Search

This study evaluates the performance of the MaskFormer model for segmenting and classifying breast lesions using ultrasound images, addressing ultrasound's limitations. Ultrasound used for breast cancer detection faces challenges like low image contrast and difficulty in the detection of small or multiple lesions, further complicated by variability based on operator skill. Initial experiments with U-Net and other CNN-based models revealed constraints, such as early plateauing in model loss, which indicated suboptimal learning and performance. In contrast, MaskFormer demonstrated continuous improvement, achieving higher precision in breast lesion segmentation and significantly reducing both false positives and false negatives. Comparative analysis showed MaskFormer's superior performance, with the highest precision and recall rates for malignant lesions and an overall mean average precision (mAP) of 0.943. The model's ability to detect a diverse range of breast lesions, including those potentially missed by the human eye, especially by less experienced practitioners, underscores its potential. These findings suggest that integrating AI models like MaskFormer could greatly enhance ultrasound performance for breast cancer detection, providing reliable, operator-independent image analysis and potentially improving patient outcomes on a global scale. [ABSTRACT FROM AUTHOR]

For successful human–robot collaboration, it is crucial to establish and sustain quality interaction between humans and robots, making it essential to facilitate human–robot interaction (HRI) effectively. The evolution of robot intelligence now enables robots to take a proactive role in initiating and sustaining HRI, thereby allowing humans to concentrate more on their primary tasks. In this paper, we introduce a system known as the Robot-Facilitated Interaction System (RFIS), where mobile robots are employed to perform identification, tracking, re-identification, and gesture recognition in an integrated framework to ensure anytime readiness for HRI. We implemented the RFIS on an autonomous mobile robot used for transporting a patient, to demonstrate proactive, real-time, and user-friendly interaction with a caretaker involved in monitoring and nursing the patient. In the implementation, we focused on the efficient and robust integration of various interaction facilitation modules within a real-time HRI system that operates in an edge computing environment. Experimental results show that the RFIS, as a comprehensive system integrating caretaker recognition, tracking, re-identification, and gesture recognition, can provide an overall high quality of interaction in HRI facilitation with average accuracies exceeding 90% during real-time operations at 5 FPS. [ABSTRACT FROM AUTHOR]

Sound localization is a crucial aspect of human auditory perception. VR (virtual reality) technologies provide immersive audio platforms that allow human listeners to experience natural sounds based on their ability to localize sound. However, the simulations of sound generated by these platforms, which are based on the general head-related transfer function (HRTF), often lack accuracy in terms of individual sound perception and localization due to significant individual differences in this function. In this study, we aimed to investigate the disparities between the perceived locations of sound sources by users and the locations generated by the platform. Our goal was to determine if it is possible to train users to adapt to the platform-generated sound sources. We utilized the Microsoft HoloLens 2 virtual platform and collected data from 12 subjects based on six separate training sessions arranged in 2 weeks. We employed three modes of training to assess their effects on sound localization, in particular for studying the impacts of multimodal error, visual, and sound guidance in combination with kinesthetic/postural guidance, on the effectiveness of the training. We analyzed the collected data in terms of the training effect between pre- and post-sessions as well as the retention effect between two separate sessions based on subject-wise paired statistics. Our findings indicate that, as far as the training effect between pre- and post-sessions is concerned, the effect is proven to be statistically significant, in particular in the case wherein kinesthetic/postural guidance is mixed with visual and sound guidance. Conversely, visual error guidance alone was found to be largely ineffective. On the other hand, as far as the retention effect between two separate sessions is concerned, we could not find any meaningful statistical implication on the effect for all three error guidance modes out of the 2-week session of training. These findings can contribute to the improvement of VR technologies by ensuring they are designed to optimize human sound localization abilities. [ABSTRACT FROM AUTHOR]

In this paper, we experimentally investigate how the 3D sound localization capabilities of the blind can improve through perceptual training. To this end, we develop a novel perceptual training method with sound-guided feedback and kinesthetic assistance to evaluate its effectiveness compared to conventional training methods. In perceptual training, we exclude visual perception by blindfolding the subjects to apply the proposed method to the visually impaired. Subjects used a specially designed pointing stick to generate a sound at the tip, indicating localization error and tip position. The proposed perceptual training aims to evaluate the training effect on 3D sound localization, including variations in azimuth, elevation, and distance. The six days of training based on six subjects resulted in the following outcomes: (1) In general, accuracy in full 3D sound localization can be improved based on training. (2) Training based on relative error feedback is more effective than absolute error feedback. (3) Subjects tend to underestimate distance when the sound source is near, less than 1000 mm, or larger than 15° to the left, and overestimate the elevation when the sound source is near or in the center, and within ±15° in azimuth estimations. [ABSTRACT FROM AUTHOR]

Liver ultrasound (US) plays a critical role in diagnosing liver diseases. However, it is often difficult for examiners to accurately identify the liver segments captured in US images due to patient variability and the complexity of US images. Our study aim is automatic, real-time recognition of standardized US scans coordinated with reference liver segments to guide examiners. We propose a novel deep hierarchical architecture for classifying liver US images into 11 standardized US scans, which has yet to be properly established due to excessive variability and complexity. We address this problem based on a hierarchical classification of 11 US scans with different features applied to individual hierarchies as well as a novel feature space proximity analysis for handling ambiguous US images. Experiments were performed using US image datasets obtained from a hospital setting. To evaluate the performance under patient variability, we separated the training and testing datasets into distinct patient groups. The experimental results show that the proposed method achieved an F1-score of more than 93%, which is more than sufficient for a tool to guide examiners. The superior performance of the proposed hierarchical architecture was demonstrated by comparing its performance with that of non-hierarchical architecture. [ABSTRACT FROM AUTHOR]

A method of location fingerprinting based on theWi-Fi received signal strength (RSS) in an indoor environment is presented. The method aims to overcome the RSS instability due to varying channel disturbances in time by introducing the concept of invariant RSS statistics. The invariant RSS statistics represent here the RSS distributions collected at individual calibration locations under minimal random spatiotemporal disturbances in time. The invariant RSS statistics thus collected serve as the reference pattern classes for fingerprinting. Fingerprinting is carried out at an unknown location by identifying the reference pattern class that maximally supports the spontaneous RSS sensed from individual Wi-Fi sources. A design guideline is also presented as a rule of thumb for estimating the number of Wi-Fi signal sources required to be available for any given number of calibration locations under a certain level of random spatiotemporal disturbances. Experimental results show that the proposed method not only provides 17% higher success rate than conventional ones but also removes the need for recalibration. Furthermore, the resolution is shown finer by 40% with the execution time more than an order of magnitude faster than the conventional methods. These results are also backed up by theoretical analysis. [ABSTRACT FROM AUTHOR]

Deep learning approaches to estimating full 3D orientations of objects, in addition to object classes, are limited in their accuracies, due to the difficulty in learning the continuous nature of three-axis orientation variations by regression or classification with sufficient generalization. This paper presents a novel progressive deep learning framework, herein referred to as 3D POCO Net, that offers high accuracy in estimating orientations about three rotational axes yet with efficiency in network complexity. The proposed 3D POCO Net is configured, using four PointNet-based networks for independently representing the object class and three individual axes of rotations. The four independent networks are linked by in-between association subnetworks that are trained to progressively map the global features learned by individual networks one after another for fine-tuning the independent networks. In 3D POCO Net, high accuracy is achieved by combining a high precision classification based on a large number of orientation classes with a regression based on a weighted sum of classification outputs, while high efficiency is maintained by a progressive framework by which a large number of orientation classes are grouped into independent networks linked by association subnetworks. We implemented 3D POCO Net for full three-axis orientation variations and trained it with about 146 million orientation variations augmented from the ModelNet10 dataset. The testing results show that we can achieve an orientation regression error of about 2.5° with about 90% accuracy in object classification for general three-axis orientation estimation and object classification. Furthermore, we demonstrate that a pre-trained 3D POCO Net can serve as an orientation representation platform based on which orientations as well as object classes of partial point clouds from occluded objects are learned in the form of transfer learning. [ABSTRACT FROM AUTHOR]

Accurately estimating the current state of local traffic scenes is one of the key problems in the development of software components for automated vehicles. In addition to details on free space and drivability, static and dynamic traffic participants and information on the semantics may also be included in the desired representation. Multi-layer grid maps allow the inclusion of all of this information in a common representation. However, most existing grid mapping approaches only process range sensor measurements such as Lidar and Radar and solely model occupancy without semantic states. In order to add sensor redundancy and diversity, it is desired to add vision-based sensor setups in a common grid map representation. In this work, we present a semantic evidential grid mapping pipeline, including estimates for eight semantic classes, that is designed for straightforward fusion with range sensor data. Unlike other publications, our representation explicitly models uncertainties in the evidential model. We present results of our grid mapping pipeline based on a monocular vision setup and a stereo vision setup. Our mapping results are accurate and dense mapping due to the incorporation of a disparity- or depth-based ground surface estimation in the inverse perspective mapping. We conclude this paper by providing a detailed quantitative evaluation based on real traffic scenarios in the KITTI odometry benchmark dataset and demonstrating the advantages compared to other semantic grid mapping approaches. [ABSTRACT FROM AUTHOR]

Tracking drivers' eyes and gazes is a topic of great interest in the research of advanced driving assistance systems (ADAS). It is especially a matter of serious discussion among the road safety researchers' community, as visual distraction is considered among the major causes of road accidents. In this paper, techniques for eye and gaze tracking are first comprehensively reviewed while discussing their major categories. The advantages and limitations of each category are explained with respect to their requirements and practical uses. In another section of the paper, the applications of eyes and gaze tracking systems in ADAS are discussed. The process of acquisition of driver's eyes and gaze data and the algorithms used to process this data are explained. It is explained how the data related to a driver's eyes and gaze can be used in ADAS to reduce the losses associated with road accidents occurring due to visual distraction of the driver. A discussion on the required features of current and future eye and gaze trackers is also presented. [ABSTRACT FROM AUTHOR]

Improving a vehicle driver's performance decreases the damage caused by, and chances of, road accidents. In recent decades, engineers and researchers have proposed several strategies to model and improve driving monitoring and assistance systems (DMAS). This work presents a comprehensive survey of the literature related to driving processes, the main reasons for road accidents, the methods of their early detection, and state-of-the-art strategies developed to assist drivers for a safe and comfortable driving experience. The studies focused on the three main elements of the driving process, viz. driver, vehicle, and driving environment are analytically reviewed in this work, and a comprehensive framework of DMAS, major research areas, and their interaction is explored. A well-designed DMAS improves the driving experience by continuously monitoring the critical parameters associated with the driver, vehicle, and surroundings by acquiring and processing the data obtained from multiple sensors. A discussion on the challenges associated with the current and future DMAS and their potential solutions is also presented. [ABSTRACT FROM AUTHOR]

An accurate and cost-effective micro sun sensor based on the extraction of the sun vector using a phenomenon called the "black sun" is presented. Unlike conventional image-based sun sensors where there is difficulty in accurately detecting the sun center, the black sun effect allows the sun center to be accurately extracted even with the sun image appearing irregular and noisy due to glare. This allows the proposed micro sun sensor to achieve high accuracy even when a 1 mm × 1 mm CMOS image sensor with a resolution of 250 × 250 pixels is used. The proposed micro sun sensor is implemented in two application modes: (1) a stationary mode targeted at tracking the sun for heliostats or solar panels with a fixed pose of single image sensor of 1 mm × 1 mm × 1.74 mm in size and (2) a non-stationary mode targeted at determining the orientation of moving platforms with six sensors on the platform, which is configured in an icosahedron geometry of 23 mm × 23 mm × 12 mm in size. For the stationary mode, we obtained an accuracy of 0.013° by applying Kalman filter to the sun sensor measurement for a particular sensor orientation. For the non-stationary mode, we obtained an improved accuracy of 0.05° by fusing the measurements from three sun sensors available at any instant of time. Furthermore, experiments indicate that the black sun effect makes the precision of sun vector extraction independent of the sun location captured on the image plane. [ABSTRACT FROM AUTHOR]

A general framework of data fusion is presented based on projecting the probability distribution of true states and measurements around the predicted states and actual measurements onto the constraint manifold. The constraint manifold represents the constraints to be satisfied among true states and measurements, which is defined in the extended space with all the redundant sources of data such as state predictions and measurements considered as independent variables. By the general framework, we mean that it is able to fuse any correlated data sources while directly incorporating constraints and identifying inconsistent data without any prior information. The proposed method, referred to here as the Covariance Projection (CP) method, provides an unbiased and optimal solution in the sense of minimum mean square error (MMSE), if the projection is based on the minimum weighted distance on the constraint manifold. The proposed method not only offers a generalization of the conventional formula for handling constraints and data inconsistency, but also provides a new insight into data fusion in terms of a geometric-algebraic point of view. Simulation results are provided to show the effectiveness of the proposed method in handling constraints and data inconsistency. [ABSTRACT FROM AUTHOR]