First we would like to thank ELC for providing an opportunity to participate in this event. For all of us it is an eye opener and awareness about breast cancer. We are very happy that we are part of this larger cause. Having Heard Dr. Beatriu and Dr. Sarah on their day to day problem we present to you our solution leveraging AI / Deep Learning for mammogram diagnosis. More importantly we also took a stab at NLP to convert medical terms to understandable terms by patients.
We have two main processing Engine 1st one is based on Deep learning which reads Diagnostic output image from various source like X-ray,MRI and ultra sound. This does ranking between High Risk,Medium Risk and low Risk. 2nd one is based on NLP which reads Doctor prescription and locate doctor term and present it in simple easy to understand text and images . This will use word cloud to be presented to user to click and follow other web portals with more details.
There is link between both the above processing so user can see result and Text with pin pointed portion of image. This help in locating terms with portion of images. Also if we have multiple image available it will create co-register/co-relation between those and present it in demographic format with proof which portion is coming from x-ray or which of the portion is coming from MRI/Ultra sound matches or overlapped . We used Python for both of them and executed code in aws utilizing various event/ lambda function .
We also took advantage for rekognition API to detect format of scanned images and used textract to get meta data out from prescription . With this Meta data we created word cloud to display to the user and fetched simple images/clinical user friendly definition.
Log in or sign up for Devpost to join the conversation.