Computer Vision News - October 2022

19 3D Reconstruction of the Ureter are acquired to assess procedural progress, and a recent uretermodel can be compared to the reference one for better progress tracking. Also, the availability of accurate 3D information during the procedure results in better surgical precision and higher confidence for the surgeon. This technology can be applied to other procedures conducted with fluoroscopy imaging. As detailed above, we already developed a 2D-to-3D reconstruction solution for the knee. Additionally, this can be applied to bladder reconstruction, coronary artery 3D modeling, ERCP, etc. Each use-case requires a custom-tailored solution, and the baseline is leveraged to fit the new need. RSIP Vision’s long track record allows the development of this tool by combining advanced segmentation algorithms with 3D reconstruction techniques in an efficient manner, providing a real-time solution for 3D ureter reconstruction. Ureteroplasty can be conducted faster, accurately, and ultimately with less complications. classic computer vision techniques. Once at least two fluoroscopy images are acquired from different projections, the ureter is automatically segmented, and the images are calibrated. The calibration can be adjusted using a pre-calibrated jig within the field-of-view, or by extracting specific c-arm geometry. Segmentation is performed using trained neural networks dedicated for this task. The second step consists of the actual reconstruction. The adjusted images and segmentations are passed to a deep neural networkthatproducesthe3Dreconstructed ureter. This model accurately portrays the ureter and can be used by the physician for procedure planning and general anatomical consideration. A 3D model of the ureter can be used as a navigational aid in other urological procedures too: stent placement, stone localization and retrieval, etc. Throughout the procedure, this model is used as a reference to verify positioning and anatomical location. As the procedure progresses, additional fluoroscopy images

RkJQdWJsaXNoZXIy NTc3NzU=