Design and quality assessment of a predictive modeling framework to support vascular access surgery planning

T.M.G. Maas

    Research output: ThesisPd Eng Thesis


    Patients with chronic kidney failure are dependent on renal replacement therapy. One of the treatment options is hemodialysis, which requires a vascular access. The preferred vascular access is an arteriovenous fistula (AVF), where a vein is surgically connected to an artery. After the surgery, the blood flow will increase and the vessel diameters will dilate. When the flow is larger than 500 ml/min and the venous diameter is larger than 4 mm the AVF can be used for hemodialysis, this is called a matured AVF. A lot of AVFs suffer from non-maturation (up to 50%). To reduce the non-maturation rate, a computer model was developed during a previous international project which was able to simulate the postoperative blood flow through the AVF. The current one year design project is part of a multicenter randomized controlled trial (RCT) to investigate the potential benefit of including model based flow predictions in the regular decision-making process. The RCT will investigate if the non-maturation rate can be reduced by adding an advice, based on the flow predictions from the model, to the clinical decision-making. The design project consisted of the following aims: 1) Design a framework which allows to perform patient specific computer simulations in a multicenter RCT to predict the postoperative blood flow in an AVF. 2) Assess the quality of the input data of the computer model. The designed framework used in the RCT showed to be able to simulate the direct and six week postoperative flows, the framework operates in a largely automated way and the framework allows to report the simulations results back to the participating hospitals within a time frame necessary in the RCT. The framework meets all the requirements and therefore, it is currently used in the RCT. The quality of the input data is investigated by assessing the accuracy. A calibration study was designed and executed to investigate the trueness error. A Gage R&R study was performed to investigate the precision error, which is caused by reproducibility and repeatability variation. The calibration study was performed by a newly created in vitro calibration setup, which was easy to transport, therefore calibration measurements could be performed in the different participating hospitals. The bias in the diameter measurements (maximum 6.6%) was acceptable for use within the computer model. The bias in the flow measurements variated from 5% to 40%, which is larger than the variation used in the model. The effect of the flow bias on the flow predictions from the model showed to be small, consequently the trueness of the measurements was assumed to be accurate enough for use within the RCT. The precision study was carried out on patients with kidney failure in one of the participating hospitals of the RCT. The results showed that the precision error for most measurements used as input in the computer model, occur within the variation defined in the model. However, the precision error of the venous diameter measurements occurred outside the specified uncertainty in the model. Therefore, the effect on the flow predictions from the computer model should be assessed. In conclusion, the RCT framework was successfully implemented and the RCT should continue with the current framework, since the efficacy of the computer model needs to be assessed at the end of the study. For future applications of the model, the precision error of the input data should be assessed in (at least) one more hospital.
    Original languageEnglish
    • Huberts, Wouter, Supervisor
    • Tordoir, Jan H.M., External supervisor, External person
    • Lammerts, Ivonne M.M., Supervisor
    • Delhaas, Tammo, Supervisor
    Award date15 Sep 2016
    Place of PublicationEindhoven
    Publication statusPublished - 15 Sep 2016

    Bibliographical note

    PdEng thesis. - Executed at
    Maastricht University Medical Center


    Dive into the research topics of 'Design and quality assessment of a predictive modeling framework to support vascular access surgery planning'. Together they form a unique fingerprint.

    Cite this