الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 130 |  |  |
| | | from | 130 | | |
Abstract
Our study was carried out on 30 patients on maintenance hemodialysis, randomly selected from the dialysis unit of Mansoura Urology and Nephrology center. Initially,.the patients were assessed by thorough clinical and laboratory investigations. The patients then divided into 2 groups the first group represent the group of patients used for training the Nellfal networks model and included 13 males an" ~ " 2 females and tlleir age were 3R.4 :t 7.85. The second group used as a testing group to the neural networks model and included 14 males and one female and their age were 36.20:t 13.51. Initially all the patients were dialyzed by the same dialyzer type (Terumo 812) all the patients were assessed by using. single pool, double pool model, direct dialyser clearance measurement and direct dialysate quantification model then a comparison was done between our neural networks model, mathematical models and the direct dialysate quantification model (which is the gold standard model). The results showed that nellfal networks model predicted the actual time needed for an adequate session with an standard error 8.33 % at blood flow 200 ml/min & 0.97 % at blood flow 400 ml/min on the other hand the mathematical model predicted the actual time of the session with standard error 10 % for the time predicted at blood flow 200 ml/min & 31.47 % for the time predicted at blood flow 400 ml/min. Also, there was linsignificant difference between solute removal index predicted by neural networks model and direct dialysate quantification model which indicate tl1at the neural networks model is as accllfate as the dialysate quantification model. . Conclusion :from our wO1-k we found that using Neural networks model in the assessment of urea kinetics gives better results and we can predict the rime needed to achieve an adequate session or to achieve a target Kt/v value even from one session and only from one blood sample before haemodialysis session with the avoidance of multiple trials which was used in the mathematical models. The neural networks model is considered a step fotward for the artificial biologically intelligent processors which will be used in the dialysis machines in the future.