,
22 tweets,
9 min read
Read on Twitter
Medical visualizations (vizs; #visualAbstract, #infographic, or #graphicalAbstract) are ⬆︎popular. In this tweetorial I'm going to detail how I make vizs to help you in your #meded/#FOAMed endeavors. Here's my portfolio for background: twitter.com/i/moments/1057…
There are four large domains that I consider when making a viz: Design, Structure, Content, and Textual Context (mostly in the parent tweet).
Let's go through each of these and discuss various facets that one might want to consider when designing a medical viz
Let's go through each of these and discuss various facets that one might want to consider when designing a medical viz
Design: Audience
Know your audience, particularly if your audience is one that *doesn't* have access to the article that inspired your viz. If they don't, then recognize that your viz may be the 1st and only exposure to the science that the reader will have.
Know your audience, particularly if your audience is one that *doesn't* have access to the article that inspired your viz. If they don't, then recognize that your viz may be the 1st and only exposure to the science that the reader will have.
Lots of readers don't have access to the article (high individual/institutional costs or slow/inconsistent internet access). Your viz must contain enough detail to allow the reader to make a bedside/chairside clinical decision in the absence of the original article.
Here's an example of a viz I made in which I added as much detail as possible to allow the healthcare professional to make a clinical decision in the absence of article access. Don't be afraid to add more detail if you forsee challenges in reading the primary literature. 
Design: Size
Stick with 8.5"x11" size (standard paper size). While you can make the viz wider and/or longer, such viz become harder to read and require multiple zoom-in/zoom-out finger gestures. Most mobile devices can fully display an 8.5"x11" image.
Stick with 8.5"x11" size (standard paper size). While you can make the viz wider and/or longer, such viz become harder to read and require multiple zoom-in/zoom-out finger gestures. Most mobile devices can fully display an 8.5"x11" image.
Design: Orientation
Portrait or Landscape? Portrait is favorable because the reader can view the viz with one🤚🏻. Landscape is favorable if you want to add more than 1 study to your viz. Pic A is a portrait viz of one article; Pic B is a landscape viz of 2 articles.
Portrait or Landscape? Portrait is favorable because the reader can view the viz with one🤚🏻. Landscape is favorable if you want to add more than 1 study to your viz. Pic A is a portrait viz of one article; Pic B is a landscape viz of 2 articles.
Design: Definition
English is considered the language of medical science, but pics/emojis/icons is not. Define your non-text "language" using color, size, shape, or a small piece of descriptive text. Here's a viz in which I use the above to define the items in my viz.
English is considered the language of medical science, but pics/emojis/icons is not. Define your non-text "language" using color, size, shape, or a small piece of descriptive text. Here's a viz in which I use the above to define the items in my viz.
Design: Definition
In the above viz I used 3 colors to define each trial (red, blue, green). I defined randomized and statistically significant at the bottom. I used black color to define connections/sections of the viz. I used a black or white stomach to define gastric acidity.
In the above viz I used 3 colors to define each trial (red, blue, green). I defined randomized and statistically significant at the bottom. I used black color to define connections/sections of the viz. I used a black or white stomach to define gastric acidity.
Design: Font
Serif v Sans Serif? I recommend the latter because it is easier to read (e.g., Helvetica, Tahoma, Calibri). Serif has overhangs on each letter, which adds to the beauty of the text but can be cumbersome to read.
Serif v Sans Serif? I recommend the latter because it is easier to read (e.g., Helvetica, Tahoma, Calibri). Serif has overhangs on each letter, which adds to the beauty of the text but can be cumbersome to read.
Structure
So what do you want to include in your medical viz? I approach this question the same way as I would a manuscript: is there a background/context section, methods, results, and conclusions section of the viz? A good medical viz touches on each of these to give....
So what do you want to include in your medical viz? I approach this question the same way as I would a manuscript: is there a background/context section, methods, results, and conclusions section of the viz? A good medical viz touches on each of these to give....
...the reader a full picture of the study. Each is important to include if your reader won't have access to the article (hence it's important to know your audience).
Here's a viz that provides a lot of context so the reader understands the value of the study better.
Here's a viz that provides a lot of context so the reader understands the value of the study better.
Content: White space is your friend
Often we want to pack as much info into a viz as possible. While this has advantages, there is a limit beyond which the lack of "white/blank space" can deter the reader or make it harder to convey your message. One needs to..
Often we want to pack as much info into a viz as possible. While this has advantages, there is a limit beyond which the lack of "white/blank space" can deter the reader or make it harder to convey your message. One needs to..
..balance the information provided with strategic "breaks" so that the reader can catch up. A good way to ascertain this is by calculating the "black-white" ratio of your proposed viz. Let's look at this closely and learn how to calculate it.
Content: Black-White ratio (B/W)
The B/W ratio converts all colored text/icons/emojis/images into black pixels. Blank areas of your viz are converted to white pixels. Divide the two to get a ratio that indicates the amount of *information density* in your viz.
The B/W ratio converts all colored text/icons/emojis/images into black pixels. Blank areas of your viz are converted to white pixels. Divide the two to get a ratio that indicates the amount of *information density* in your viz.
Content: B/W
Too low an information density score and you haven't conveyed enough info to the reader to allow for a clinical decision to be made. Too high and you have overwhelmed the reader.
To calculate B/W, download ImageJ: a free software from the @NIH imagej.nih.gov/ij/
Too low an information density score and you haven't conveyed enough info to the reader to allow for a clinical decision to be made. Too high and you have overwhelmed the reader.
To calculate B/W, download ImageJ: a free software from the @NIH imagej.nih.gov/ij/
Content: Info Density & ImageJ
Launch ImageJ and open your proposed medical viz file (must be in JPG or PNG format). Once you see your image, you'll need to convert it to a binary format (either black or white). See the video to learn how to do that.
Launch ImageJ and open your proposed medical viz file (must be in JPG or PNG format). Once you see your image, you'll need to convert it to a binary format (either black or white). See the video to learn how to do that.
Content: Info Density & ImageJ
Now you have your proposed medical viz in black and white pixels. Black is the content and white is the blank space. To calculate the ratio, you need the number of pixels of each color. See the video to learn how to do that.
Now you have your proposed medical viz in black and white pixels. Black is the content and white is the blank space. To calculate the ratio, you need the number of pixels of each color. See the video to learn how to do that.
Content: Info Density
The density for my viz is 0.107. Is that too low/high or just right? There isn't a threshold level as yet, but a group of researchers led by @virenkaul are creating a model that incorporates the info density score to predict engagement on #SoMe. For now...
The density for my viz is 0.107. Is that too low/high or just right? There isn't a threshold level as yet, but a group of researchers led by @virenkaul are creating a model that incorporates the info density score to predict engagement on #SoMe. For now...
..I'd recommend an information density (B/W) score ≤ 0.5 and wait for the regression model to be published.
Textual Content
A few points to address in the tweet/facebook post:
• link to the article
• tag the journal & any authors who are on #SoMe
• briefly describe the article & viz
• use one of the 3 hashtags: #infographic, #visualAbstract, #graphicalAbstract
A few points to address in the tweet/facebook post:
• link to the article
• tag the journal & any authors who are on #SoMe
• briefly describe the article & viz
• use one of the 3 hashtags: #infographic, #visualAbstract, #graphicalAbstract
