Urolithiasis, or kidney stones, is a common pathology affecting nearly 10% of the population in the USA. There are various treatment options for urolithiasis. The preferable method is natural-passage – simply urinate the stones. If stone size renders natural-passage impossible, other options exist. Stones can be removed using ureteroscopy (URS) or extracorporeal shock wave lithotripsy (ESWL), or in a percutaneous manner – PCNL (percutaneous nephrolithotomy). In URS, an endoscope is inserted through the urethra and ureters until the stone is reached. Then small tools are used to grab the stone and remove it. ESWL requires real-time stone tracking while shockwaves shatter the stone, until they can be passed naturally or removed safely. In PCNL a needle is inserted percutaneously towards the stone and is then removed. All methods require careful visual inspection of the stone via the relevant imaging modality, accurate navigation towards the stone, and careful removal of the stone and/or its fragments.
Stone tracking throughout the procedure is essential for verification of complete stone removal to reduce recurrences and additional procedures. There are several artificial intelligence (AI) algorithms for tracking objects in images, which can be used for this task.
URS utilizes direct visualization of the stone using an endoscope, and tools for extracting the stones. However, stone tracking can be impacted by blood and debris during the procedure. Using deep learning, networks can be trained to detect the stone and debris within the field-of-view, even when the object is obstructed. The urologist can focus on removing the stones while the algorithm continuously oversees the procedure.
During ESWL, the stones are targeted using ultrasound or fluoroscopy, and are broken using shockwaves. Due to the nature of this procedure, it is difficult to visually track the stones and their fragments, especially on US or fluoroscopy images. By implementing sophisticated tracking and registration algorithms, the position of the stone can be presented in real time on the pre-op CT scan, and continuously tracked in the different modalities. More advanced algorithms can even predict stone fragmentation patterns and success rates.
PCNL is more invasive than the other alternatives as it requires percutaneous penetration into the kidney, and usually is carried on when stone size is too large for the other methods. Similar to ESWL, thanks to AI the stones can be tracked on the fluoroscopic or US images during the procedure and registered to other imaging modalities, ensuring no stones are missed or lost during the procedure.
Urolithiasis care can benefit tremendously from the introduction of artificial intelligence technology to the procedures. The advantages of this technology include increased procedural accuracy and fewer complications. Implementation of this technology can be complex, but RSIP Vision has a proven track record of development of such solutions and assimilation into clinical settings. Contact us today for faster development and better clinical results!