Hallym University Kangnam Sacred Heart Hospital

Ureteral stone is a mineral mass in the ureter, which may or may not have originated in the kidney and traveled down into the ureter, causing stabbing pain. A stone begins when particles of minerals in stagnated urine crystallize and form a mass. There are three ways to remove the stones. The first one is to wait until the patient pass the ureteral stones. The second one is breaking the stones by using extracorporeal shockwave lithotripsy(ESWL). The last one is ureteroscopy, where a doctor inserts a thin, flexible scope into the patient’s bladder and ureter. Among the three types of treatment, ESWL is the most favored by the patients, since it can remove the stones without a scratch on body. However, ESWL itself can't treat all ureteral stones, and it's also hard to precisely predict results before the treatment, eventually causing the patients to receive additional surgery in addition to wasting time and money.

To overcome those limitations, the research team led by Professor Seong Ho Lee of Urology who is also the director of Hallym University Dongtan Sacred Heart Hospital and Prof. Jin Kim of Computer Science at Hallym University developed a forecasting model that uses AI to enable the doctors to predict either success or failure in ESWL procedures beforehand. The research paper titled "A Prediction Model Using Machine Learning Algorithm for Assessing Stone-Free Status after Single Session Shock Wave Lithotripsy to Treat Ureteral Stones" was published in the recent issue of 'The Journal of Urology', an international journal of the United States that is ranked the second in the list of highest impact factor journals with the score of 5.381.

The research team analyzed 791 patients with ureteral stones who received ESWL treatment at Dongtan Hospital from October 2012 to August 2016. The result showed that 509 patients(64.3%) had the ureteral stones removed successfully, whereas 282 patients(35.7%) had failed to do so.

These data were then analyzed through AI using decision making trees algorithms. It means that once the program puts the new data through machine learning, it tries to anticipate the results as accurate as possible. As a result, a prediction model for assessing stone-free status after single session shock wave lithotripsy to treat ureteral stones was developed, which analyzes 15 factors including a patient's name, gender, and status. When applied to patients with ureteral stones, the model anticipated the results with an accuracy rate of 92.29 percent.

Director Seong Ho Lee said, "Thanks to the advances in AI, we are able to anticipate the results of treatment for ureteral stones. Patients no longer have to waste their money or endure a long period of suffering."

The prediction model can be found on the following website: http://pisces.hallym.ac.kr/ESWL/. This research outcome has been published in the latest issue of 'Urology', an academic journal of the American Urological Association.

By Chul Kwon, Int’l Cooperation Team, HUMC (chris@hallym.or.kr)