Machine learning to prevent possible complications arising from arteriovenous fistula for haemodialysis

José Ibeas


    José Ibeas


    Fundació Institut d’Investigació i Innovació Parc Taulí (I3PT), Sabadell, Spain


    Patients with end-stage renal disease requiring renal replacement therapy through haemodialysis need an arteriovenous fistula, a surgically created connection that allows the patient's vascular system to be adapted to connect it to the haemodialysis machine. However, this connection has a high likelihood of failure and can also cause or exacerbate pre-existing cardiac pathologies due to the blood flows it can generate.

    This project aims to create a clinical decision support system based on machine learning which enables the risk of failure of this component and the potential cardiac implications for each patient to be individually determined. To achieve this, the plan is to extensively use data from various sources, including clinical, analytical, imaging and biometric data, all collected over decades of clinical experience, to assist professionals in their decision-making.

    This solution could benefit patients' health and quality of life, as well as the management of healthcare resource. The prototype, designed by the team using information from around a hundred patients, has produced results with a high degree of accuracy and precision. In this stage of the project, the team intend to increase the clinical evidence with data from a larger number of patients, refine the system by moving the tool towards a clinical trial and establish a spin-off to facilitate the necessary steps to start marketing the system so that it can be used by haemodialysis centres.


    Arteriovenous Fistula for Dialysis: Machine Learning based models for failure and Cardiac events


    Stage 2