Search

We introduce an active learning framework for content-based image retrieval for video surveillance that can be trained ad-hoc for a single camera in a matter of minutes. This technique allows searching for both, known and unknown objects, given a region of interest. The process does not require prior labelled data and treats image retrieval as a binary classification task, in which frames can be similar or different from a query image. The technique is compatible with any pre-trained deep feature extractor. In addition, we propose a novel label propagation algorithm that benefits from (1) visual similarity of image pairs and (2) the semantic representation of the feature vectors from a pre-trained deep feature extractor. This approach allows to reduce the amount of labels needed, while avoiding the propagation of errors. Our experiments with three use-cases from a nuclear facility show the validity of the proposed method, which achieves high precision and recall while requiring minimal amounts of labelled data. [Display omitted] • Present a relevant video frame retrieval pipeline that can be trained interactively. • Propose an active learning algorithm that leverages visual and semantic similarity. • Visual similarity makes label propagation less prone to introducing errors. [ABSTRACT FROM AUTHOR]

Introduction: Tools for improving operative performance for surgical trainees are increasingly desirable, particularly in the context of EWTD and 'run-through' training programmes. In addition, positive direct trainer feedback to trainees can improve skill acquisition and motivation, whilst negative feedback may have the opposite effect.1 We aimed to examine the impact of targeted trainer feedback based on video analysis on trainee confidence and objective operative performance in laparoscopic cholecystectomy.Methods: Selected procedures designated as training cases were recorded. These were assessed by the trainers using the Independence-Scaled Procedural Assessment Score for laparoscopic cholecystectomy. Targeted feedback based on video review of selected procedures was then delivered by the trainers to the trainees. Trainees completed a self-reported questionnaire based on their response to this feedback. Subsequent to the feedback intervention, further training procedures were recorded and assessed.Results: A total of 6 trainees and 4 trainers participated in the study. For the pre-intervention assessment 15 cases were recorded, with a further 13 for the post-intervention assessment (total n = 28). The overall scores for the procedures performed post video feedback were improved, with a trend towards statistical significance (p = 0.08). However, there was a statistically significant improvement in the scores for performance of the triangle of Calot dissection after the feedback intervention (p = 0.009). The response rate to the questionnaire was 100%, with all trainees agreeing that they felt more confident and competent after the feedback intervention.Conclusion: Targeted feedback to trainees based on post-procedure video review improves trainee confidence and may also improve performance. ACGME Core Competencies; Patient Care and Procedural Skills; Practice Based Learning and Improvement. [ABSTRACT FROM AUTHOR]

Introduction: Emergency department thoracotomy (EDT) must be rapid and well-executed. Currently there are no defined benchmarks for EDT procedural milestones. We hypothesized that trauma video review (TVR) can be used to define the 'normative EDT' and generate procedural benchmarks. As a secondary aim, we hypothesized that data collected by TVR would have less missingness and bias than data collected by review of the Electronic Medical Record (EMR).Methods: We used continuously recording video to review all EDTs performed at our centre during the study period. Using skin incision as start time, we defined four procedural milestones for EDT: 1. Decompression of the right chest (tube thoracostomy, finger thoracostomy, or clamshell thoracotomy with transverse sternotomy performed in conjunction with left anterolateral thoracotomy) 2. Retractor deployment 3. Pericardiotomy 4. Aortic Cross-clamp. EDTs with any milestone time ≥ 75th percentile of time or during which a milestone was omitted were identified as outliers. We compared rates of missingness in data collected by TVR and EMR using McNemar's test.Results: 44 EDTs were included from the study period. Patients had a median age of 30 [IQR 25-44] and were predominantly African-American (95%) males (93%) with penetrating trauma (95%). From skin incision, median times in minutes to milestones were as follows: right chest decompression: 2.11 [IQR 0.68-2.83], retractor deployment 1.35 [IQR 0.96-1.85], pericardiotomy 2.35 [IQR 1.85-3.75], aortic cross-clamp 3.71 [IQR 2.83-5.77]. In total, 28/44 (64%) of EDTs were either high outliers for one or more benchmarks or had milestones that were omitted. For all milestones, rates of missingness for TVR data were lower than EMR data (p < 0.001).Conclusions: Video review can be used to define normative times for the procedural milestones of EDT. Steps exceeding the 75th percentile of time were common, with over half of EDTs having at least one milestone as an outlier. Data quality is higher using TVR compared to EMR collection. Future work should seek to determine if minimizing procedural technical outliers improves patient outcomes. [ABSTRACT FROM AUTHOR]

We aimed to evaluate the duration and frequency of communication between EMS (Emergency Medical Services) and ED (Emergency Department) staff during handoff and the subsequent time to critical cardiac care (rhythm determination, defibrillation) using CA (cardiac arrest) video review. A single-center retrospective study of video-recorded adult CAs between August 2020 and December 2022 was performed. Two investigators assessed the communication of 17 data points, time intervals, EMS initiation of handoff, and type of EMS agency. Median times from initiation of handoff to first ED rhythm determination and defibrillation were compared between the groups above versus below the median number of data points communicated. Overall, 95 handoffs were reviewed. The handoff was initiated in a median of 2 seconds (interquartile range (IQR) 0–10) after arrival. EMS initiated handoff in 65 (69.2%) patients. The median number of data points communicated was 9 and median duration was 66 seconds (IQR 50–100). Age, location of arrest, estimated down time, and medications administered were communicated > 80% of the time, initial rhythm 79%, and bystander cardiopulmonary resuscitation and witnessed arrest < 50%. The median times from initiation of handoff to first ED rhythm determination and defibrillation were 188 (IQR 106–256) and 392 (IQR 247–725) seconds, though not statistically different between handoffs with <9 vs. ≥9 data points communicated (p > 0.40). There is no standardization for handoff reports from EMS to ED staff for CA patients. Using video review, we demonstrated the variable communication during handoff. Improvements to this process could reduce the time to critical cardiac care interventions. [ABSTRACT FROM AUTHOR]

Abstract: Study aim: Adherence to Advanced Trauma Life Support (ATLS) protocol has been associated with improved management of injured patients. The objective of this study is to determine factors associated with delayed and omitted ATLS primary and secondary survey tasks at a level 1 pediatric trauma center. Methods: Video recorded resuscitations of 237 injured patients<18 years old obtained over a four month period at our hospital were evaluated to assess completeness and timeliness of essential tasks in the primary and secondary survey of ATLS. Multivariate analyses were performed to identify features associated with decreased ATLS performance. Results: Primary survey findings were stated less often in patients with burn injuries compared to those with blunt injuries (RR=1.72; 95% CI: 1.26–2.35) and less often during the overnight shift [11PM–7AM] (RR=1.22; 95% CI: 1.02–1.46). Secondary survey findings were verbalized less often in patients with penetrating injures (RR=2.30; 95% CI: 1.06–5.00). Time to statement of primary surveys findings was delayed in patients with burn injuries (HR=0.69; 95% CI: 0.48–0.98) and among those transferred from another hospital. Completeness and timeliness of ATLS task performance were not associated with age or injury severity score. Conclusions: Mechanism of injury and hospital factors are associated with incomplete and delayed primary and secondary surveys. Interventions that address deficient ATLS adherence related to these factors may lead to a reduction in errors during this critical period of patient care. [Copyright &y& Elsevier]