| Summary: | 碩士 === 國立清華大學 === 工業工程與工程管理學系碩士在職專班 === 107 === According to the statistics in the 2017 United States Renal Data System Annual Report, the incidence rate of end-stage renal disease (ESRD) in Taiwan in 2015 was 476 per million people while the prevalence rate of dialysis was 3,185 per million people. Both indices were the highest in the world, way surpassing the figures for the US and Japan.
Due to their damaged renal functions, end-stage renal disease patients are unable to metabolize the waste in their blood and have to undergo hemodialysis. Such patients need areriovenous fistulas to stay alive. Therefore, creation of dialysis fistulas is necessary before patients start to receive dialysis. Hemodialysis fistulas are surgically created communications between the native artery and vein in an extremity to allow blood in the artery to run to the vein and expand the lumen diameter of the vein. After surgery and ball grasping training, blood vessel
maturation takes about 1-2 months and a vein with expanded lumen diameter finally can be used as a fistula.
Narrowness of fistulas is a major problem in hemodialysis. The high frequency of use of fistulas (three times a week) causes the fistulas of some patients to narrow down. So far, narrowing down of fistulas can be discovered only after it has happened for some time and clinical symptoms have appeared or after specialists
conduct invasive inspections to check whether fistulas have narrowed down. If there is an easier method to predict narrowing down of fistulas, arrangements can be made in advance for patients to undergo percutaneous transluminal angioplasty (PTA) to open up the fistula and make sure there is no blockage before the following dialysis treatment in order to protect the safety of patients.
In this study, 31 key factors associated with time domain and frequency domain signals are adopted in conjunction with an artificial neural network to predict stenosis of fistulas. The time domain signal eiganvalues TR(x2 (t′), a = 0.2, c = 0.9)
and TR(x2t′ , c = 60) and frequency domain signal eiganvalue F6 are the main key factors in single-factor forecasting. In two-factor forecasting, combining time and frequency domain factors produce good results. The key two-factor in four test are (1)TR(x2 (t′), a = 0.2, c = 0.9)、F6 (2)TR(x2t′ , c = 70), F9 (3)TR(x2t′ , c = 80), F9 (4)WER, F7.
|
|---|
