Low Quality Video Research Articles

Facial Identity Verification Robust to Pose Variations and Low Image Resolution: Image Comparison Based on Anatomical Facial Landmarks

Face comparison/face mapping is one of the promising methods in face biometrics which needs relatively little effort compared with face identification. Various factors may be used to verify whether two faces are of the same person, among which facial landmarks are one of the most objective indicators due to the same anatomical definition for every face. This study identified major landmarks from 2D and 3D facial images of the same Korean individuals and calculated the distance between the reciprocal landmarks of two images to examine their acceptable range for identifying an individual to obtain standard values from diverse facial angles and image resolutions. Given that reference images obtained in the real-world could be from various angles and resolutions, this study created a 3D face model from multiple 2D images of different angles, and oriented the 3D model to the angle of the reference image to calculate the distance between reciprocal landmarks. In addition, we used the super-resolution method of artificial intelligence to address the inaccurate assessments that low-quality videos can yield. A portion of the process was automated for speed and convenience of face analysis. We conclude that the results of this study could provide a standard for future studies regarding face-to-face analysis to determine if different images are of the same person.

Detection and Mosaicing Techniques for Low-Quality Retinal Videos.

Ideally, to carry out screening for eye diseases, it is expected to use specialized medical equipment to capture retinal fundus images. However, since this kind of equipment is generally expensive and has low portability, and with the development of technology and the emergence of smartphones, new portable and cheaper screening options have emerged, one of them being the D-Eye device. When compared to specialized equipment, this equipment and other similar devices associated with a smartphone present lower quality and less field-of-view in the retinal video captured, yet with sufficient quality to perform a medical pre-screening. Individuals can be referred for specialized screening to obtain a medical diagnosis if necessary. Two methods were proposed to extract the relevant regions from these lower-quality videos (the retinal zone). The first one is based on classical image processing approaches such as thresholds and Hough Circle transform. The other performs the extraction of the retinal location by applying a neural network, which is one of the methods reported in the literature with good performance for object detection, the YOLO v4, which was demonstrated to be the preferred method to apply. A mosaicing technique was implemented from the relevant retina regions to obtain a more informative single image with a higher field of view. It was divided into two stages: the GLAMpoints neural network was applied to extract relevant points in the first stage. Some homography transformations are carried out to have in the same referential the overlap of common regions of the images. In the second stage, a smoothing process was performed in the transition between images.

Quality-Oriented Task Allocation and Scheduling in Transcoding Servers With Heterogeneous Processors

Dynamically adaptive streaming over HTTP requires a large-scale server to transcode various bitrate versions in which different preset parameters can be used to provide different video qualities at each resolution. When transcoding servers contain a heterogeneous mix of CPUs and GPUs, the task scheduler must choose a processor and preset parameter for each transcoding task to meet the transcoding deadlines while achieving the best possible video quality. We apply regression analysis to sample variable-bit-rate videos to provide accurate (mean absolute percentage error values from 1.3% to 13.9%) model for predicting bitrate, transcoding time and video quality at each resolution on different processors. We build this into a greedy allocation and scheduling algorithm which first satisfies deadlines with low video quality, and then redistributes the workload to improve that quality while continuing to meet the deadlines. This scheme was both simulated and implemented on a testbed server. It satisfies all deadlines while outperforming standard algorithms by between 3.12% and 15.59% in terms of popularity-weighted video quality divided by bitrate.

End-to-end Characterization of Game Streaming Applications on Mobile Platforms

With the advent of 5G, supporting high-quality game streaming applications on edge devices has become a reality. This is evidenced by a recent surge in cloud gaming applications on mobile devices. In contrast to video streaming applications, interactive games require much more compute power for supporting improved rendering (such as 4K streaming) with the stipulated frames-per second (FPS) constraints. This in turn consumes more battery power in a power-constrained mobile device. Thus, the state-of-the-art gaming applications suffer from lower video quality (QoS) and/or energy efficiency. While there has been a plethora of recent works on optimizing game streaming applications, to our knowledge, there is no study that systematically investigates the <QoS, Energy> design pairs on the end-to-end game streaming pipeline across the cloud, network, and edge devices to understand the individual contributions of the different stages of the pipeline for improving the overall QoS and energy efficiency. In this context, this paper presents a comprehensive performance and power analysis of the entire game streaming pipeline consisting of the server/cloud side, network, and edge. Through extensive measurements with a high-end workstation mimicking the cloud end, an open-source platform (Moonlight-GameStreaming) emulating the edge device/mobile platform, and two network settings (WiFi and 5G) we conduct a detailed measurement-based study with seven representative games with different characteristics. We characterize the performance in terms of frame latency, QoS, bitrate, and energy consumption for different stages of the gaming pipeline. Our study shows that the rendering stage and the encoding stage at the cloud end are the bottlenecks to support 4K streaming. While 5G is certainly more suitable for supporting enhanced video quality with 4K streaming, it is more expensive in terms of power consumption compared to WiFi. Further, fluctuations in 5G network quality can lead to huge frame drops thus affecting QoS, which needs to be addressed by a coordinated design between the edge device and the server. Finally, the network interface and the decoder units in a mobile platform need more energy-efficient design to support high quality games at a lower cost. These observations should help in designing more cost-effective future cloud gaming platforms.

YouTube as a Source of Information About Air Pollution

YouTube is an important source of information on air pollution. The information presented on youtube may be of high or low quality. In our study, we aimed to evaluate the YouTube videos for their qualities and compare their view numbers, likes, and comment numbers according to high- and low-quality groups. In our study, 32.2% of all videos were in the high-quality group, 15.1% in the intermediate quality group, and 52.7% in the low-quality group according to Global Quality Scale (GQS). Significant differences were found among the quality groups in video length (p < 0.001), dislikes/day (p = 0.043), comment/day (p = 0.005), DISCERN (DS) tool scores (p < 0.001). The duration and DS tool scores were higher in high-quality videos; dislikes/day and comment/day averages were higher in low-quality videos. On the other hand, there was no significant difference in views, views/day, likes/day, likes ratios among quality groups. On this important matter, low-quality videos are abundant. Considering that there are lots of nonreliable information on YouTube, it is needed that individuals should be referred to reliable videos on air pollution.

Performance Evaluation of Action Recognition Models on Low Quality Videos

In the design of action recognition models, the quality of videos is an important issue; however, the trade-off between the quality and performance is often ignored. In general, action recognition models are trained on high-quality videos, hence it is not known how the model performance degrades when tested on low-quality videos, and how much the quality of training videos affects the performance. The issue of video quality is important, however, it has not been studied so far. The goal of this study is to show the trade-off between the performance and the quality of training and test videos by quantitative performance evaluation of several action recognition models for transcoded videos in different qualities. First, we show how the video quality affects the performance of pre-trained models. We transcode the original validation videos of Kinetics400 by changing quality control parameters of JPEG (compression strength) and H.264/AVC (CRF). Then we use the transcoded videos to validate the pre-trained models. Second, we show how the models perform when trained on transcoded videos. We transcode the original training videos of Kinetics400 by changing the quality parameters of JPEG and H.264/AVC. Then we train the models on the transcoded training videos and validate them with the original and transcoded validation videos. Experimental results with JPEG transcoding show that there is no severe performance degradation (up to -1.5%) for compression strength smaller than 70 where no quality degradation is visually observed, and for larger than 80 the performance degrades linearly with respect to the quality index. Experiments with H.264/AVC transcoding show that there is no significant performance loss (up to -1%) with CRF30 while the total size of video files is reduced to 30%.

Collective servicing of heterogenous traffic streams over 3GPP LTE network and application of access control

Due to the progressions in the cellular technologies in recent times, communicating networks have seen vast expansion in the volumes and multiplicities of the data transmitted over the enhanced smart network especially IoT. The regulation and organization of such enormous amounts of data is posing substantial challenges across the various network industries. Because majority of data traffic transferred over such complex networks is comprised of "Big Data" That is continuously growing and evolving. 3GPP have acknowledged such scenarios and have given several recommendations and procedures to manage resources in such situations. But regrettably there are no concrete explanations and techniques on how these resources should be distributed in heterogenous networks. The mathematical modeling that considers the characteristics of traffic flows originating and accepting for servicing is deemed as one of the critical answers to such problems. To address this issue, we have designed a model of resource allocation and sharing strategy for conjoint servicing of real time video traffic referred as "Heavy traffic" of surveillance cameras and low-quality video traffic, referred as "Light traffic" which are transmitted as files over LTE cell facilities in a surveillance operator system. In the model, the access control is incorporated to create the conditions for differentiated servicing of heterogenous traffic incoming sessions. The mathematical model can be utilized to study reservation-based resource allocation scenarios over a shared heterogenous network.

Efficient video quality assessment with deeper spatiotemporal feature extraction and integration

The challenge of video quality assessment (VQA) modeling for user-generated content (UGC) (i.e., UGC-VQA) is how to accurately extract discriminative features and elaborately quantify interfeature interactions by following the behavior patterns of human eye-brain vision perception. To address this issue, we propose the Deeper Spatial–Temporal Scoring Network (DSTS-Net) to give a precise VQA. Concretely, we first deploy the multiscale feature extraction module to characterize content-aware features accounting for nonlinear reverse hierarchy theory in video perception process, which is not fully considered in the reported UGC-VQA models. Hierarchical handcraft and semantic features are simultaneously considered using content adaptive weighting. Second, we develop a feature integration structure, i.e., deeper gated recurrent unit (DGRU), to fully imitate the interfeature interactions in visionary perception, including feedforward and feedback processes. Third, the dual DGRU structure is employed to further account for interframe interactions of hierarchical features, imitating the nonlinearity of perception as much as possible. Finally, improved pooling is achieved in the local adaptive smoothing module accounting for the temporal hysteresis. Holistic validation of the proposed method on four public challenging UGC-VQA datasets presents a comparable performance over the state-of-the-art no-reference VQA methods, especially our method can give an accurate prediction of the low quality videos with weak temporal correlation. To promote reproducible research and public evaluation, an implementation of our method has been made available online: https://github.com/liu0527aa/DSTS-Net.

Anti-jamming heart rate estimation using a spatial–temporal fusion network

Remote heart rate estimation based on face video has attracted increasing research attention. The previous methods are mostly based on ideal face features, where face videos are obtained under controllable experimental conditions, with uniform lighting and little noise interference. However, there is a big deviation from the real world scene. The lighting changes, head movements, and object occlusion are inevitable in complex scenes, which will lead to the failure of face positioning and feature extraction from video, and the instability of the facial signal required for heart rate estimation. In this paper, an anti-jamming network is proposed to improve the robustness of handling less-constrained scenarios. Specifically, a new spatial–temporal map generation mechanism is designed to enhance the spatial and temporal features representation by equivalent padding for low-quality video frame fragments. Meanwhile, a heart rate estimation stability module is built to evaluate the quality of the face signals and assign reasonable weights to video clips. Our approach significantly outperforms all current state-of-the-art methods on VIPL-HR dataset.

Abstract 10715: The Quality of Coronary Artery Bypass Grafting on YouTube

Objective: YouTube, the second most accessed site on the internet, has become a widespread source of healthcare information for patients. Videos for Coronary Artery Bypass Grafts (CABG) have accrued tens of millions of views on the platform, yet their educational quality is unknown. This study investigates the level of bias and misinformation in videos regarding CABG procedures on YouTube. Methods: YouTube was searched for various terms such as “Coronary Bypass Graft Procedure.” After applying exclusion criteria, 73 videos were assessed. Two independent reviewers rated the material with the Global Quality Scale (GQS; 5 = high quality, 3 = low quality) to judge educational value. A ratio of view count to days since upload was applied to assess video popularity. Source, modality, and date of upload were recorded for each video as well. Results: An average GQS score of 2.95 was found, indicating poor educational quality amongst the 73 YouTube videos for CABG procedures. Videos uploaded by physicians (56 or 76.7%) had a significantly higher average GQS score than those not uploaded by physicians (p&lt;0.001). In grouping cohorts by modality, physician-led presentations (24 or 32.9%) produced the highest average GQS score of 3.35; conversely, patient-friendly delivery methods (18 or 24.7%) yielded the lowest average GQS score of 2.36 (p&lt;0.001) (Fig 2.). Neither view ratio nor days since upload significantly correlated with educational quality of the video. Conclusion: Although Coronary Artery Bypass Graft videos are readily available on YouTube, they often contain considerable biases and misleading information. With online sources for healthcare education now commonplace, physicians must be aware of the vast quantities of low-quality videos patients often encounter when weighing different treatment options. Further analysis of CABG videos on YouTube may allow physicians to ameliorate this gap by producing videos that are not only high quality, but highly viewed on the platform.

A CNN and LSTM Network for Eye-Blink Classification from MRI Scanner Monitoring Videos.

Eye closure changes brain activity, so eye-blink tracking of subjects undergoing resting-state functional magnetic resonance imaging (fMRI) is relevant for identifying when a subject blinks, falls asleep, or keeps their eyes closed. Existing MRI eye-tracking solutions use commercially available MR-compatible video cameras with tracking software that can fail on low-quality videos. In this paper, we propose a two-stage convolutional recurrent neural network to classify open and closed eyes from frames of MRI eye-tracking videos under variable camera conditions. The model extracts visual features from each video frame using a convolutional neural network based on the Inception-v3 model, then uses a long short-term memory network to incorporate temporal information encoded in the sequence of visual features over time. Our model is implemented in Keras and demonstrated on a dataset of MRI eye-tracking videos from the Human Connectome Project. We manually labelled frames from the dataset for training and evaluation. The network was able to classify eye-blink states with a precision of 0.739 and recall of 0.835 on a previously unseen holdout dataset under varying camera conditions, eye position, and video quality.Clinical relevance- Functional mapping studies in psychiatry and neuro-development which rely on a resting state fMRI protocol may yield divergent results depending on whether the subject keeps their eyes closed or open or whether the subject falls asleep. The clinical relevance of this work is to introduce the eye state (closed or open) in brain imaging studies as a prospective covariate, and as a feature that can potentially control for sleep state as a confounding factor.

Quality and Reliability of Halitosis Videos on YouTube as a Source of Information.

Halitosis is a common condition that adversely affects quality of life. Through YouTube, people access oral health information. This study aimed to analyze the quality and comprehensiveness of the content, and reliability of halitosis videos hosted on YouTube. In total, 300 videos were collected based on three search keywords, and the number of likes, dislikes, views, duration, interaction index, viewing rate, and source category. Subsequently, the video score, which represents the content quality and comprehensiveness was used to categorize videos as poor, good, and excellent. DISCERN score was used to assess video reliability. Of the 105 videos analyzed, 68 (64.8%) were uploaded by personal users. In general, videos were categorized as poor and have low reliability. Videos sourced from healthcare professionals showed the highest quality, comprehensiveness of the content, and reliability (p < 0.05). There was no difference in the number of viewers for video duration less than or more than 4 min (p > 0.05). YouTube users prefer low-quality videos over high-quality ones, indicating that viewers cannot differentiate between reliable and potentially biased content. It is critical to enable viewers to critically assimilate information hosted on YouTube to make effective oral healthcare decisions.

YouTube as a source of information on COVID-19 vaccination in rheumatic diseases

As the most well-known and popular video-sharing platform around the world, YouTube is an influential tool for the dissemination of health-related information. In addition, considering the increase in obtaining information from internet-based sources in pandemic conditions, YouTube has become more important in the presentation of information related to COVID-19. Therefore, the aim of this study was to evaluate videos related to COVID-19 vaccination in rheumatic diseases (RD) on YouTube. In this descriptive study, 334 video URLs listed with six search terms were recorded (26 July 2021). Three quality groups (high, intermediate, and low) were created based on the Global Quality Scores (GQS). Video sources were identified and various video parameters were compared between the quality groups. Following the implementation of the exclusion criteria, 56 videos remained for further analysis; of which 37 (66.07%) were evaluated as high quality, 12 (21.42%) as intermediate quality, and 7 (12.51%) as low quality. No significant difference was determined between the quality groups in per day values of views, likes, dislikes, and comments. The sources of high-quality videos were pharmaceutical company (n = 1; 100%), pharmacist (n = 1; 100%), society-organization (n = 17; 85%), and academic (n = 3; 75%). Although two-thirds of the videos were high quality, it should be kept in mind that intermediate and low-quality videos are also available. Users should not assume the quality of the videos based on the number of views, likes, dislikes, and comments, but should focus more on video sources.

Are YouTube videos enough to learn anakinra self-injection?

Anakinra is a drug that can be administered subcutaneously as a self-injection in both children and adults. We aimed to evaluate the content, reliability, and quality of the videos most viewed on “anakinra self-injection” on YouTube, which is an easily accessible source of information. We performed a YouTube search using the keywords “anakinra”, “anakinra injection”, and “anakinra self-injection” in addition to the generic and commercial names of the biologic agent in September 2021. The quality and reliability of the videos were assessed according to the Global Quality Score (GQS) and DISCERN score. Video power index was used to assess both the view and the like ratio of the videos. A total of 51 videos were analyzed, a majority of which were uploaded by physicians (56.9%). The median GQS was 3 and total DISCERN score was 49. According to the GQS, 21.6% of the videos were of low quality, 35.3% were of fair quality, and 43.1% were of high quality. High-quality videos had higher DISCERN scores and longer duration of videos (p < 0.05). 41 (80.4%) videos were categorized as useful information, and 8 (15.7%) as useful as per patients’ opinion. Further, GQS and DISCERN scores of videos that had useful information were significantly higher. There are numerous YouTube videos with helpful information that can be a source of knowledge on the safe and correct technique of daily anakinra self-injection for both adults and children. There was no significant difference in patient interaction between useful and misleading videos. This indicates that patients do not differentiate between high- and low-quality videos.Supplementary InformationThe online version contains supplementary material available at 10.1007/s00296-021-04999-w.

YouTube as a Source of Patient Information Regarding Exercises and Compensated Maneuvers for Dysphagia.

Objective: Acquiring online health-related information has become increasingly widespread. In this study, we aimed to evaluate the quality of the most-viewed YouTube videos on dysphagia regarding exercises and compensated maneuvers. Method: We searched for the keywords “dysphagia exercise”, “dysphagia rehabilitation”, “dysphagia maneuver”, “dysphagia therapy”, and “dysphagia compensation” on YouTube on 5 February 2021. The educational quality of videos on YouTube was investigated based on the Global Quality Scale (GQS) and categorized into three groups: high-, intermediate-, and low-quality. The modified DISCERN tool was used to evaluate the reliability of the YouTube videos. Video parameters were compared between the groups according to the quality of the videos. Results: Of the 51 videos evaluated, according to the GQS, 54.9% (n = 28) were of high-quality, 35.3% (n = 18) were of intermediate-quality, and 9.8% (n = 5) were of low-quality, respectively. When the video parameters were compared among the groups, there were no significant differences in the number of views, likes, dislikes, or comments per day (p > 0.05). However, there was a significant difference in the DISCERN scores between the groups (p < 0.001). Conclusion: YouTube can be deemed as a predominant source for high-quality videos on dysphagia exercise and compensated maneuvers. However, YouTube should be accepted as a mixed pool, with high-, intermediate-, and low-quality videos. Therefore, healthcare professionals, such as physicians and therapists, should verify the suitability and quality of the video, and suggest it to the patient, to ensure that the patient obtains the appropriate information.

Analysis of the Wavelet Domain Filtering Approach for Video Super-Resolution

The wavelet domain-centered algorithms for the super-resolution research area give better visual quality and have been explored by different researchers. The visual quality is achieved with increased complexity and cost as most of the systems embed different pre- and post-processing techniques. The frequency and spatial domain-based methods are the usual approaches for super-resolution with some benefits and limitations. Considering the benefits of wavelet domain processing, this paper deals with a new algorithm that depends on wavelet residues. The methodology opts for wavelet domain filtering and residue extraction to get super-resolved frames for better visuals without embedding other techniques. The avoidance of noisy high-frequency components from low-quality videos and the consideration of edge information in the frames are the main targets of the super-resolution process. This inverse process is carried with a proper combination of information present in low-frequency bands and residual information in the high-frequency components. The efficient known algorithms always have to sacrifice simplicity to achieve accuracy, but in the proposed algorithm efficiency is achieved with simplicity. The robustness of the algorithm is tested by analyzing different wavelet functions and at different noise levels. The proposed algorithm performs well in comparison to other techniques from the same domain.

Impact of video quality when evaluating video-assisted cardiopulmonary resuscitation: a randomized, controlled simulation trial

BackgroundAlthough not routinely established during cardiopulmonary resuscitation (CPR), video-assisted CPR has been described as beneficial in the communication with emergency medical service (EMS) authorities in out-of-hospital cardiac arrest scenarios. Since the influence of video quality has not been investigated systematically and due to variation of quality of a live-stream video during video-assisted CPR, we investigated the influence of different video quality levels during the evaluation of CPR performance in video sequences.MethodsSeven video sequences of CPR performance were recorded in high quality and artificially reduced to medium and low quality afterwards. Video sequences showed either correct CPR performance or one of six typical errors: too low and too high compression rate, superficial and increased compression depth, wrong hand position and incomplete release. Video sequences were randomly assigned to the different quality levels. During the randomised and double-blinded evaluation process, 46 paramedics and 47 emergency physicians evaluated seven video sequences of CPR performance in different quality levels (high, medium and low resolution).ResultsOf 650 video sequences, CPR performance was evaluable in 98.2%. CPR performance was correctly evaluated in 71.5% at low quality, in 76.8% at medium quality, and in 77.3% at high quality level, showing no significant differences depending on video quality (p = 0.306). In the subgroup analysis, correct classification of increased compression depth showed significant differences depending on video quality (p = 0.006). Further, there were significant differences in correct CPR classification depending on the presented error (p < 0.001). Allegedly errors, that were not shown in the video sequence, were classified in 28.3%, insignificantly depending on video quality. Correct evaluation did not show significant interprofessional differences (p = 0.468).ConclusionVideo quality has no significant impact on the evaluation of CPR in a video sequence. Even low video quality leads to an acceptable rate of correct evaluation of CPR performance. There is a significant difference in evaluation of CPR performance depending on the presented error in a video sequence.Trial registrationGerman Clinical Trial Register (Registration number DRKS00015297) Registered on 2018-08-21.

The benefit of online education in twin-to-twin transfusion syndrome: could it be an eye-opening platform?

Objective This study intends to objectively review the quality and reliability of laser photocoagulation for twin-to-twin transfusion syndrome (TTTS) videos on the YouTube® platform in terms of integrity and educational value for patients and medical professionals. Methods YouTube® (http://www.youtube.com) search was performed for videos pertaining to fetal laser surgery in TTTS, using the terms fetoscopic laser ablation or/and fetoscopic laser surgery or/and fetoscopic laser twin-to-twin transfusion syndrome. A total of 42 videos were analyzed. Video sources were categorized as follows: INDIVIDUAL, ACADEMIC INSTITUTION, and COMMERCIAL. Source of videos, time since upload (days), total duration of videos (seconds), total views, and likes were recorded. Video power index (VPI) was used to report video popularity. Educational quality and integrity were reviewed using the DISCERN, JAMA, GQS, and a novel YouTube® TTTS content - integrity score (TTTS-IS). Results The mean duration was 336.43±351.8 seconds, and the total views were 172.076±82.6. Surgical risks were stated only in 33% of videos, and the success rate of surgery was mentioned in 35%. In all groups, the median value of DISCERN scores is 32.50, 42, and 26, which refers to poor, fair, and very poor quality, respectively. Videos uploaded by an academic institution had statistically insignificant DISCERN and JAMA scores along with lower GQS and TTTS-IS (p&gt;0.05). In regards to the video source, academic institutions were prominent (47.6%), and the most popularized videos were conducted by commercial websites and individuals related to VPI. The VPI was poorly correlated with all scoring systems. Negative correlations were detected between total views and JAMA scores and positive correlations were determined with the GQS and TTTS-IS. Conclusion Despite having VPI score highly in each group, YouTube® provides low-quality videos regarding fetal laser surgery of TTTS. There is an urgent need to regulate the context of those videos according to medical guidelines.

Deep-HR: Fast heart rate estimation from face video under realistic conditions

This paper presents a novel method for remote heart rate (HR) estimation. Recent studies have proved that blood pumping by the heart is highly correlated to the intense color of face pixels, and surprisingly can be utilized for remote HR estimation. Researchers successfully proposed several methods for this task, but making it work in realistic situations is still a challenging problem in computer vision community. Furthermore, learning to solve such a complex task on a dataset with very limited annotated samples is not reasonable. Consequently, researchers do not prefer to use the deep learning approaches for this problem. In this paper, we propose a simple yet efficient approach to benefit the advantages of the Deep Neural Network (DNN) by simplifying HR estimation from a complex task to learning from very correlated representation to HR. Inspired by previous work, we learn a component called Front-End (FE) to provide a discriminative representation of face videos, afterward a light deep regression auto-encoder as Back-End (BE) is learned to map the FE representation to HR. Regression task on the informative representation is simple and could be learned efficiently on limited training samples. Beside of this, to be more accurate and work well on low-quality videos, two deep encoder–decoder networks are trained to refine the output of FE. We also introduce a challenging dataset (HR-D) to show that our method can efficiently work in realistic conditions. Experimental results on HR-D and MAHNOB datasets confirm that our method could run as a real-time method and estimate the average HR better than state-of-the-art ones.

Educational quality of YouTube videos on musculoskeletal ultrasound.

Introduction/objectivesA progressively growing number of rheumatologists have integrated musculoskeletal ultrasound (MSUS) as a prized tool in their daily clinical practice over the past two decades. Lately, YouTube is often used by healthcare professionals as an educational tool. This study aimed to analyze the quality of educational information available on YouTube pertaining to MSUS.MethodWe performed a search using the keyword “musculoskeletal ultrasound” on YouTube in January 2021. Of the first 200 videos screened, 147 met the inclusion criteria. The quality and reliability of the videos were evaluated using the global quality score (GQS) and modified DISCERN score, and their popularity was assessed using the Video Power Index (VPI).ResultsOf the 147 videos evaluated, 58 (39.5%) were of high quality, 30 (20.4%) were of moderate quality, while 59 (40.1%) were of low quality. The modified DISCERN score and the duration of the videos were significantly higher in the high-quality videos than the moderate- and low-quality videos (p < 0.001 for all). The median number of views per day, likes, comments, and the popularity of the videos assessed with VPI were significantly higher in the high- and moderate-quality group compared to the low-quality videos (p < 0.001, p < 0.001, p = 0.007, and p < 0.001, respectively).ConclusionThe quality of YouTube videos as an educational source on MSUS is unsatisfactory. Professional organizations and physicians should consider creating more high-quality MSUS videos. YouTube should consider filtering out poor-quality videos in collaboration with experts in the field.Key Points• Although rheumatologists are increasingly using ultrasound as an extension of their exam, accessing musculoskeletal ultrasound training is still facing some challenges.• YouTube is commonly used as an educational source by medical students and healthcare professionals.• Approximately 60% of the videos related to musculoskeletal ultrasound on YouTube revealed high and moderate quality, while the remaining 40% were of low quality and not suitable for medical education.• A filtering mechanism is required to improve the quality of YouTube content on MSUS.