STORZ MEDICAL – Literature Databases
STORZ MEDICAL – Literature Databases
Literature Databases
Literature Databases

[Article only available in Russian] Ershov AV. et al., 2021: [The use of neural algorithms when choosing a method of surgical treatment of urolithiasis].

Ershov AV, Neymark AI, Kapsargin FP, Berezhnoy AG, Vinnik YY.
FGBOU VO Krasnoyarsk State Medical University named after professor V.F. Voyno-Yasenetsky of the Ministry of Health of Russia, Krasnoyarsk, Russia.
Altai State Medical University, Barnaul, Russia.


Aim: to assess the possibility of using neural network algorithms in choosing a method for surgical treatment of urolithiasis.

Materials and methods: treatment results of 625 patients with kidney stones were analyzed in the study. Information about each patient was presented in the form of a multidimensional vector characterized by following preoperative investigations: questionnaires, clinical examination, instrumental and laboratory studies. A register was created where information on more than 50 parameters for each patient was added. Each example has an output parameter representing a predefined treatment strategy (extracorporeal shock-wave lithotripsy [ESWL] - 1, percutaneous nephrolithotomy [PCNL] - 2, pyelolithotomy or nephrolithotomy - 3). The initial database served as the basis for training the neural network estimation technique.

Results: A prospective trial was conducted to assess the clinical efficiency of the recommendations of neural network. A cohort of 150 patients admitted to the urology department was divided into two groups of 75 people. In the group 1, patients received treatment according to the standard recommendations. In group 2, treatment strategy was chosen based on the results of neural network analysis. In the group 1, ESWL was performed in 40 (53.3%) patients. The average number of sessions was 1.8. At the discharge, residual fragments were diagnosed in 12 (30%) cases. In 4 patients, acute pyelonephritis developed, which required performing ureteral catheterization and subsequent treatment. In group 1, the efficiency of ESWL was 75%. In the group 2, where the neural network assessment technique was used, the average number of sessions was 1.4. At the discharge, 7 (15.6%) patients had residual fragments: 4 in the kidney, in 3 in the lower ureter ("steinstrasse"). In 4 cases, a conversion for PCNL was performed. ESWL efficiency was 91.1%. Stone-free rate for ESWL in the second group was significantly higher due to the greater number of stone fragmentation. In addition, number of shock waves was lower (the average number of sessions was 0.4 less). Improvement of treatment tactics through the use of neural network algorithms led to a decrease in hospitalization times, as well as to an improvement in the quality of treatment. The low efficiency of ESWL, as the first-line method, led to a change in treatment tactics in 25% of patients in group 1 and only in 8.9% of patients in group 2. Using these algorithms, it was possible to reduce hospitalization time, need for changing treatment strategy, number of auxiliary procedures, readmission rates, the incidence of inflammatory complications, and the number of residual fragments after ESWL.

Conclusions: The possibility of using the neural network prediction technique at the preoperative stage in patients with kidney stones has been shown. This technique allows practicing urologist to make a decision on the choice of the optimal treatment method on an individual basis, thereby minimizing the risk of early postoperative complications.
Urologiia. 2021 Sep;(4):47-52.PMID: 34486274 Russian.



No comments made yet. Be the first to submit a comment
Friday, 14 June 2024