Medical Imaging Assistant Basic Features

Our project that deals with the use of a recently proposed technique in literature: Deep Convolutional Neural Networks (CNN). These deep networks will incorporate information from several different modes by a User Interface (UI) with AI-Assisted techniques behind. Therefore, we aim to develop a Medical Imaging (MI) Assistant that will assist physicians across diagnosis. The requirements are as follows: 1. Dataset Reader; 2. Assistant Animation; 3. Message Notification; 4. Medical Imaging Answer Animation; For the issue (1), the goal is to create the first Dataset Reader feature. This feature will be of chief importance since it will read the dataset resulted from the Machine Learning (ML) algorithms and will provide the structure information on a dataset from the results. On the issue (2), the goal is to create the new assistant animation. A wanted simulation for this feature can be watched here. The idea can be to do this animation using HTML, CSS and JavaScript technologies, however, if considered, we can do in whatever would be better. On the other hand, the issue (3) aims to create the Message Notification feature source. This feature will implement the Dataset Reader, the issue (1), results and will give the user visualization of it. However, is important to underline that this message feature will have several templates, and, therefore, must be implemented with this level of scale. Prepared to receive as many templates as possible. Finally, the issue (4), we aim to create the Medical Imaging Answer Animation new feature. For this feature, we need to think deeper, while it can be postponed on this phase. However, if considered, the idea here is to implement some kind of "detailed answer" for the BIRADS. It means that when the user wants to know why that value of the BIRADS was given by the machine, the Assistant will note where are the lesions, or at least, where is probable to have a lesion on the image. Another point, that we also need to deeper think, could be some patterns of the clinical covariates that can trigger a high cancer probability. We need to think about how to accurately give this information to the users.