Computer Vision News - December 2021

37 Procedural Planning in Urology Segmentation and 3D reconstruction from CT/MRI Recent developments in computer vision and AI allow accurate segmentation of desired organs and lesions within. Neural networks can be trained to segment the organs within the urological tract, and even sub-organs within them. Also, lesions or tumors can be segmented, resulting in a 3D model of the patient, with the interesting regions segmented and highlighted. This model is a map of the patient. Biopsy Planning The biopsy requires prior knowledge of where the organ is located (e.g. prostate), where the suspected lesion is within the organ, and what is the least harmful direction to get to the lesion (without damaging nearby structures). In real-time the physician uses experience and ultrasound guidance to perform this task accurately, but all this information exists in the CT/MRI data! Based on the scan and the segmentations, the best path from the skin to the lesion can be determined, automatically or by the physician. The system can then extract the exact point of penetration on the skin, the ideal angle of the needle, and the distance from skin to the lesion which the needle needs to penetrate. The physician can use these directions as a baseline for the biopsy and be certain that the needle is pointing in the right direction. Additional modules can use Ultrasound or X-ray for needle tracking. The probe position can be tracked using stereo cameras or י sensors on the probe, and this position can be used to verify needle position. Needle distance from skin can be calculated using the Ultrasound or X-ray calibration and needle tracking, so that the physician knows exactly how deep the needle is penetrating. All these tools can improve the procedural accuracy and reduce time and error- rate. The benefits of this module, aside from the obvious clinical advantage, is the relatively simple implementation. It requires minimal hardware additions for needle tracking, and it utilizes the available means – a PC - to improve the procedure. By simply adding AI you can get a superior procedure. More projects in AI for urology here. All this information exists in the CT/MRI data!