- 标题
- 摘要
- 关键词
- 实验方案
- 产品
-
Multi-label chest X-ray image classification via category-wise residual attention learning
摘要： This paper considers the problem of multi-label thorax disease classification on chest X-ray images. Identifying one or more pathologies from a chest X-ray image is often hindered by the pathologies unrelated to the targets. In this paper, we address the above problem by proposing a category-wise residual attention learning (CRAL) framework. CRAL predicts the presence of multiple pathologies in a class-specific attentive view. It aims to suppress the obstacles of irrelevant classes by endowing small weights to the corresponding feature representation. Meanwhile, the relevant features would be strengthened by assigning larger weights. Specifically, the proposed framework consists of two modules: feature embedding module and attention learning module. The feature embedding module learns high-level features with a convolutional neural network (CNN) while the attention learning module focuses on exploring the assignment scheme of different categories. The attention module can be flexibly integrated into any feature embedding networks with end-to-end training. The comprehensive experiments are conducted on the Chest X-ray14 dataset. CRAL yields the average AUC score of 0.816 which is a new state of the art.
关键词： Image classification,Chest X-ray,Convolutional neural network,Residual attention
更新于2025-09-23 15:23:52
-
The Role of Exercise-Induced Arousal and Exposure to Blue-Enriched Lighting on Vigilance
摘要： It is currently assumed that exposure to an artificial blue-enriched light enhances human alertness and task performance, but recent research has suggested that behavioral effects are influenced by the basal state of arousal. Here, we tested whether the effect of blue-enriched lighting on vigilance performance depends on participants’ arousal level. Twenty-four participants completed four sessions (blue-enriched vs. dim light × low vs. high arousal) at 10 pm on four consecutive days, following a repeated-measures design. Participants’ arousal was manipulated parametrically through the execution of a cycling task at two intensities (low vs. moderate), and was checked by monitoring their heart rate. On each session, distal and proximal skin temperatures were recorded as a neuroergonomic index of vigilance, while participants performed a 20-min psychomotor vigilance task (PVT) under either blue-enriched light or dim light conditions. The Positive and Negative Affect Schedule (PANAS), Karolinska Sleepiness Scale (KSS), and Borg Rating of Perceived Exertion Scale (RPE) were used to measure subjective psychological state. The results showed that the exercise-induced manipulation of arousal produced robust alerting effects in most measures, while the lighting manipulation only attenuated subjective sleepiness and enhanced positive affect, but it did not influence behavior or physiology. Acute exposure to a blue-enriched light was practically ineffective when the arousal level was over baseline. The present research favored the use of acute physical exercise over acute exposure to blue-enriched lighting in order to boost humans’ levels of alertness when necessary, as in work settings where maintaining optimal attention is difficult (shift work, night-work, vigilance tasks) and necessary to prevent human error and accidents.
关键词： circadian,light,attention,heart rate,temperature,ergonomics,alertness,physical activity
更新于2025-09-23 15:22:29
-
Enhanced spatial focusing increases feature-based selection in unattended locations
摘要： Attention is a multifaceted phenomenon, which operates on features (e.g., colour or motion) and over space. A fundamental question is whether the attentional selection of features is confined to the spatially-attended location or operates independently across the entire visual field (global feature-based attention, GFBA). Studies providing evidence for GFBA often employ feature probes presented at spatially unattended locations, which elicit enhanced brain responses when they match a currently-attended target feature. However, the validity of this interpretation relies on consistent spatial focusing onto the target. If the probe were to temporarily attract spatial attention, the reported effects could reflect transient spatial selection processes, rather than GFBA. Here, using magnetoencephalographic recordings (MEG) in humans, we manipulate the strength and consistency of spatial focusing to the target by increasing the target discrimination difficulty (Experiment 1), and by demarcating the upcoming target’s location with a placeholder (Experiment 2), to see if GFBA effects are preserved. We observe that motivating stronger spatial focusing to the target did not diminish the effects of GFBA. Instead, aiding spatial pre-focusing with a placeholder enhanced the feature response at unattended locations. Our findings confirm that feature selection effects measured with spatially-unattended probes reflect a true location-independent neural bias.
关键词： Global feature-based attention,Attention,Spatial focusing,Magnetoencephalographic recordings,Visual processing
更新于2025-09-23 15:21:21
-
Multiscale Visual Attention Networks for Object Detection in VHR Remote Sensing Images
摘要： Object detection plays an active role in remote sensing applications. Recently, deep convolutional neural network models have been applied to automatically extract features, generate region proposals, and predict corresponding object class. However, these models face new challenges in VHR remote sensing images due to the orientation and scale variations and the cluttered background. In this letter, we propose an end-to-end multiscale visual attention networks (MS-VANs) method. We use skip-connected encoder–decoder model to extract multiscale features from a full-size image. For feature maps in each scale, we learn a visual attention network, which is followed by a classification branch and a regression branch, so as to highlight the features from object region and suppress the cluttered background. We train the MS-VANs model by a hybrid loss function which is a weighted sum of attention loss, classification loss, and regression loss. Experiments on a combined data set consisting of Dataset for Object Detection in Aerial Images and NWPU VHR-10 show that the proposed method outperforms several state-of-the-art approaches.
关键词： object detection,VHR remote sensing image,visual attention,Multiscale feature
更新于2025-09-23 15:21:21
-
Airline Head-Up Display Systems: Human Factors Considerations
摘要： As the key source of information for pilots, the human visual system has necessarily driven much of the evolution in cockpit technology. In contrast to the complicated, gauge-based systems of the past, the electronic flight displays of today’s modern airliners are testament to advances in human factors engineering. The next step in flight instrumentation, although already used for some 50 years in the military, is just beginning to emerge in civil transport aircraft. Head-up displays (HUDs) allow pilots to see key flight instrumentation while viewing the outside world. The need to look down at the flight instruments is removed by the HUD, resulting in increased situational awareness and greater precision in aircraft control. While the head-up display is a welcome development, as with many instances of human-technology interface, the benefits provided by HUDs are not without potential drawbacks. The present report provides a thorough grounding in key areas of importance to the development and operation of HUD systems, including the human visual system, the history and development of conventional cockpit instrumentation, and a summary of the technical aspects and human factors considerations relating to HUD systems. This basis is built upon with a more practical assessment of the HUD system used in a modern airliner, the Boeing 787.
关键词： aviation,Boeing 787,Attention capture,Head-up display,Airline,Visual attention,Human factors,Cognitive tunneling,HUD
更新于2025-09-23 15:21:01
-
[Lecture Notes in Computer Science] Pattern Recognition and Computer Vision Volume 11257 (First Chinese Conference, PRCV 2018, Guangzhou, China, November 23-26, 2018, Proceedings, Part II) || Prohibited Item Detection in Airport X-Ray Security Images via Attention Mechanism Based CNN
摘要： Automation of security inspections is crucial for improving the efficiency and reducing security risks. In this paper, we focus on automatically recognizing and localizing prohibited items in airport X-ray security images. A top-down attention mechanism is applied to enhance a CNN classifier to additionally locate the prohibited items. We introduce a high-level semantic feedback loop to map the targets semantic signal to the input X-ray image space for generating task-specic attention maps. And the attention maps indicate the location and general outline of prohibited items in the input images. Furthermore, to obtain more accurate location information, we combine the lateral inhibition and contrastive attention to suppress noise and non-target interference in attention maps. The experiments on the GDX-ray image dataset have demonstrated the efficiency and stability of the proposed scheme in both single target detection and multi-target detection.
关键词： CNN,Detection,Prohibited item,Attention
更新于2025-09-23 15:21:01
-
Rating the Intensity of a Laser Stimulus, but Not Attending to Changes in Its Location or Intensity Modulates the Laser-Evoked Cortical Activity
摘要： Top-down attention towards nociceptive stimuli can be modulated by asking participants to pay attention to specific features of a stimulus, or to provide a rating about its intensity/unpleasantness. Whether and how these different top-down processes may lead to different modulations of the cortical response to nociceptive stimuli remains an open question. We recorded electroencephalographic (EEG) responses to brief nociceptive laser stimuli in 24 healthy participants while they performed a task in which they had to compare two subsequent stimuli on their Spatial location (Location task) or Intensity (Intensity Task). In two additional blocks (Location + Ratings, and Intensity + Ratings) participants had to further provide a rating of the perceived intensity of the stimulus. Such a design allowed us to investigate whether focusing on spatial or intensity features of a nociceptive stimulus and rating its intensity would exert different effects on the EEG responses. We did not find statistical evidence for an effect on the signal while participants were focusing on different features of the signal. We only observed a significant cluster difference in frontoparietal leads at approximately 300–500 ms post-stimulus between the magnitude of the signal in the Intensity and Intensity + Rating conditions, with a less negative response in the Intensity + Rating condition in frontal electrodes, and a less positive amplitude in parietal leads. We speculatively propose that activity in those electrodes and time window reflects magnitude estimation processes. Moreover, the smaller frontal amplitude in the Intensity + Rating condition can be explained by greater working memory engagement known to reduce the magnitude of the EEG signal. We conclude that different top-down attentional processes modulate responses to nociceptive laser stimuli at different electrodes and time windows depending on the underlying processes that are engaged.
关键词： top-down modulation,pain,ratings,attention,EEG,laser evoked potentials (LEPs)
更新于2025-09-23 15:19:57
-
Aging effects on prefrontal cortex oxygenation in a posture-cognition dual-task: an fNIRS pilot study
摘要： Background: The aging process alters upright posture and locomotion control from an automatically processed to a more cortically controlled one. The present study investigated a postural-cognitive dual-task paradigm in young and older adults using functional Near-Infrared Spectroscopy (fNIRS). Methods: Twenty healthy participants (10 older adults 72 ± 3 y, 10 young adults 23 ± 3 y) performed a cognitive (serial subtractions) and a postural task (tandem stance) as single-tasks (ST) and concurrently as a dual-task (DT) while the oxygenation levels of the dorsolateral prefrontal cortex (DLPFC) were measured. Results: In the cognitive task, young adults performed better than older adults in both conditions (ST and DT) and could further increase the number of correct answers from ST to DT (all ps ≤ 0.027) while no change was found for older adults. No significant effects were found for the postural performance. Cerebral oxygenation values (O2Hb) increased significantly from baseline to the postural ST (p = 0.033), and from baseline to the DT (p = 0.031) whereas no changes were found in deoxygenated hemoglobin (HHb). Finally, the perceived exertion differed between all conditions (p ≤ 0.003) except for the postural ST and the DT (p = 0.204). Conclusions: There was a general lack of age-related changes except the better cognitive performance under motor-cognitive conditions in young compared to older adults. However, the current results point out that DLPFC is influenced more strongly by postural than cognitive load. Future studies should assess the different modalities of cognitive as well as postural load.
关键词： Attention,Functional near-infrared spectroscopy (fNIRS),Postural control,Aging,Dual-tasking,Executive control,Balance,Elderly
更新于2025-09-19 17:15:36
-
Validating attentive locomotion training using interactive treadmill: an fNIRS study
摘要： Background: Existing treadmill-based locomotion training, which has been used for gait function recovery, still has limitations, such as less attentive training. Interactive treadmills (ITMs) were developed to overcome these limitations, but it has not yet been verified that ITMs can make the user pay closer attention to walk training. Methods: An experimental comparison between ITMs and conventional treadmills was conducted by measuring the level of the user’s attention using functional near-infrared spectroscopy (fNIRS). To consider the effect of task complexity on the subject’s attention, we provided two (slow and fast) speed conditions for walking on both treadmills. Results: Both the cortical activity images and oxygenated hemoglobin (oxyHb) changes showed that the level of attention to walking induced by the ITM was significantly higher than that induced by the conventional treadmill. We found that the walking speed on the ITM also affected the level of attention. Conclusion: ITM-based locomotion training would be a promising solution to the limitations of existing treadmill-based locomotion training currently used to improve gait function recovery.
关键词： Gait training,Attention,Interactive treadmill (ITM),Task complexity,fNIRS,Walking speed
更新于2025-09-19 17:15:36
-
[Lecture Notes in Computer Science] Computer Vision – ECCV 2018 Workshops Volume 11133 (Munich, Germany, September 8-14, 2018, Proceedings, Part V) || Deep Residual Attention Network for Spectral Image Super-Resolution
摘要： Spectral imaging sensors often suffer from low spatial resolution, as there exists an essential tradeoff between the spectral and spatial resolutions that can be simultaneously achieved, especially when the temporal resolution needs to be retained. In this paper, we propose a novel deep residual attention network for the spatial super-resolution (SR) of spectral images. The proposed method extends the classic residual network by (1) directly using the 3D low-resolution (LR) spectral image as input instead of upsampling the 2D bandwise images separately, and (2) integrating the channel attention mechanism into the residual network. These two operations fully exploit the correlations across both the spectral and spatial dimensions and greatly promote the performance of spectral image SR. In addition, for the scenario when stereo pairs of LR spectral and high-resolution (HR) RGB measurements are available, we design a fusion framework based on the proposed network. The spatial resolution of the spectral input is enhanced in one branch, while the spectral resolution of the RGB input is enhanced in the other. These two branches are then fused together through the attention mechanism again to reconstruct the final HR spectral image, which achieves further improvement compared to using the single LR spectral input. Experimental results demonstrate the superiority of the proposed method over plain residual networks, and our method is one of the winning solutions in the PIRM 2018 Spectral Super-resolution Challenge.
关键词： Spectral image,Super-resolution,Channel attention
更新于2025-09-19 17:15:36