Computer Vision News - March 2020

3 Summary Real-Time SLAM in Endoscopy 9 for example, often prefers a red image that shows the veins more distinctively. Currently, most SLAM algorithms in research target outdoor scenes, which relyonhighvariationin-depthandrequire rigid objects. However, images typically captured during endoscopic surgeries lack both of these characteristics. To overcome these difficulties, custom- made SLAM algorithms use state of the art deep learning models that detect additional information of the image. “In conjunction with prior knowledge of the real-life topography of the human body, the algorithm can optimize and create a precise 3D model of the endoscopic environment” , shares Barel. Real-Time SLAM Has Real Benefits Using these algorithms has multiple benefits in medical imaging. A 3D model canmeasure the dimension of any object from any angle in real-time, simply using a video stream. This allows for a patient- friendly, non-invasive, and painless investigation of the patient’s body. Barel concludes: “Thanks to these types of algorithms, we can introduce location tracking that only uses internal sensors, or we can use error detection and correction in systems that use external localization. As a result, the devices provide more detailed information about the patient’s body, allowing for increased accuracy during surgery. ” Read more about AI applications for endoscopy . "The algorithm can optimize and create a precise 3D model of the endoscopic environment."