A Preliminary Study of Deep Learning Sensor Fusion for Pedestrian Detection.

Most pedestrian detection methods focus on bounding boxes based on fusing RGB with lidar. These methods do not relate to how the human eye perceives objects in the real world. Furthermore, lidar and vision can have difficulty detecting pedestrians in scattered environments, and radar can be used to overcome this problem. Therefore, the motivation of this work is to explore, as a preliminary step, the feasibility of fusing lidar, radar, and RGB for pedestrian detection that potentially can be used for autonomous driving that uses a fully connected convolutional neural network architecture for multimodal sensors. The core of the network is based on SegNet, a pixel-wise semantic segmentation network. In this context, lidar and radar were incorporated by transforming them from 3D pointclouds into 2D gray images with 16-bit depths, and RGB images were incorporated with three channels. The proposed architecture uses a single SegNet for each sensor reading, and the outputs are then applied to a fully connected neural network to fuse the three modalities of sensors. Afterwards, an up-sampling network is applied to recover the fused data. Additionally, a custom dataset of 60 images was proposed for training the architecture, with an additional 10 for evaluation and 10 for testing, giving a total of 80 images. The experiment results show a training mean pixel accuracy of 99.7% and a training mean intersection over union of 99.5%. Also, the testing mean of the IoU was 94.4%, and the testing pixel accuracy was 96.2%. These metric results have successfully demonstrated the effectiveness of using semantic segmentation for pedestrian detection under the modalities of three sensors. Despite some overfitting in the model during experimentation, it performed well in detecting people in test mode. Therefore, it is worth emphasizing that the focus of this work is to show that this method is feasible to be used, as it works regardless of the size of the dataset. Also, a bigger dataset would be necessary to achieve a more appropiate training. This method gives the advantage of detecting pedestrians as the human eye does, thereby resulting in less ambiguity. Additionally, this work has also proposed an extrinsic calibration matrix method for sensor alignment between radar and lidar based on singular value decomposition.

Overview of Biofluids and Flow Sensing Techniques Applied in Clinical Practice.

This review summarizes the current knowledge on biofluids and the main flow sensing techniques applied in healthcare today. Since the very beginning of the history of medicine, one of the most important assets for evaluating various human diseases has been the analysis of the conditions of the biofluids within the human body. Hence, extensive research on sensors intended to evaluate the flow of many of these fluids in different tissues and organs has been published and, indeed, continues to be published very frequently. The purpose of this review is to provide researchers interested in venturing into biofluid flow sensing with a concise description of the physiological characteristics of the most important body fluids that are likely to be altered by diverse medical conditions. Similarly, a reported compilation of well-established sensors and techniques currently applied in healthcare regarding flow sensing is aimed at serving as a starting point for understanding the theoretical principles involved in the existing methodologies, allowing researchers to determine the most suitable approach to adopt according to their own objectives in this broad field.

Acoustic surveillance of cough for detecting respiratory disease using artificial intelligence.

Research question: Can smartphones be used to detect individual and population-level changes in cough frequency that correlate with the incidence of coronavirus disease 2019 (COVID-19) and other respiratory infections? Methods: This was a prospective cohort study carried out in Pamplona (Spain) between 2020 and 2021 using artificial intelligence cough detection software. Changes in cough frequency around the time of medical consultation were evaluated using a randomisation routine; significance was tested by comparing the distribution of cough frequencies to that obtained from a model of no difference. The correlation between changes of cough frequency and COVID-19 incidence was studied using an autoregressive moving average analysis, and its strength determined by calculating its autocorrelation function (ACF). Predictors for the regular use of the system were studied using a linear regression. Overall user experience was evaluated using a satisfaction questionnaire and through focused group discussions. Results: We followed-up 616 participants and collected >62 000 coughs. Coughs per hour surged around the time cohort subjects sought medical care (difference +0.77 coughs·h-1; p=0.00001). There was a weak temporal correlation between aggregated coughs and the incidence of COVID-19 in the local population (ACF 0.43). Technical issues affected uptake and regular use of the system. Interpretation: Artificial intelligence systems can detect changes in cough frequency that temporarily correlate with the onset of clinical disease at the individual level. A clearer correlation with population-level COVID-19 incidence, or other respiratory conditions, could be achieved with better penetration and compliance with cough monitoring.

Implicit learning of predictable sound sequences modulates human brain responses at different levels of the auditory hierarchy.

Deviant stimuli, violating regularities in a sensory environment, elicit the mismatch negativity (MMN), largely described in the Event-Related Potential literature. While it is widely accepted that the MMN reflects more than basic change detection, a comprehensive description of mental processes modulating this response is still lacking. Within the framework of predictive coding, deviance processing is part of an inference process where prediction errors (the mismatch between incoming sensations and predictions established through experience) are minimized. In this view, the MMN is a measure of prediction error, which yields specific expectations regarding its modulations by various experimental factors. In particular, it predicts that the MMN should decrease as the occurrence of a deviance becomes more predictable. We conducted a passive oddball EEG study and manipulated the predictability of sound sequences by means of different temporal structures. Importantly, our design allows comparing mismatch responses elicited by predictable and unpredictable violations of a simple repetition rule and therefore departs from previous studies that investigate violations of different time-scale regularities. We observed a decrease of the MMN with predictability and interestingly, a similar effect at earlier latencies, within 70 ms after deviance onset. Following these pre-attentive responses, a reduced P3a was measured in the case of predictable deviants. We conclude that early and late deviance responses reflect prediction errors, triggering belief updating within the auditory hierarchy. Beside, in this passive study, such perceptual inference appears to be modulated by higher-level implicit learning of sequence statistical structures. Our findings argue for a hierarchical model of auditory processing where predictive coding enables implicit extraction of environmental regularities.

Effects of prone position and positive end-expiratory pressure on lung perfusion and ventilation.

OBJECTIVES: Prone positioning is frequently used during acute respiratory distress syndrome. However, mechanisms by which it improves oxygenation are poorly understood, as well as its interaction with positive end-expiratory pressure. This study was conducted to decipher the respective effects of positive end-expiratory pressure and posture during lung injury on regional lung ventilation, perfusion and recruitment assessed by positron emission tomography. DESIGN: Experimental study. SETTING: Research laboratory of a university hospital. SUBJECTS: Six female piglets. INTERVENTIONS: After oleic acid-induced lung injury, all animals were studied in supine and prone position at both positive end-expiratory pressure 0 and positive end-expiratory pressure 10 cm H2O. MEASUREMENTS AND MAIN RESULTS: In each experimental condition, regional lung perfusion and ventilation were assessed with positron emission tomograph using intravenous 15O-labeled water and inhaled nitrogen-13. Nonaerated lung weight was assessed with positron emission tomograph, and alveolar recruitment was defined as the difference of nonaerated lung weight between conditions. Positive end-expiratory pressure was associated with significant alveolar recruitment (130 +/- 85 and 65 +/- 29 g of lung in supine and prone position, respectively [p < 0.05 vs. 0]), whereas recruitment induced by posture was not statistically significant (77 +/- 97 g with positive end-expiratory pressure 0 and 13 +/- 19 g with positive end-expiratory pressure 10 [p > 0.05 vs. 0]). Regardless the posture, positive end-expiratory pressure redistributed both perfusion and ventilation toward dependent regions. Recruitment by positive end-expiratory pressure was restricted to dorsal regions in supine position, but extended diffusely along the ventral-to-dorsal dimension in prone position. Prone position was associated with recruitment in dorsal regions with concomitant derecruitment in ventral regions, magnitude of this being reduced by positive end-expiratory pressure. Prone position redistributed ventilation toward dorsal and ventral regions at positive end-expiratory pressure 0 and positive end-expiratory pressure, respectively. Finally, prone position redistributed perfusion toward ventral regions, to an extent amplified by positive end-expiratory pressure. CONCLUSIONS: Positive end-expiratory pressure and posture act synergistically by redistributing lung regional perfusion toward ventral regions, but have antagonistic effects on regional ventilation.

Effect of activated protein C on pulmonary blood flow and cytokine production in experimental acute lung injury.

OBJECTIVE: In acute lung injury (ALI) activated protein C (APC) may reopen occluded lung vessels and minimize lung inflammation. We aimed at assessing the effect of APC on regional lung perfusion, aerated lung volume, cytokine production and oxygenation in experimental ALI. DESIGN AND SETTING: Prospective, controlled study in an imaging facility. PARTICIPANTS: Pigs tracheotomized and mechanically ventilated. INTERVENTION: Pigs were randomly given intravenously APC (n = 8) or saline (n = 8). Thirty minutes later, ALI was induced by injecting oleic acid. MEASUREMENTS AND RESULTS: Lung perfusion and aerated lung volume measured with positron emission tomography, plasma cytokines and arterial blood gas were determined just before ALI and 110 and 290 min thereafter. Lung cytokines were measured at the end of the experiment. PaO2 under F I O2 1 was significantly lower in the APC group before lung injury (473+/-129 vs. 578+/-54 mmHg) and 110 min (342+/-138 vs. 446+/-103 mmHg) and 290 min (303+/-171 vs. 547+/-54 mmHg) thereafter (p < 0.05). Lung perfusion nonsignificantly tended to redistribute towards dorsal lung regions with APC. Total aerated lung volume was not different between APC and control before ALI (10.0+/-1.5 vs. 11.0+/-2.5 ml/kg) (p > 0.05) or thereafter. Plasma IL-6 and IL-8 at 110 min were greater with APC (p < 0.05). CONCLUSIONS: In contrast to studies using other models, pretreatment with APC was associated with worsening oxygenation in the present investigation. This might be due to ventilation-perfusion mismatch, with more perfusion to dependent nonaerated areas.

MRI intensity nonuniformity correction using simultaneously spatial and gray-level histogram information.

A novel approach for correcting intensity nonuniformity in magnetic resonance imaging (MRI) is presented. This approach is based on the simultaneous use of spatial and gray-level histogram information. Spatial information about intensity nonuniformity is obtained using cubic B-spline smoothing. Gray-level histogram information of the image corrupted by intensity nonuniformity is exploited from a frequential point of view. The proposed correction method is illustrated using both physical phantom and human brain images. The results are consistent with theoretical prediction, and demonstrate a new way of dealing with intensity nonuniformity problems. They are all the more significant as the ground truth on intensity nonuniformity is unknown in clinical images.

Creation and application of a simulated database of dynamic [18f]MPPF PET acquisitions incorporating inter-individual anatomical and biological variability.

During the process of validation of a new tracer, estimation of performance and validation of processing algorithms have to be investigated with data sets representative of the ground truth. Because this ground truth is hardly accessible in positron emission tomography (PET), validations of processing algorithms often rely on the use of simulated data sets. Considering that Monte Carlo simulators are very time consuming and are not very easy to use, the building of publicly available databases of simulated PET volumes are becoming highly desirable. We present here the methodology employed for the creation of a database of simulated dynamic [18F]MPPF-PET data, including inter-individual anatomical and biological variability which meets the criteria of a gold standard database as defined by Lehmann: reliance, equivalence, independence, relevance, significance. The assessment of the realism of the built database against actual MPPF PET data is also presented here. Whereas the database was specifically created for the investigations of quantification of activity and binding of ligand-receptor with the [18F]MPPF PET tracer, it may serve the community with countless purposes. The full strength of this database, does not only stem from the knowledge of important information such as the true activity map and underlying anatomical data, but also from the possibility to fully control the biological difference between sets of simulated PET data. Indeed, time activity curves included in the simulated data sets are controlled by a multicompartmental model of ligand-receptor exchanges. This latter feature is of a great interest in the context of the improvement of the detectability of biological variation in PET.

Alveolar recruitment assessed by positron emission tomography during experimental acute lung injury.

OBJECTIVES: To compare changes in aerated lung volumes measured by positron emission tomography (PET) and inflation volume-pressure curve (V-P) of the respiratory system, and to evaluate the reliability of PET to assess alveolar recruitment. DESIGN AND SETTING: Experimental study in six anesthetized and mechanically ventilated pigs in a PET facility in an experimental university laboratory. INTERVENTIONS: Lung injury was induced by oleic acid. Animals were randomly studied in four conditions: PEEP 0cmH(2)O (ZEEP) in supine position (SP), PEEP 10cmH(2)O in SP, ZEEP in prone position (PP) and PEEP in PP, each applied for 30min. MEASUREMENTS AND RESULTS: With PET aerated lung volume was obtained from pulmonary density analysis using transmission scan (VA(trans)) and from nitrogen-13 kinetics on emission scan (VA(em)). Changes in VA(trans) and VA(em) were computed as the difference in aerated volume between conditions. VA(trans) and VA(em) did not differ between SP and PP, on either ZEEP or PEEP, suggesting no modification in relaxation volume of the respiratory system induced by posture. Changes in VA(trans) or VA(em) were significantly correlated with changes in aerated volume assessed from superimposed V-P curves (R (2)=0.74 and 0.75, respectively). Alveolar recruitment assessed by PET was significantly correlated with both PaO(2) (R (2)=0.61) and PaCO(2) (R (2)=0.40) variations induced by PEEP. CONCLUSIONS: PET is a new reliable tool of scientific interest to image lung volume and alveolar recruitment during acute lung injury.

Computation of transmitted and received B1 fields in magnetic resonance imaging.

Computation of B1 fields is a key issue for determination and correction of intensity nonuniformity in magnetic resonance images. This paper presents a new method for computing transmitted and received B1 fields. Our method combines a modified MRI acquisition protocol and an estimation technique based on the Levenberg-Marquardt algorithm and spatial filtering. It enables accurate estimation of transmitted and received B1 fields for both homogeneous and heterogeneous objects. The method is validated using numerical simulations and experimental data from phantom and human scans. The experimental results are in agreement with theoretical expectations.

Quantitative assessment of regional alveolar ventilation and gas volume using 13N-N2 washout and PET.

UNLABELLED: Measurement of alveolar volume (Va) and regional ventilation (a) is crucial to understanding the pathophysiology of acute lung injury and ventilator-induced lung injury. PET has previously been used as a noninvasive, quantitative method to assess a, but formal validation of this technique in experimental lung injury is lacking. This study aims to validate Va and a regional assessment with PET, using inhaled (13)N-N(2) in pigs. METHODS: Two normal and 2 oleic acid-injured pigs were tracheotomized, mechanically ventilated, and studied in 5 different levels of ventilation by changing respiratory rate. In each experimental condition, lungs were washed-in and then washed-out with (13)N-N(2) through an open circuit in the ventilator. Using this method, multiframe images were acquired with a dedicated PET camera. Regions of interest (ROIs) were drawn on each lung. Regional time-activity curves during washout were generated for each ROI and fitted to a mono- and a bicompartmental model. Validation of this method was performed in 2 ways. First, regional values of predicted Va (Va(emission)) were compared with regional volume obtained independently from density analysis on a transmission scan (Va(trans)). Second, regional values of predicted a were summed in each animal during each experimental condition and compared with minute-ventilation values set on the ventilator. RESULTS: The bicompartmental model best fitted the experimental values in normal (94.7% [62.2%-100.0%] (median [interquartile range]) of the ROIs) as well as in injured animals (90.7% [81.6%-97.4%] of the ROIs) (P = 0.49). Va(emission) significantly correlated with Va(trans) (R(2) = 0.89, P < 0.001) but exceeded Va(trans) by 10%. Finally, a strongly and positively correlated with minute-ventilation in both normal (R(2) = 0.96, P < 0.001) and injured (R(2) = 0.96, P < 0.001) animals. CONCLUSION: Measurement of (13)N-N(2) washout using PET is accurate to assess regional alveolar volume and ventilation during experimental acute lung injury.

Characterization of PVA cryogel for intravascular ultrasound elasticity imaging.

The present study characterizes the mechanical properties of polyvinyl alcohol (PVA) cryogel in order to show its utility for intravascular elastography. PVA cryogel becomes harder with an increasing number of freeze-thaw cycles, and Young's modulus and Poisson's ratio are measured for seven samples. Mechanical tests were performed on cylindrical samples with a pressure column and on a hollow cylinder with the calculation of an intravascular elastogram. An image of the Young's modulus was obtained from the elastogram using cylinder geometry properties. Results show the mechanical similitude of PVA cryogel with the biological tissues present in arteries. A good agreement between Young's modulus obtained from pressure column and from elastogram was also observed.