Brittany Wenger
Brittany Wenger, a high-school senior, is well on her way to making the diagnosis of breast cancer less painful and more accurate. Wenger began studying neural networks when she was in the seventh grade. And this year, she won the grand prize in the 2012 Google Science Fair for her project, “Global Neural Network Cloud Service for Breast Cancer.” The resulting Cloud4Cancer service aggregates data from biopsies done with the fine-needle aspiration process, instead of the traditional and more painful surgical option. The goal of Wenger’s network is to ensure that fine-needle aspiration biopsies are as accurate as traditional biopsies, which would allow more women to be diagnosed and treated early. In preliminary trials, the service has achieved 99.11 percent sensitivity to malignancy, and its over 7.6 million trials have demonstrated the network improves as more data is deposited. And because it is delivered as a cloud service, Cloud4Cancer is built to support usage by every hospital in the world.
14 Responses to Brittany Wenger
-
-
Booger Francais says:
You can find the proportion of false negatives here on step 8.
-
-
Shubh says:
Hi Brittany,
Congratulations on winning the Google Science Fair 2013… and I really think you’ve done a great job for society. In fact, I am also interested in coding and right now I am still learning Html on codeacademy. Can you please explain to me how you made an app that could think a bit like the human brain? I want to understand how that works. My knowledge about coding is still limited because I have only been doing this for a few months and I am in the 8th grade…
Thanks,
Shubh-
Nicholas Pequeno says:
@Shubh
I recommend the online ‘course’
Stanford’s Machine Learning on coursera.org
Also checkout udacity.com for programming courses in general.
Brittany’s network actually seems pretty ‘simple’. It’s just very well designed. Implementing it should be fairly easy with an adequate knowledge of programming and artificial neural networks. (and in this case, cloud resources)I wish I could meet Brittany in person. I’ve been working with similar topics (ANNs and cloud computing). I just haven’t started on anything as profound as this, and I’m about a year older. (and not a woman)
-
-
Hi Snubh,
Right now, GSF 2013′s submission is running (https://www.googlesciencefair.com/). To start with Artificial Intelligence computer programming, I recommend you to learn JAVA /programming language/ 1st. More info, look here – http://docs.oracle.com/javase/tutorial/java/index.html.
I’m sure you’ve also chance to change the world like her.
Good luck to your future endeavors,Programming Enthusiast
From Mongolia,
-
Hi RJ,
I personally wouldn’t recommend starting with Java to anyone. It’s horrendously complex. I would recommend something like Python to start with. Python has some awesome AI and stats modules such as statsmodels, scikits-learn, and pybrain, but beyond that it has a simple, clean syntax and is very easy to learn. It is also platform independent, as is Java. Definitely something you can move on to after learning the basic concepts using Python.
Cheers,
Steve
-
-
marilyn raymond, M.D. says:
Brittany, Congratulations.I am an Oncologist in Florida. I focus my pracyice on Breast Cancer and a friend forwarded me your link. Your Cloud is brilliant !! I hope you continue to work in the field of Oncology. We need bright, intuitive thinkers like you. I wish you the best and will look forward to seeing your name again in the head lines . Bravo.
-
Buford Olivarri says:
Although one programmer has the necessary skills and knowledge to work competently on a problem or even create a program, he or she can only do so much. Creating the source code for an operating system, for example, will require thousands of manhours from a single programmer and most probably, he or she will only be halfway through. There just isn’t enough time for one or even two programmers to work effectively to produce a usable program.”"
Keep it up
<http://www.caramoan.co/ -
Duffy Godshall says:
Hello Ms. Wenger
I’m a TED fan and a fan of your work. Just thought I’d comment. I’ll keep it brief.
When working on my online MIT classes and getting discouraged, I’ll think of you. You inspire me. Thank you for that. I hope to make the world a better place too.
Best wishes,
Duf. -
The Grim Snark says:
Kids these days…
-
Rhonda H in Virginia says:
Check out Brittnay’s code for yourself ( http://cloud4cancer.appspot.com/ )
-
Hurbault Julien says:
Brittany,
If I understood you correctly, your first goal was to provide a tool to enhance the cytologic evaluation (less expensive, less invasive, but also less accurate for the moment compared to the biopsie evaluation) in the diagnosis of breath cancer, i.e. is the tumor malignant or benign.
You’re saying, as a result, that your software has a sensitivity of 99,1% :
* As Caroline Rowlands was asking, What is the specificity of it ? A good sensitivity is required for a screening test (and fine needle test would be kind of really invasive for a screening test), but for a Diagnosis you need a good Specificity.
* The attributes (inputs) are still coming from Human interpretation of a sample (such as the attribute about the cell size. How do take these consideration ?
* What was the process, precisely, to evaluate your test ? For example, do you consider in your result the variation due to the fine needle procedure (how it’s gonna be in rea life..)..?
By the way, I’m impressed by your maturity and your skills, bravo.
Julien H.
-
Were the commercial neural network products specifically for cancer detection or were they more general use?
Also did all the products have the same input parameters?You must be really smart to understand neural networks at the age of 17. I still can’t and I’m 22 ha!
-
Oh ok. All the inputs are human given. That’s still really cool. I have to wonder if it’s be more effective on average than human diagnosis.
-
You do not mention the specificity of your test. I suspect it is very low?