ARTIFICIAL NEURAL NETWORK BASIS FOR LONG-TERM VIABILITY PREDICTION OF VASCULAR ACCESS IN HEMODIALYSED POPULATION

The important number of end-stage renal disease population and subsequently, the modality of performing an optimal vascular access adequate for dialysis represent important healthcare problems that are on continuous debate. Additionally, considering that only half of the newly created arteriovenous fistula are well-developed and even a smaller number is free of related complications, there is an increased interest in achieving a suitable vascular access. Therefore, the aim of the present study was to create a computational model to monitor hemodialysed patients and to predict arteriovenous fistula evolution. The research included 52 newly initiated hemodialysed patients on central venous catheter or arteriovenous fistula, followed along I year. The statistically analyzed data and also the described artificial neural models related to this particular category of individuals emphasized a correlation between various bioumoral factors, such as hemoglobin, creatinine, urea, uric acid, serum albumin, erythrocytes sedimentation rate, serum iron, and the future growth rate of the vein diameter, implicated in creating a patency fistula. Although the findings were promising, more complex researches are needed in the future designed to take into account several other aspects related to the underlying pathology or to the dialysis procedure itself and to study a larger number of hemodialysis patients to be followed for a longer period long.