Automatic analysis of neonatal video data to evaluate resuscitation performance

An approach to human action classification in videos is presented, based on knowledge-aware initial features extracted from human skeleton data and on further processing by convolutional networks. The proposed smart tracking of skeleton joints, approximation of missing joints and normalization of skeleton data are important steps of feature extraction. Three neural network models—based on LSTM, Transformer and CNN—are developed and experimentally verified. The models are trained and tested on the well-known NTU-RGB+D (Shahroudy et al., 2016) dataset in the cross-view mode. The obtained results show a competitive performance with other SOTA methods and verify the efficiency of proposed feature engineering. The network has a five times lower number of trainable parameters than other proposed methods to reach nearly similar performance and twenty times lower number than the currently best performing solutions. Thanks to the lightness of the classifier, the solution only requires relatively small computational resources.

Action Recognition in Still Images using Residual Neural Network Features

Sparsity-inducing dictionaries for effective action classification

BackgroundInfection control and neonatal resuscitation are essential midwifery practices that can reduce maternal and neonatal mortality. However, in Ethiopia, theory-heavy midwifery education leads to limited clinical competence. To address this, Debre Tabor University implemented a competency-based curriculum in 2013. This study examines whether competency-based midwifery education produces graduates with significantly better performance in neonatal resuscitation and infection prevention compared to conventional education, thereby framing a testable argument about the curriculum’s effectiveness.MethodsA comparative cross-sectional study assessed the infection prevention and neonatal resuscitation performance of 68 BSc midwifery graduates (32 competency-based vs. 36 conventional) from third-generation Ethiopian universities. Performance was measured using a validated observation tool in clinical settings for infection prevention and simulations for neonatal resuscitation. Mean percentage scores were compared using t-tests, with effect size illustrated via Gardner–Altman plots.ResultsOverall, midwives demonstrated 63.6% of essential neonatal resuscitation tasks, with competency-based curriculum graduates (CBCGs) having higher performance than conventional curriculum graduates (CCGs) (71.6% vs. 56.5%; t (66.0) = 3.82, p < .001; difference = 15.1%), particularly in airway suctioning and chest rise assessment. For infection prevention, midwives performed 71.7% of the required tasks, with CBCGs again scoring higher (76.9% vs. 67.0%; t (55.4) = 2.79, p < .01; difference = 9.9%). Key differences were observed in hand hygiene, the use of personal protective equipment, and apron decontamination. Despite these improvements, persistent deficiencies remained in both groups, particularly in checking breathing/pulse during neonatal resuscitation and in disinfecting aprons during infection prevention practices.ConclusionsCBCGs demonstrated better performance in neonatal resuscitation and infection prevention compared to those from the conventional program, suggesting clinical relevance. However, performance gaps in both groups underscore the need for enhanced simulation training, ongoing skill reinforcement, and curriculum refinement.

Infection control and neonatal resuscitation are essential midwifery practices that can reduce maternal and neonatal mortality. However, in Ethiopia, theory-heavy midwifery education leads to limited clinical competence. To address this, Debre Tabor University implemented a competency-based curriculum in 2013. This study examines whether competency-based midwifery education produces graduates with significantly better performance in neonatal resuscitation and infection prevention compared to conventional education, thereby framing a testable argument about the curriculum's effectiveness. A comparative cross-sectional study assessed the infection prevention and neonatal resuscitation performance of 68 BSc midwifery graduates (32 competency-based vs. 36 conventional) from third-generation Ethiopian universities. Performance was measured using a validated observation tool in clinical settings for infection prevention and simulations for neonatal resuscitation. Mean percentage scores were compared using t-tests, with effect size illustrated via Gardner-Altman plots. Overall, midwives demonstrated 63.6% of essential neonatal resuscitation tasks, with competency-based curriculum graduates (CBCGs) having higher performance than conventional curriculum graduates (CCGs) (71.6% vs. 56.5%; t (66.0) = 3.82, p < .001; difference = 15.1%), particularly in airway suctioning and chest rise assessment. For infection prevention, midwives performed 71.7% of the required tasks, with CBCGs again scoring higher (76.9% vs. 67.0%; t (55.4) = 2.79, p < .01; difference = 9.9%). Key differences were observed in hand hygiene, the use of personal protective equipment, and apron decontamination. Despite these improvements, persistent deficiencies remained in both groups, particularly in checking breathing/pulse during neonatal resuscitation and in disinfecting aprons during infection prevention practices. CBCGs demonstrated better performance in neonatal resuscitation and infection prevention compared to those from the conventional program, suggesting clinical relevance. However, performance gaps in both groups underscore the need for enhanced simulation training, ongoing skill reinforcement, and curriculum refinement.

Reply

After completing this article, readers should be able to: 1. Define the indeterminate class of recommendations for neonatal resuscitation. 2. Describe the two areas of current investigation within the indeterminate class recommendations. 3. Describe the application of two techniques from other settings within the indeterminate class recommendations. 4. Describe the indeterminate class recommendation for which conflicting evidence is emerging. With the shift to evidence-based guidelines, the process of revising the scientific framework for neonatal resuscitation and the derivative educational efforts will become more predictable and accessible. Beginning with the International Guidelines 2000, an Indeterminate Class of recommendations appeared. These focused on areas of intense scientific research that may lead to clinically important therapies; technological developments widely adopted for use in other age groups that may find a role in neonatal resuscitation; or emerging evidence that conflicts substantially with previous data, resulting in a revision of recommendations to withdraw support of a particular therapeutic approach. The advent of changes in evidence-based guidelines carries the obligation to monitor the impact of such changes. Finally, entirely new questions and proposed guideline recommendations will be submitted to evidence evaluation in the future. Five Indeterminate Class recommendations appeared in the neonatal resuscitation portion of the International Guidelines 2000 (Table⇓ ). Cerebral hypothermia following hypoxic-ischemic insult and positive-pressure ventilation with room air represent proposals in the translational research phase, moving from animal and molecular models into clinical trials. The recommendations relating to adjunctive airway techniques, laryngeal mask airway and exhaled carbon dioxide detection, recognize the importance of these techniques in the older pediatric and adult populations, but acknowledge the significant limitations in their application to neonates. The statement regarding high-dose epinephrine reinforces the conflicting nature of evidence relating to this therapy, yet it acknowledges that available evidence is extrapolated largely from older age groups and falls short of supporting …

Are You Ready for the Change? Embracing the Neonatal Resuscitation Program Guidelines of Simulation and Debrief

Image and video data contains abundant, rich information for data miners to explore. On one hand, the rich literature on image and video data analysis will naturally provide many advanced methods that may help mining other kinds of data. On the other hand, recent research on data mining will also provide some new, interesting methods that may benefit image and video data retrieval and analysis. In this talk we explore the latter, and discuss whether the new results obtained in data mining research could be useful in image and video data retrieval and analysis. Our discussion will be focused on the following aspects: (1) how frequent pattern, sequential pattern, and structural pattern analysis methods may help image and video data analysis; (2) how data mining may help construction of effective and efficient indexing and similarity search mechanisms for image and video retrieval; (3) how discriminative pattern-based classification methods may shed new light on image and video classification; and (4) how pattern-based analysis methods may help high-dimensional clustering in image and video analysis. Our goal is to promote collaborative research between these two research communities.

As an important issue in video classification, human action recognition is becoming a hot topic in computer vision. The ways of effectively representing the spatial static and temporal dynamic information of videos are important problems in video action recognition. This paper proposes an attention mechanism based convolutional LSTM action recognition algorithm to improve the accuracy of recognition by extracting the salient regions of actions in videos effectively. First, GoogleNet is used to extract the features of video frames. Then, those feature maps are processed by the spatial transformer network for the attention. Finally the sequential information of the features is modeled via the convolutional LSTM to classify the action in the original video. To accelerate the training speed, we adopt the analysis of temporal coherence to reduce the redundant features extracted by GoogleNet with trivial accuracy loss. In comparison with the state-of-the-art algorithms for video action recognition, competitive results are achieved on three widely-used datasets, UCF-11, HMDB-51 and UCF-101. Moreover, by using the analysis of temporal coherence, desirable results are obtained while the training time is reduced.

This study aimed to evaluate whether a custom warmer height improves the quality and consistency of chest compressions (CCs) compared with a standard warmer height during simulated neonatal resuscitation.Cross-over study using simulated neonatal resuscitation. A controlled research environment equipped with a 12-camera motion capture system, four in-floor multi-axis force plates, a neonatal manikin, and resuscitation equipment. Biomechanical assessments were recorded every 2 minutes during a 20-minute simulation for each condition. Twenty Neonatal Resuscitation Program (NRP)-trained providers. Each participant performed two 20-minute CC sessions-one with the warmer at the standard 100 cm height and one at a custom height selected by the participant. CC depth, force, and rate; participant back angle, heart rate, and self-reported exertion, were analyzed at 2-minute intervals.Compared with the standard height, the custom height resulted in greater and more consistent CC depth and force while maintaining compression rate. Participants also exhibited a greater back angle, and lower heart rate, and reported reduced exertion under the custom height condition.Allowing NRP-trained providers to adjust warmer heights led to improved CC quality and consistency, suggesting that customizable warmer heights may enhance neonatal resuscitation performance. KEY POINTS: · Custom warmer height chosen by NRP-trained providers resulted in more consistent and greater CC depth and force.. · It also was associated with less provider fatigue, compared with standard height.. · During neonatal resuscitation, frontline healthcare professionals changed.. · Participant heart rate was lower when using the custom versus standard height.. · Our findings support the need for guidelines on adjusting warmer height during neonatal cardiopulmonary resuscitation..

Pediatric resuscitation guidelines in the new millennium.

Introduction: Anaesthetists are among several health care practitioners responsible for neonatal resuscitation in Canada. The Neonatal resuscitation program (NRP) courses are the North American educational standard. NRP has been shown to be an effective way of learning skills and knowledge but retention has been found to be problematic [1]. The use of cognitive aids is mandatory in industries such as aviation, to avoid dependence on memory when decision making in critical situations. Visual cognitive aids have been studied retrospectively in resuscitation and performance was found to correlate to the frequency of use of the aid [2]. Cognitive aids have been found to be of benefit in an unblinded prospective study [3]. We aimed to conduct the first blinded study on the effect of a cognitive aid on the performance of simulated resuscitation. Methods: We conducted a single-blind randomized controlled trial to investigate whether the presence of a cognitive aid improved performance in a simulated neonatal resuscitation. After ethics board approval we recruited 32 anaesthesia residents who had previously passed the NRP. Subjects were randomized to an intervention group that had a poster detailing the NRP algorithm and a control group without the poster. The cognitive aid was positioned so that it could not be seen on the video recordings of the simulation that was used to assess performance. The scenario was piloted to confirm adequate blinding. Both groups had their performance in a simulated neonatal resuscitation recorded and subsequently analyzed by a peer, an expert anaesthetist and an expert neonatologist, using a previously validated checklist. A further rater observed the scenario in real time to examine frequency of use of the cognitive aid. Results: The inter-rater reliability of the checklist was excellent with an intraclass correlation coefficient of 0.88. Consequently the mean of the scores assigned by all three raters was used for analysis. The median checklist score in the control group 18.2 [15.0–20.5 (10.7–25.3)] was not significantly different from that in the intervention group 20.3 [18.3–21.3 (15.0–24.3)] (P = 0.08). Retention of NRP skills and knowledge of was poor: when evaluated by the neonatologist none of the subjects correctly performed all life-saving interventions necessary to pass the checklist. Although only one subject in the intervention group did not use the aid at all, only 26.7% used the aid frequently and none used it extensively. Discussion: Retention of skills after NRP training was poor. Our study confirms previous findings of poor retention of skills after NRP training: Kaczorowski et al. investigated family medicine trainees and found that none of 44 residents that were retested 6–8 months after an NRP course would have passed the course due to errors in life-saving interventions [1]. Previous research has shown that the presence of a cognitive aid can improve performance in the simulated management of a rare, high stakes scenario: malignant hyperthermia [3]. Our negative findings contrast with this and another previous study [2]. A potential reason for this discrepancy is that the raters in the previous studies were not blinded to group allocation, nor were the rating scales used validated. The infrequent use of the cognitive aid may be the reason that it did not improve performance in. Further research is required to investigate whether cognitive aids can be useful if their use is incorporated into NRP training. Conclusion: A randomized single-blinded trial found that a cognitive aid did not improve performance at simulated resuscitation, in contrast to previous retrospective and unblended studies. Retention of skills and knowledge after resuscitation training remains an ongoing challenge for medical educators.

The aim of this paper is to address recognition of natural human actions in diverse and realistic video settings. This challenging but important subject has mostly been ignored in the past due to several problems one of which is the lack of realistic and annotated video datasets. Our first contribution is to address this limitation and to investigate the use of movie scripts for automatic annotation of human actions in videos. We evaluate alternative methods for action retrieval from scripts and show benefits of a text-based classifier. Using the retrieved action samples for visual learning, we next turn to the problem of action classification in video. We present a new method for video classification that builds upon and extends several recent ideas including local space-time features, space-time pyramids and multi-channel non-linear SVMs. The method is shown to improve state-of-the-art results on the standard KTH action dataset by achieving 91.8% accuracy. Given the inherent problem of noisy labels in automatic annotation, we particularly investigate and show high tolerance of our method to annotation errors in the training set. We finally apply the method to learning and classifying challenging action classes in movies and show promising results.

Decision support tools (DST) may aid compliance of teams with the Neonatal Resuscitation Program (NRP) algorithm but have not been adequately tested in this population. Furthermore, the optimal team size for neonatal resuscitation is not known. Our aim was to determine whether use of a tablet-based DST or team size altered adherence to the NRP algorithm in teams of healthcare providers (HCPs) performing simulated neonatal resuscitation. One hundred nine HCPs were randomized into a team of 2 or 3 and into using a DST or memory alone while performing 2 simulation scenarios. The primary outcome was NRP compliance, assessed by the modified Neonatal Resuscitation Performance Evaluation (NRPE). Secondary outcomes were the subcomponents of the NRPE score, cumulative time error (the cumulative time in seconds to perform resuscitation tasks in error, early or late, from NRP guidelines), and the interaction between DST and team size. Decision support tool use improved total NRPE score when compared with memory alone (p = 0.015). There was no difference in NRPE score within teams of 2 compared with 3 HCPs. Cumulative time error was decreased with DST use compared with memory alone but was not significant (p = 0.057). Team size did not affect time error. Teams with the DST had improved NRP adherence compared with teams relying on memory alone in 1 of 2 scenarios. Two and 3 HCP teams performed similarly. Given the positive results observed in the simulated environment, further testing the DST in the clinical environment is warranted.