Computer Vision News - August 2020
RSIP Vision makes pioneering work in these applications: we provide (for several years now) these innovative features to leading manufacturers of medical devices and help them embed those features in their applications, in order to enable their next breakthrough. One of these important applications is the video analysis in endoscopy : a very rich set of information comes from the video, but its exploitation was considered until recently a very difficult task, due tomotion and to the ambiguity of the images: previous technology was not perfectly able to say what are you looking at and even where the precise point being shown is found. Analyzing the video with our new AI technologies enables us to understand the actual flow of the procedure, locate the place being displayed and also give us precise information about the vicinity around the area inspected by the probe . The prominent models being used are built on the U-Net Convolutional Neural Network (CNN) to make a state-of-the- art segmentation of the image shown. For the video work, much evidence can be derived from LSTM (Long Short Term Memory) networks , a Recurrent Neural Network (RNN) architecture used in deep learning and able to give information along the sequence of the images. Adjusting an LSTM for a specific task requires the work of an expert, but after running the model results are often impressive. RSIP Vision uses these technologies to analyze thedifferent videos successfully. It tells the interventional urologist or the surgeon specific information about the area explored by the probe and about the phase of the task being performed. Read more about AI solutions for Endoscopy Procedures Video Analysis for Medical 11
Made with FlippingBook
RkJQdWJsaXNoZXIy NTc3NzU=