Dataset Information

Automated processing of social media content for radiologists: applied deep learning to radiological content on twitter during COVID-19 pandemic.

ABSTRACT:

Purpose

The purpose of this study was to develop an automated process to analyze multimedia content on Twitter during the COVID-19 outbreak and classify content for radiological significance using deep learning (DL).

Materials and methods

Using Twitter search features, all tweets containing keywords from both "radiology" and "COVID-19" were collected for the period January 01, 2020 up to April 24, 2020. The resulting dataset comprised of 8354 tweets. Images were classified as (i) images with text (ii) radiological content (e.g., CT scan snapshots, X-ray images), and (iii) non-medical content like personal images or memes. We trained our deep learning model using Convolutional Neural Networks (CNN) on training dataset of 1040 labeled images drawn from all three classes. We then trained another DL classifier for segmenting images into categories based on human anatomy. All software used is open-source and adapted for this research. The diagnostic performance of the algorithm was assessed by comparing results on a test set of 1885 images.

Results

Our analysis shows that in COVID-19 related tweets on radiology, nearly 32% had textual images, another 24% had radiological content, and 44% were not of radiological significance. Our results indicated a 92% accuracy in classifying images originally labeled as chest X-ray or chest CT and a nearly 99% accurate classification of images containing medically relevant text. With larger training dataset and algorithmic tweaks, the accuracy can be further improved.

Conclusion

Applying DL on rich textual images and other metadata in tweets we can process and classify content for radiological significance in real time.

SUBMITTER: Khurana S

PROVIDER: S-EPMC7811945 | biostudies-literature | 2021 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Automated processing of social media content for radiologists: applied deep learning to radiological content on twitter during COVID-19 pandemic.

Khurana Shikhar S Chopra Rohan R Khurana Bharti B

Emergency radiology 20210118 3

<h4>Purpose</h4>The purpose of this study was to develop an automated process to analyze multimedia content on Twitter during the COVID-19 outbreak and classify content for radiological significance using deep learning (DL).<h4>Materials and methods</h4>Using Twitter search features, all tweets containing keywords from both "radiology" and "COVID-19" were collected for the period January 01, 2020 up to April 24, 2020. The resulting dataset comprised of 8354 tweets. Images were classified as (i) ...[more]

PMID: 33459907

Similar Datasets

Project description:BackgroundThe COVID-19 pandemic has drawn attention to various inequalities in global societies, highlighting discrepancies in terms of safety, accessibility, and overall health. In particular, sex workers are disproportionately at risk due to the nature of their work and the social stigma that comes alongside it.ObjectiveThis study examines how public social media can be used as a tool of professional and personal expression by sex workers during the COVID-19 pandemic. We aimed to explore an underresearched topic by focusing on sex workers' experiences with the ongoing COVID-19 pandemic on the social media platform Twitter. In particular, we aimed to find the main issues that sex workers discuss on social media in relation to the COVID-19 pandemic.MethodsA literature review followed by a qualitative analysis of 1458 (re)tweets from 22 sex worker Twitter accounts was used for this study. The tweets were qualitatively coded by theme through the use of intercoder reliability. Empirical, experimental, and observational studies were included in this review to provide context and support for our findings.ResultsIn total, 5 major categories were identified as a result of the content analysis used for this study: concerns (n=542, 37.2%), solicitation (n=336, 23.0%), herd mentality (n=231, 15.8%), humor (n=190, 13.0%), and blame (n=146, 10.0%). The concerns category was the most prominent category, which could be due to its multifaceted nature of including individual concerns, health issues, concerns for essential workers and businesses, as well as concerns about inequalities or intersectionality. When using gender as a control factor, the majority of the results were not noteworthy, save for the blame category, in which sexual and gender minorities (SGMs) were more likely to post content.ConclusionsThough there has been an increase in the literature related to the experiences of sex workers, this paper recommends that future studies could benefit from further examining these 5 major categories through mixed methods research. Examining this phenomenon could recognize the challenges unique to this working community during the COVID-19 pandemic and potentially reduce the widespread stigma associated with sex work in general.

Project description:BACKGROUND:Although gun violence has been identified as a major public health concern, the scope and significance of internet gun advertising is not known. OBJECTIVE:This study aimed to quantify the characteristics of gun advertising on social media and to compare the reach of posts by manufacturers with those of influencers. METHODS:Using a systematic search, we created a database of recent and popular Twitter and YouTube posts made public by major firearm manufacturers and influencers. From our sample of social media posts, we reviewed the content of the posts on the basis of 19 different characteristics, such as type of gun, presence of women, and military or police references. Our content analysis summarized statistical differences in the information conveyed in posts to compare advertising approaches across social media platforms. RESULTS:Sample posts revealed that firearm manufacturers use social media to attract audiences to websites that sell firearms: 14.1% (131/928; ±2.9) of Twitter posts, 53.6% (228/425; ±6.2) of YouTube videos, and 89.5% (214/239; ±5.1) of YouTube influencer videos link to websites that facilitate sales. Advertisements included women in efforts to market handguns and pistols for the purpose of protection: videos with women included protection themes 2.5 times more often than videos without women. Top manufacturers of domestic firearms received 98 million channel views, compared with 6.1 billion channel views received by the top 12 YouTube influencers. CONCLUSIONS:Firearm companies use social media as an advertising platform to connect viewers to websites that sell guns. Gun manufacturers appropriate YouTube servers, video streaming services, and the work of YouTube influencers to reach large audiences to promote the widespread sale of consumer firearms. YouTube and Twitter subsidize gun advertising by offering server and streaming services at no cost to gun manufacturers, to the commercial benefit of Google and Twitter's corporate ownership.

Project description:BACKGROUND:Respecting patient privacy and confidentiality is critical for doctor-patient relationships and public trust in medical professionals. The frequency of potentially identifiable disclosures online during periods of active engagement is unknown. OBJECTIVE:The objective of this study was to quantify potentially identifiable content shared on social media by physicians and other health care providers using the hashtag #ShareAStoryInOneTweet. METHODS:We accessed and searched Twitter's API using Symplur software for tweets that included the hashtag #ShareAStoryInOneTweet. We identified 1206 tweets by doctors, nurses, and other health professionals out of 43,374 tweets shared in May 2018. Tweet content was evaluated in January 2019 to determine the incidence of instances where names or potentially identifiable information about patients were shared; content analysis of tweets in which information about others had been disclosed was performed. The study also evaluated whether participants raised concerns about privacy breaches and estimated the frequency of deleted tweets. The study used dual, blinded coding for a 10% sample to estimate intercoder reliability using Cohen ? statistic for identifying the potential identifiability of tweet content. RESULTS:Health care professionals (n=656) disclosing information about others included 486 doctors (74.1%) and 98 nurses (14.9%). Health care professionals sharing stories about patient care disclosed the time frame in 95 tweets (95/754, 12.6%) and included patient names in 15 tweets (15/754, 2.0%). It is estimated that friends or families could likely identify the clinical scenario described in 242 of the 754 tweets (32.1%). Among 348 tweets about potentially living patients, it was estimated that 162 (46.6%) were likely identifiable by patients. Intercoder reliability in rating the potential identifiability demonstrated 86.8% agreement, with a Cohen ? of 0.8 suggesting substantial agreement. We also identified 78 out of 754 tweets (6.5%) that had been deleted on the website but were still viewable in the analytics software data set. CONCLUSIONS:During periods of active sharing online, nurses, physicians, and other health professionals may sometimes share more information than patients or families might expect. More study is needed to determine whether similar events arise frequently and to understand how to best ensure that patients' rights are adequately respected.

Project description:BackgroundFalse claims about COVID-19 vaccines can undermine public trust in ongoing vaccination campaigns, posing a threat to global public health. Misinformation originating from various sources has been spreading on the web since the beginning of the COVID-19 pandemic. Antivaccine activists have also begun to use platforms such as Twitter to promote their views. To properly understand the phenomenon of vaccine hesitancy through the lens of social media, it is of great importance to gather the relevant data.ObjectiveIn this paper, we describe a data set of Twitter posts and Twitter accounts that publicly exhibit a strong antivaccine stance. The data set is made available to the research community via our AvaxTweets data set GitHub repository. We characterize the collected accounts in terms of prominent hashtags, shared news sources, and most likely political leaning.MethodsWe started the ongoing data collection on October 18, 2020, leveraging the Twitter streaming application programming interface (API) to follow a set of specific antivaccine-related keywords. Then, we collected the historical tweets of the set of accounts that engaged in spreading antivaccination narratives between October 2020 and December 2020, leveraging the Academic Track Twitter API. The political leaning of the accounts was estimated by measuring the political bias of the media outlets they shared.ResultsWe gathered two curated Twitter data collections and made them publicly available: (1) a streaming keyword-centered data collection with more than 1.8 million tweets, and (2) a historical account-level data collection with more than 135 million tweets. The accounts engaged in the antivaccination narratives lean to the right (conservative) direction of the political spectrum. The vaccine hesitancy is fueled by misinformation originating from websites with already questionable credibility.ConclusionsThe vaccine-related misinformation on social media may exacerbate the levels of vaccine hesitancy, hampering progress toward vaccine-induced herd immunity, and could potentially increase the number of infections related to new COVID-19 variants. For these reasons, understanding vaccine hesitancy through the lens of social media is of paramount importance. Because data access is the first obstacle to attain this goal, we published a data set that can be used in studying antivaccine misinformation on social media and enable a better understanding of vaccine hesitancy.

Project description:BackgroundSocial media is often used for health communication and can facilitate fast information exchange. Despite its increasing use, little is known about child health information sharing and engagement over social media.ObjectiveThe primary objectives of this study are to systematically describe the content of social media posts about child pain and sleep and identify the level of research evidence in these posts. The secondary objective is to examine user engagement with information shared over social media.MethodsTwitter, Instagram, and Facebook were searched by members of the research team over a 2-week period using a comprehensive search strategy. Codes were used to categorize the content of posts to identify the frequency of content categories shared over social media platforms. Posts were evaluated by content experts to determine the frequency of posts consistent with existing research evidence. User engagement was analyzed using Netlytic, a social network analysis program, to examine visual networks illustrating the level of user engagement.ResultsFrom the 2-week period, nearly 1500 pain-related and 3800 sleep-related posts were identified and analyzed. Twitter was used most often to share knowledge about child pain (639/1133, 56.40% of posts), and personal experiences for child sleep (2255/3008, 75.00% of posts). For both topics, Instagram posts shared personal experiences (53/68, 78% pain; 413/478, 86.4% sleep), Facebook group posts shared personal experiences (30/49, 61% pain; 230/345, 66.7% sleep) and Facebook pages shared knowledge (68/198, 34.3% pain; 452/1026, 44.05% sleep). Across platforms, research evidence was shared in 21.96% (318/1448) of pain- and 9.16% (445/4857) of sleep-related posts; 5.38% (61/1133) of all pain posts and 2.82% (85/3008) of all sleep posts shared information inconsistent with the evidence, while the rest were absent of evidence. User interactions were indirect, with mostly one-way, rather than reciprocal conversations.ConclusionsSocial media is commonly used to discuss child health, yet the majority of posts do not contain research evidence, and user engagement is primarily one-way. These findings represent an opportunity to expand engagement through open conversations with credible sources. Research and health care communities can benefit from incorporating specific information about evidence within social media posts to improve communication with the public and empower users to distinguish evidence-based content better. Together, these findings have identified potential gaps in social media communication that may be informative targets to guide future strategies for improving the translation of child health evidence over social media.

Dataset Information

Automated processing of social media content for radiologists: applied deep learning to radiological content on twitter during COVID-19 pandemic.

Purpose

Materials and methods

Results

Conclusion

Publications

Automated processing of social media content for radiologists: applied deep learning to radiological content on twitter during COVID-19 pandemic.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets