive strategies, and attention training. Participants were assessed at pre-treatment, post-treatment, after 6-months, and after 2-years, with measures assessing self-reported residual cognitive deficits, rumination, symptoms of MDD and relapse. Overall, 32 participants completed the 2-year follow-up assessment. Between the pre-treatment and 2-year follow-up assessments, there was a reduction in cognitive deficits and rumination, while there was an increase in symptoms of MDD. Cognitive deficits were stable between the 6-month and the 2-year follow-up, while there was an increase in rumination and symptoms of MDD. Thirteen of 32 participants reported a new episode of MDD during the follow-up period. The relapse group reported longer duration of MDD at pre-treatment and showed a difference in all outcomes after 2 years compared to the no-relapse group. The no-relapse group showed improvement in MDD symptoms at post-treatment, while the relapse group did not. Delivering cognitive enhancement interventions over the internet is potentially related to stable improvements in residual cognitive deficits. The effects on rumination and symptoms of MDD are less certain. Lack of improvement in MDD symptoms after the intervention period should be investigated as an indicator of relapse. Results should be interpreted with caution due to the lack of control group and sample size.

Virtual reality-based music attention training for acquired brain injury: A protocol for randomized cross-over trial.

Attention training is the primary step in the rehabilitation for patients with acquired brain injury (ABI). While active music performance has been reported to aid neural and functional recovery, its efficacy for patients with ABI remains uncertain due to methodological concerns. The purpose of the study is to develop a virtual reality-based music attention training (VR-MAT), which utilizes a visually guided, bilateral drumming in an immersive environment to train attention and executive functions. We also aims to examine the feasibility and effectiveness of the VR-MAT with a small sample size of participants (3-60 months after ABI, N = 20 approximately). Participants will be randomly assigned to either a waitlist control or music group, in which VR-MAT will take place five times weekly over 4 weeks (randomized crossover design). The evaluation of VR-MAT performance will include accuracy and response time in music responses. Neurocognitive outcome measures will be administered to quantify pre-post changes in attention, working memory, and executive functions. Additionally, functional near-infrared spectroscopy will be employed to explore the relationships between musical behavior, neurocognitive function, and neurophysiological responses.

Neural Mechanisms Related to the Enhanced Auditory Selective Attention Following Neurofeedback Training: Focusing on Cortical Oscillations.

Selective attention can be a useful tactic for speech-in-noise (SiN) interpretation as it strengthens cortical responses to attended sensory inputs while suppressing others. This cortical process is referred to as attentional modulation. Our earlier study showed that a neurofeedback training paradigm was effective for improving the attentional modulation of cortical auditory evoked responses. However, it was unclear how such neurofeedback training improved attentional modulation. This paper attempts to unveil what neural mechanisms underlie strengthened auditory selective attention during the neurofeedback training paradigm. Our EEG time-frequency analysis found that, when spatial auditory attention was focused, a fronto-parietal brain network was activated. Additionally, the neurofeedback training increased beta oscillation, which may imply top-down processing was used to anticipate the sound to be attended selectively with prior information. When the subjects were attending to the sound from the right, they exhibited more alpha oscillation in the right parietal cortex during the final session compared to the first, indicating improved spatial inhibitory processing to suppress sounds from the left. After the four-week training period, the temporal cortex exhibited improved attentional modulation of beta oscillation. This suggests strengthened neural activity to predict the target. Moreover, there was an improvement in the strength of attentional modulation on cortical evoked responses to sounds. The Placebo Group, who experienced similar attention training with the exception that feedback was based simply on behavioral accuracy, did not experience these training effects. These findings demonstrate how neurofeedback training effectively improves the neural mechanisms underlying auditory selective attention.

Virtual reality-based attention training in patients with neurological damage: A pilot study

ABSTRACT Virtual Reality has been shown to be a valid tool to assess cognitive functions in an ecologically valid way. However, evidence regarding its effectiveness as a treatment option for cognitive rehabilitation has been limited. Furthermore, its potential to facilitate the transfer of training effects to patients’ everyday life is still poorly studied. This study aimed to evaluate the efficacy of a VR-based attention training and its transfer to attentional functioning in everyday life. Nineteen inpatients with neurological disorders and attentional deficits underwent daily 30-min VR training sessions over a two-week period. Attentional functions were assessed before and after the training period using several computerized tests, two self-constructed behavioral tasks, and a questionnaire assessing patients’ subjective attentional functioning. Pre-Post-analyses show significant decreases in reaction times in the computerized alertness and selective attention tests. Transfer to the behavioral tasks and self-report data could not be observed. Despite the specificity of the changes, confounding effects cannot entirely be ruled out due to the lack of a control group. Results suggest that training was effective in improving attentional functioning along neuropsychological measures, but did not elicit transfer to an ecologically valid or subjective level. Implications for the future development of VR interventions are discussed.

Gamified Web-Delivered Attentional Bias Modification Training for Adults with Chronic Pain: A Randomised, Double-Blind, Placebo-Controlled Trial (Preprint)

Gamified Web-Delivered Attentional Bias Modification Training for Adults with Chronic Pain: A Randomised, Double-Blind, Placebo-Controlled Trial (Preprint)

Adaptive P300-Based Brain-Computer Interface for Attention Training: Protocol for a Randomized Controlled Trial.

The number of people with cognitive deficits and diseases, such as stroke, dementia, or attention-deficit/hyperactivity disorder, is rising due to an aging, or in the case of attention-deficit/hyperactivity disorder, a growing population. Neurofeedback training using brain-computer interfaces is emerging as a means of easy-to-use and noninvasive cognitive training and rehabilitation. A novel application of neurofeedback training using a P300-based brain-computer interface has previously shown potential to improve attention in healthy adults. This study aims to accelerate attention training using iterative learning control to optimize the task difficulty in an adaptive P300 speller task. Furthermore, we hope to replicate the results of a previous study using a P300 speller for attention training, as a benchmark comparison. In addition, the effectiveness of personalizing the task difficulty during training will be compared to a nonpersonalized task difficulty adaptation. In this single-blind, parallel, 3-arm randomized controlled trial, 45 healthy adults will be recruited and randomly assigned to the experimental group or 1 of 2 control groups. This study involves a single training session, where participants receive neurofeedback training through a P300 speller task. During this training, the task's difficulty is progressively increased, which makes it more difficult for the participants to maintain their performance. This encourages the participants to improve their focus. Task difficulty is either adapted based on the participants' performance (in the experimental group and control group 1) or chosen randomly (in control group 2). Changes in brain patterns before and after training will be analyzed to study the effectiveness of the different approaches. Participants will complete a random dot motion task before and after the training so that any transfer effects of the training to other cognitive tasks can be evaluated. Questionnaires will be used to estimate the participants' fatigue and compare the perceived workload of the training between groups. This study has been approved by the Maynooth University Ethics Committee (BSRESC-2022-2474456) and is registered on ClinicalTrials.gov (NCT05576649). Participant recruitment and data collection began in October 2022, and we expect to publish the results in 2023. This study aims to accelerate attention training using iterative learning control in an adaptive P300 speller task, making it a more attractive training option for individuals with cognitive deficits due to its ease of use and speed. The successful replication of the results from the previous study, which used a P300 speller for attention training, would provide further evidence to support the effectiveness of this training tool. ClinicalTrials.gov NCT05576649; https://clinicaltrials.gov/ct2/show/NCT05576649. DERR1-10.2196/46135.

Effects of SRSD writing interventions in grade 3: examining the added value of attention vs. transcription training components

Children’s writing depends on strategic and self-regulated behaviors as well as on attention and transcription skills. Despite the well-established effectiveness of interventions based on the self-regulated strategy development model (SRSD), little is known about their effects when combined with the explicit teaching and systematic training of attention or transcription. This study compared the effects of SRSD interventions that taught students a planning strategy and how to use it compose text, with additional components of attention (via focused meditation exercises) or transcription (via copy and alphabet exercises). We used a quasi-experimental design with three groups of Portuguese third graders: SRSD + attention (n = 42), SRSD + transcription (n = 37), and wait list (n = 34). Measures of transcription, executive functions, motivation, written production, and academic achievement were collected before and after the interventions. Analyses of co-variance revealed that the SRSD + transcription group surpassed the others in terms of handwriting fluency as well as in terms of spelling accuracy, though only among the weakest spellers at pretest. Moreover, in comparison to the wait list group, both SRSD groups showed better planning skills, wrote more complete texts, and displayed better executive functioning at posttest. After the intervention, though the SRSD + transcription group produced better texts than their peers, the SRSD + attention group achieved better academic performance, but only among those with poorer grades at pretest. Regarding motivation, we only found an effect of the SRSD + transcription intervention on self-efficacy for ideation, among students with lower self-efficacy at pretest. These findings suggest the importance of articulating writing and self-regulation strategies with attention and transcription training in early schooling.

Multi-path decoder U-Net: A weakly trained real-time segmentation network for object detection and localization in ultrasound scans.

Detecting and localizing an anatomical structure of interest within the field of view of an ultrasound scan is an essential step in many diagnostic and therapeutic procedures. However, ultrasound scans suffer from high levels of variabilities across sonographers and patients, making it challenging for sonographers to accurately identify and locate these structures without extensive experience. Segmentation-based convolutional neural networks (CNNs) have been proposed as a solution to assist sonographers in this task. Despite their accuracy, these networks require pixel-wise annotations for training; an expensive and labor-intensive operation that requires the expertise of an experienced practitioner to identify the precise outline of the structures of interest. This complicates, delays, and increases the cost of network training and deployment. To address this problem, we propose a multi-path decoder U-Net architecture that is trained on bounding box segmentation maps; not requiring pixel-wise annotations. We show that the network can be trained on small training sets, which is the case in medical imaging datasets; reducing the cost and time needed for deployment and use in clinical settings. The multi-path decoder design allows for better training of deeper layers and earlier attention to the target anatomical structures of interest. This architecture offers up to a 7% relative improvement compared to the U-Net architecture in localization and detection performance, with an increase of only 0.75% in the number of parameters. Its performance is on par with, or slightly better than, the more computationally expensive U-Net++, which has 20% more parameters; making the proposed architecture a more computationally efficient alternative for real-time object detection and localization in ultrasound scans.