An Introduction to Adversarially Robust Deep Learning.

The widespread success of deep learning in solving machine learning problems has fueled its adoption in many fields, from speech recognition to drug discovery and medical imaging. However, deep learning systems are extremely fragile: imperceptibly small modifications to their input data can cause the models to produce erroneous output. It is very easy to generate such adversarial perturbations even for state-of-the-art models, yet immunization against them has proven exceptionally challenging. Despite over a decade of research on this problem, our solutions are still far from satisfactory and many open problems remain. In this work, we survey some of the most important contributions in the field of adversarial robustness. We pay particular attention to the reasons why past attempts at improving robustness have been insufficient, and we identify several promising areas for future research.

Improving Turn Movement Count Using Cooperative Feedback.

In this paper, we propose a new cooperative method that improves the accuracy of Turn Movement Count (TMC) under challenging conditions by introducing contextual observations from the surrounding areas. The proposed method focuses on the correct identification of the movements in conditions where current methods have difficulties. Existing vision-based TMC systems are limited under heavy traffic conditions. The main problems for most existing methods are occlusions between vehicles that prevent the correct detection and tracking of the vehicles through the entire intersection and the assessment of the vehicle's entry and exit points, incorrectly assigning the movement. The proposed method intends to overcome this incapability by sharing information with other observation systems located at neighboring intersections. Shared information is used in a cooperative scheme to infer the missing data, thereby improving the assessment that would otherwise not be counted or miscounted. Experimental evaluation of the system shows a clear improvement over related reference methods.

An Unsupervised Transfer Learning Framework for Visible-Thermal Pedestrian Detection.

Dual cameras with visible-thermal multispectral pairs provide both visual and thermal appearance, thereby enabling detecting pedestrians around the clock in various conditions and applications, including autonomous driving and intelligent transportation systems. However, due to the greatly varying real-world scenarios, the performance of a detector trained on a source dataset might change dramatically when evaluated on another dataset. A large amount of training data is often necessary to guarantee the detection performance in a new scenario. Typically, human annotators need to conduct the data labeling work, which is time-consuming, labor-intensive and unscalable. To overcome the problem, we propose a novel unsupervised transfer learning framework for multispectral pedestrian detection, which adapts a multispectral pedestrian detector to the target domain based on pseudo training labels. In particular, auxiliary detectors are utilized and different label fusion strategies are introduced according to the estimated environmental illumination level. Intermediate domain images are generated by translating the source images to mimic the target ones, acting as a better starting point for the parameter update of the pedestrian detector. The experimental results on the KAIST and FLIR ADAS datasets demonstrate that the proposed method achieves new state-of-the-art performance without any manual training annotations on the target data.

Shared decision-making in advance care planning for persons with dementia in nursing homes: a cross-sectional study.

BACKGROUND: Shared decision-making provides an approach to discuss advance care planning in a participative and informed manner, embodying the principles of person-centered care. A number of guided approaches to achieve shared decision-making already exist, such as the three-talk model. However, it is uncertain whether daily practice methods in nursing home wards for persons with dementia comply with the underpinnings of this model. It is also uncertain whether professionals consider shared decision-making to be important in this context, and whether they perceive themselves sufficiently competent to practice this approach frequently. METHODS: The study has a cross-sectional design, with 65 wards (46 Belgian nursing homes) participating in the study. We compared nursing home professionals' and residents' perspectives on the level of shared decision-making during advance care planning conversations with ratings from external raters. Residents and professionals rated the level of shared decision-making by means of a questionnaire, which included the topic of the conversation. External raters assessed audio recordings of the conversations. Professionals filled in an additional self-report questionnaire on the importance of shared decision-making, their competence in practicing the approach, and with what frequency. RESULTS: At ward level, professionals and residents rated the average achieved level of shared decision-making 71.53/100 (σ = 16.09) and 81.11/100 (σ = 19.18) respectively. Meanwhile, raters gave average scores of 26.97/100 (σ = 10.45). Only 23.8% of residents referred to advance care planning as the topic of the conversation. Professionals considered shared decision-making to be important (x̄=4.48/5, σ = 0.26). This result contrasted significantly with the frequency (x̄=3.48/5, σ = 0.51) and competence (x̄=3.76/5, σ = 0.27) with which these skills were practiced (P < 0.001). CONCLUSIONS: Residents with dementia are grateful when involved in discussing their care, but find it difficult to report what is discussed during these conversations. Receiving more information about advance care planning could provide them with the knowledge needed to prepare for such a conversation. External raters observe a discrepancy between the three-talk model and daily practice methods. Training programs should focus on providing professionals with better knowledge of and skills for shared decision-making. They should also promote team-based collaboration to increase the level of person-centered care in nursing home wards for persons with dementia.

An interactive ImageJ plugin for semi-automated image denoising in electron microscopy.

The recent advent of 3D in electron microscopy (EM) has allowed for detection of nanometer resolution structures. This has caused an explosion in dataset size, necessitating the development of automated workflows. Moreover, large 3D EM datasets typically require hours to days to be acquired and accelerated imaging typically results in noisy data. Advanced denoising techniques can alleviate this, but tend to be less accessible to the community due to low-level programming environments, complex parameter tuning or a computational bottleneck. We present DenoisEM: an interactive and GPU accelerated denoising plugin for ImageJ that ensures fast parameter tuning and processing through parallel computing. Experimental results show that DenoisEM is one order of magnitude faster than related software and can accelerate data acquisition by a factor of 4 without significantly affecting data quality. Lastly, we show that image denoising benefits visualization and (semi-)automated segmentation and analysis of ultrastructure in various volume EM datasets.

Improving shared decision-making in advance care planning: Implementation of a cluster randomized staff intervention in dementia care.

OBJECTIVE: Increasing staff engagement level of shared decision-making in advance care planning for persons with dementia in nursing homes. Perceived importance, competence and frequency of staff members applying shared decision-making were measured. Additionally, facilitators and barriers in the implementation process were described. METHODS: In this pretest-posttest cluster randomized trial, 311 staff members from 65 Belgian nursing home wards participated. Key components of the intervention were knowledge on shared decision-making, role-play exercises and internal policies on advance care planning. Audio recordings of advance care planning conversations between residents, families and staff were compared before and after the intervention. Participants filled in questionnaires and provided feedback. RESULTS: Wards demonstrated a higher level of shared decision-making after the intervention (p < 0.001) while time spent on the conversations did not increase. This effect persisted at 6 months follow-up (p < 0.001). Participants perceived shared decision-making as more important (p = 0.031) and felt more competent (p = 0.010), though frequency of use did not change (p = 0.201). High staff turnover and difficult co-operation with GP's were barriers. CONCLUSION: Nursing home staff benefits from this training in shared decision-making. PRACTICE IMPLICATIONS: Learning shared decision-making in advance care planning for persons with dementia is possible and sustainable in the time-constricted context of nursing homes.

'We DECide optimized' - training nursing home staff in shared decision-making skills for advance care planning conversations in dementia care: protocol of a pretest-posttest cluster randomized trial.

BACKGROUND: Due to the gradual loss of function, it is crucial for persons with dementia to discuss advance care planning in due course. However, nursing home staff, residents and their families feel uncomfortable to start this type of conversation, resulting in unknown (care) preferences. 'We DECide optimized' will provide tools to nursing home staff for discussing advance care planning. The primary objective is to enhance the level of shared decision-making in advance care planning conversations. We hypothesize that the training will enhance the perception of the importance, competence and frequency in which participants engage in advance care planning conversations. The secondary objective is to assess barriers and facilitators in the implementation of advance care planning policies at the ward level. METHODS: 'We DECide optimized' will consist of two four-hour workshops and a homework assignment between sessions. Training components will include information on advance care planning and shared decision-making, role-play exercises and group discussions on implementation barriers at the ward level. Participating wards will receive supporting materials to stimulate residents and their families to initiate conversations. The study uses a cluster randomized controlled design, with 65 Flemish nursing home wards taking part (311 staff members). Data will be collected through a pretest-posttest model, with measurements up to 9 months after training. The RE-AIM framework will be used to evaluate the effectiveness of the implementation. Quantitative and qualitative data at the clinical, organizational and resident level will be collected. DISCUSSION: This study describes a hands-on, in-depth and multi-level training approach to improve shared decision-making in advance care planning conversations. By providing tools to ward staff, engaging the management and informing residents and their families, 'We DECide optimized' aims to decrease evidence-based barriers and to provide all stakeholders with incentives to engage in conversations about (care) preferences in an informative and participatory manner.

Line-Constrained Camera Location Estimation in Multi-Image Stereomatching.

Stereomatching is an effective way of acquiring dense depth information from a scene when active measurements are not possible. So-called lightfield methods take a snapshot from many camera locations along a defined trajectory (usually uniformly linear or on a regular grid-we will assume a linear trajectory) and use this information to compute accurate depth estimates. However, they require the locations for each of the snapshots to be known: the disparity of an object between images is related to both the distance of the camera to the object and the distance between the camera positions for both images. Existing solutions use sparse feature matching for camera location estimation. In this paper, we propose a novel method that uses dense correspondences to do the same, leveraging an existing depth estimation framework to also yield the camera locations along the line. We illustrate the effectiveness of the proposed technique for camera location estimation both visually for the rectification of epipolar plane images and quantitatively with its effect on the resulting depth estimation. Our proposed approach yields a valid alternative for sparse techniques, while still being executed in a reasonable time on a graphics card due to its highly parallelizable nature.

Coxiella burnetii (Q fever) prevalence in associated populations of humans and small ruminants in The Gambia.

OBJECTIVES: To simultaneously estimate the prevalence of antibodies against Coxiella burnetii (Q fever) among adults and small ruminants, and C. burnetii shedding prevalence among small ruminants in households in the Kiang West district of The Gambia, and to assess associated risk factors. METHODS: Sera of 599 adults and 615 small ruminants from 125 compounds within 12 villages were tested for antibodies against C. burnetii using ELISA. Vaginal swabs and milk samples of 155 small ruminants were tested using PCR to investigate shedding of C. burnetii. RESULTS: A total of 3.8-9.7% of adults, depending on ELISA test cut-off, and 24.9% of small ruminants in Kiang West were seropositive. Having at least one seropositive animal in one's compound was a risk factor for human seropositivity (OR: 3.35, 95% CI: 1.09-14.44). A grazing area within a village was a risk factor for seropositivity in small ruminants (OR: 2.07, 95% CI: 1.26-3.50); others were having lambed (OR: 2.75, 95% CI: 1.37-5.76) and older age of the animals (OR: 2.75, 95% CI: 1.37-5.76 for 1-3 years and OR 5.84, 95% CI: 3.10-11.64 for >3 years); 57.4% of sampled small ruminants were shedding C. burnetii. CONCLUSION: Coxiella burnetii infection is endemic among both humans and small ruminants in this area of The Gambia. Human and animal exposure to C. burnetii were related at compound level. Further research into the clinical relevance of C. burnetii infection in West Africa is needed.

MATE: Machine Learning for Adaptive Calibration Template Detection.

The problem of camera calibration is two-fold. On the one hand, the parameters are estimated from known correspondences between the captured image and the real world. On the other, these correspondences themselves-typically in the form of chessboard corners-need to be found. Many distinct approaches for this feature template extraction are available, often of large computational and/or implementational complexity. We exploit the generalized nature of deep learning networks to detect checkerboard corners: our proposed method is a convolutional neural network (CNN) trained on a large set of example chessboard images, which generalizes several existing solutions. The network is trained explicitly against noisy inputs, as well as inputs with large degrees of lens distortion. The trained network that we evaluate is as accurate as existing techniques while offering improved execution time and increased adaptability to specific situations with little effort. The proposed method is not only robust against the types of degradation present in the training set (lens distortions, and large amounts of sensor noise), but also to perspective deformations, e.g., resulting from multi-camera set-ups.

Low Seroprevalence of Brucellosis in Humans and Small Ruminants in the Gambia.

BACKGROUND: Brucellosis is a worldwide zoonosis with significant impact on rural livelihoods and a potentially underestimated contributor to febrile illnesses. The aim of this study was to estimate the seroprevalence of brucellosis in humans and small ruminants in The Gambia. METHODS: The study was carried out in rural and urban areas. In 12 rural villages in Kiang West district, sera were collected from humans (n = 599) and small ruminants (n = 623) from the same compounds. From lactating small ruminants, milk samples and vaginal swabs were obtained. At the urban study sites, sera were collected from small ruminants (n = 500) from slaughterhouses and livestock markets. Information on possible risk factors for seropositivity was collected through questionnaires. Sera were screened for antibodies against Brucella spp. with the Rose Bengal Test, ELISA and Micro Agglutination Test (human sera only). PCR was performed on 10 percent of the milk samples and vaginal swabs from small ruminants. RESULTS: One human and 14 sheep sera were positive by the Rose Bengal Test. The rest were negative in all serological tests used. The PCR results were all negative. CONCLUSIONS: The results suggest that brucellosis is currently not a generalized problem in humans or small ruminants in The Gambia.

Optimal Joint Detection and Estimation That Maximizes ROC-Type Curves.

Combined detection-estimation tasks are frequently encountered in medical imaging. Optimal methods for joint detection and estimation are of interest because they provide upper bounds on observer performance, and can potentially be utilized for imaging system optimization, evaluation of observer efficiency, and development of image formation algorithms. We present a unified Bayesian framework for decision rules that maximize receiver operating characteristic (ROC)-type summary curves, including ROC, localization ROC (LROC), estimation ROC (EROC), free-response ROC (FROC), alternative free-response ROC (AFROC), and exponentially-transformed FROC (EFROC) curves, succinctly summarizing previous results. The approach relies on an interpretation of ROC-type summary curves as plots of an expected utility versus an expected disutility (or penalty) for signal-present decisions. We propose a general utility structure that is flexible enough to encompass many ROC variants and yet sufficiently constrained to allow derivation of a linear expected utility equation that is similar to that for simple binary detection. We illustrate our theory with an example comparing decision strategies for joint detection-estimation of a known signal with unknown amplitude. In addition, building on insights from our utility framework, we propose new ROC-type summary curves and associated optimal decision rules for joint detection-estimation tasks with an unknown, potentially-multiple, number of signals in each observation.

MRI segmentation of the human brain: challenges, methods, and applications.

Image segmentation is one of the most important tasks in medical image analysis and is often the first and the most critical step in many clinical applications. In brain MRI analysis, image segmentation is commonly used for measuring and visualizing the brain's anatomical structures, for analyzing brain changes, for delineating pathological regions, and for surgical planning and image-guided interventions. In the last few decades, various segmentation techniques of different accuracy and degree of complexity have been developed and reported in the literature. In this paper we review the most popular methods commonly used for brain MRI segmentation. We highlight differences between them and discuss their capabilities, advantages, and limitations. To address the complexity and challenges of the brain MRI segmentation problem, we first introduce the basic concepts of image segmentation. Then, we explain different MRI preprocessing steps including image registration, bias field correction, and removal of nonbrain tissue. Finally, after reviewing different brain MRI segmentation methods, we discuss the validation problem in brain MRI segmentation.

Automatic high-bandwidth calibration and reconstruction of arbitrarily sampled parallel MRI.

Today, many MRI reconstruction techniques exist for undersampled MRI data. Regularization-based techniques inspired by compressed sensing allow for the reconstruction of undersampled data that would lead to an ill-posed reconstruction problem. Parallel imaging enables the reconstruction of MRI images from undersampled multi-coil data that leads to a well-posed reconstruction problem. Autocalibrating pMRI techniques encompass pMRI techniques where no explicit knowledge of the coil sensivities is required. A first purpose of this paper is to derive a novel autocalibration approach for pMRI that allows for the estimation and use of smooth, but high-bandwidth coil profiles instead of a compactly supported kernel. These high-bandwidth models adhere more accurately to the physics of an antenna system. The second purpose of this paper is to demonstrate the feasibility of a parameter-free reconstruction algorithm that combines autocalibrating pMRI and compressed sensing. Therefore, we present several techniques for automatic parameter estimation in MRI reconstruction. Experiments show that a higher reconstruction accuracy can be had using high-bandwidth coil models and that the automatic parameter choices yield an acceptable result.

Coxiella burnetii seroprevalence in small ruminants in The Gambia.

BACKGROUND: Q fever is a zoonosis caused by Coxiella burnetii, a Gram negative bacterium present worldwide. Small ruminants are considered the main reservoirs for infection of humans. This study aimed to estimate the extent of C. burnetii infection among sheep and goats in part of The Gambia. METHODOLOGY/PRINCIPAL FINDINGS: This survey was carried out from March to May 2012 at two areas in The Gambia. The first area comprised a cluster of seven rural villages situated 5-15 km west of Farafenni as well as the local abattoir. A second sampling was done at the central abattoir in Abuko (30 km from the capital, Banjul) in the Western Region. Serum samples were obtained from 490 goats and 398 sheep. In addition, 67 milk samples were obtained from lactating dams. Sera were tested with a Q fever ELISA kit. C. burnetii DNA was extracted from milk samples and then detected using a specific quantitative multiplex PCR assay, targeting the IS1111a element. A multivariable mixed logistic regression model was used to examine the relationship between seropositivity and explanatory variables. An overall seroprevalence of 21.6% was found. Goats had a significantly higher seroprevalence than sheep, respectively 24.2% and 18.5%. Seropositive animals were significantly older than seronegative animals. Animals from the villages had a significantly lower seroprevalence than animals from the central abattoir (15.1% versus 29.1%). C. burnetii DNA was detected in 2 out of 67 milk samples, whereas 8 samples gave a doubtful result. CONCLUSION/SIGNIFICANCE: A substantial C. burnetii seroprevalence in sheep and goats in The Gambia was demonstrated. People living in close proximity to small ruminants are exposed to C. burnetii. Q fever should be considered as a possible cause of acute febrile illness in humans in The Gambia. Future studies should include a simultaneous assessment of veterinary and human serology, and include aetiology of febrile illness in local clinics.

Nonparametric estimation receiver operating characteristic analysis for performance evaluation on combined detection and estimation tasks.

In an effort to generalize task-based assessment beyond traditional signal detection, there is a growing interest in performance evaluation for combined detection and estimation tasks, in which signal parameters, such as size, orientation, and contrast are unknown and must be estimated. One motivation for studying such tasks is their rich complexity, which offers potential advantages for imaging system optimization. To evaluate observer performance on combined detection and estimation tasks, Clarkson introduced the estimation receiver operating characteristic (EROC) curve and the area under the EROC curve as a summary figure of merit. This work provides practical tools for EROC analysis of experimental data. In particular, we propose nonparametric estimators for the EROC curve, the area under the EROC curve, and for the variance/covariance matrix of a vector of correlated EROC area estimates. In addition, we show that reliable confidence intervals can be obtained for EROC area, and we validate these intervals with Monte Carlo simulation. Application of our methodology is illustrated with an example comparing magnetic resonance imaging [Formula: see text]-space sampling trajectories. MATLAB® software implementing the EROC analysis estimators described in this work is publicly available at http://code.google.com/p/iqmodelo/.

Channelized model observer for the detection and estimation of signals with unknown amplitude, orientation, and size.

As a task-based approach for medical image quality assessment, model observers (MOs) have been proposed as surrogates for human observers. While most MOs treat only signal-known-exactly tasks, there are few studies on signal-known-statistically (SKS) MOs, which are clinically more relevant. In this paper, we present a new SKS MO named channelized joint detection and estimation observer (CJO), capable of detecting and estimating signals with unknown amplitude, orientation, and size. We evaluate its estimation and detection performance using both synthesized (correlated Gaussian) backgrounds and real clinical (magnetic resonance) backgrounds. The results suggest that the CJO has good performance in the SKS detection-estimation task.

Computationally efficient locally adaptive demosaicing of color filter array images using the dual-tree complex wavelet packet transform.

Most digital cameras use an array of alternating color filters to capture the varied colors in a scene with a single sensor chip. Reconstruction of a full color image from such a color mosaic is what constitutes demosaicing. In this paper, a technique is proposed that performs this demosaicing in a way that incurs a very low computational cost. This is done through a (dual-tree complex) wavelet interpretation of the demosaicing problem. By using a novel locally adaptive approach for demosaicing (complex) wavelet coefficients, we show that many of the common demosaicing artifacts can be avoided in an efficient way. Results demonstrate that the proposed method is competitive with respect to the current state of the art, but incurs a lower computational cost. The wavelet approach also allows for computationally effective denoising or deblurring approaches.

Short communication: prevalence of antibodies against Coxiella burnetii (Q fever) in children in The Gambia, West Africa.

OBJECTIVE: To estimate the prevalence of antibodies against Coxiella burnetii (Q fever) among children in eight villages in The Gambia, West Africa. METHODS: Sera of 796 children aged 1-15 years were tested for presence of antibodies against phase II of C. burnetii by ELISA. RESULTS: IgG and/or IgM phase II antibodies against C. burnetii were detectable in 8.3% (66/796) of all serum samples analysed with significant differences in seroprevalence between villages. Highest prevalence was found in the age group 1-4 years. CONCLUSIONS: Exposure to C. burnetii is considerable in the early years of life in The Gambia, and further studies are warranted to estimate the role of Q fever in acute febrile illness in The Gambia and elsewhere in Africa.

Channelized Hotelling observers for the assessment of volumetric imaging data sets.

Current clinical practice is rapidly moving in the direction of volumetric imaging. For two-dimensional (2D) images, task-based medical image quality is often assessed using numerical model observers. For three-dimensional (3D) images, however, these models have been little explored so far. In this work, first, two novel designs of a multislice channelized Hotelling observer (CHO) are proposed for the task of detecting 3D signals in 3D images. The novel designs are then compared and evaluated in a simulation study with five different CHO designs: a single-slice model, three multislice models, and a volumetric model. Four different random background statistics are considered, both gaussian (noncorrelated and correlated gaussian noise) and non-gaussian (lumpy and clustered lumpy backgrounds). Overall, the results show that the volumetric model outperforms the others, while the disparity between the models decreases for greater complexity of the detection task. Among the multislice models, the second proposed CHO could most closely approach the volumetric model, whereas the first new CHO seems to be least affected by the number of training samples.