Medical Imaging Platform Undo Redo Annotating Feature & Priorities List - 09/01/2020 17:09 EST

Budget: 60$ - 100$ This project proposes a tiny improvement of our already full implemented User Interface (UI) which is a complete platform on a Medical Imaging (MI) scope. The purpose of this UI is to provide clinicians an efficient interaction among the most used MI technologies. In this platform, clinicians can open the medical images and annotate the several lesions available on each image. The annotations can be a towfold: (1) masses [1]; and (2) calcifications [2]. For the (1) masses, our system has a mechanism providing clinicians with a drawing tool, which we called "Freehand" [3] feature. For the (2) calcifications, clinicians just need to do bullet points on the image, which we called "Probe" [4] feature. This is the first task (Image "[login to view URL]"). The second task (Image "[login to view URL]"), further described, will be to develop a list of priorities for each Patient ID in our assistant regarding several variables. Editing our medical images is a fundamental step to higher and better annotations dataset. One important part of this use case is the need for REDO and UNDO our annotations among the image. Our purpose is to guarantee the reliability of those annotations so that we need to create an agile way to edit and improve the annotations. The goal of this project is to create a REDO and UNDO mechanism so that each clinician can edit their own annotations on the UI. The platform already has an arcaice machanism for that. However, it was not accuratly implemented and, therefore, is not robust enough. At the moment, when a clinician start the annotation, he/she can press the ESC button to cancel the last click. Altough, when the clinician closes and finish the annotation, the mechanism is not properly developed. Furthermore, this mechanism is not implemeted for the (2) calcifications. Regarding the second task, with the presented project we aim at exploring the acquired sets of data among system functionalities and features. The purpose is to develop a list of priorities (i.e., the results that are more important for the final BIRADS on a top-down order) for our decision model. Each priority has the percentage of the achieved value, resulted from computing our AI models. Users can manage the provided priorities so that, it is possible to re-configure them (see uploaded images). For instance, if the system considers wrongly the Shape as more important in comparison to Margin, the user can change and put the Margin on the top of the list. As stated, we need to develop a list of priorities for each Patient ID. For the purpose, we will use the JSON format as usual. Each Patient ID will be available on a foo/ folder, to be designated. For instance, inside the foo/ folder, we will have [login to view URL] file (see uploaded example), where 12345 is the Patient ID. Inside this file, the structure is simple (see next source code). For now, we will have an attr variable that represents the priority. Each attr priority will have the respective value associated. Both model_order and user_order are representing the priority order on the list. The lower the value, the higher the priority. This means that a "model_order": 0 will have top priority. The difference between model_order and user_order is that the model_order represents the attributed value that came from the one considered by the AI model. On the other hand, the user_order starts by default with the same value as model_order. However, when the user changes the order on the UI, it will change this user_order value. Note: The Shape and Margin values are patient-wise, instead of being lesion-wise. References [1] [login to view URL] [2] [login to view URL] [3] [login to view URL] [4] [login to view URL]