腎臟位於人體內的腹腔後壁的脊椎兩側,是身體新陳代謝、排除廢物、毒素、調節血壓、和維持體內液體平衡的重要器官。一旦腎臟功能異常,無法正常運作,身體便會產生毒素,危害器官,甚至導致死亡。為了延長或挽救腎臟病患者的生命,常使用腎臟替代療法:腎臟移植 (Kidney Transplantation)、血液透析(Hemodialysis, HD)及腹膜透析(Peritoneal Dialysis, PD)進行醫治。血液透析就是常稱的「洗腎」,利用人工腎(透析)設備將尿毒素、水份排出體外,減輕尿毒症狀。至於何時該進行血液透析,一般而言,當尿素氮(Blood Urea Nitrogen, BUN)大於90 mg/dL,血中肌酐酸(Blood Creatinine, CRE(Blood))大於9 mg/dL時,肌酐酸清除率(Creatinine Clearance, CC)小於0.17 ml/s或尿酸中的肌酐酸(Uric Acid Creatinine, CRE(UA))大於707.2 mg/dL,醫生就會建議可以開始進行血液透析治療。但是,腎臟組織在一般正常人約僅發揮30%的功能,就能維持日常的排毒功能,因此腎臟功能要在損壞高達70%時,血中尿毒指數才會明顯升高。當血中尿毒指數開始升高時,腎臟已是相當脆弱,稍微不慎就會迅速惡化。換句話說,當上述的指數到達可以洗腎的門檻時,腎臟其實已經損壞了1/3 以上。是否存在其他關鍵檢驗項目,例如腎臟功能的蛋白比值(A/G ratio);血液檢查的紅血球數量(Red Blood Cells, RBC);尿液檢查的尿液白血球數值(White Blood Cells, WBC)等,與腎臟功能有著潛在的關係,這些指標判斷腎臟功能是否異常的能力會較BUN、CRE或CC更好呢?因此,本論文提出一個血液透析關鍵因子分析技術,採用熵函數找出關鍵檢驗項目,並利用關鍵項目進行透析病患分群,用以判斷關鍵因子之準確性及分群效果。此外,進一步利用關聯式規則探勘演算法,找出各群內的規則,幫助病患提升察覺疾病發生的可能性。
The human kidneys are situated on the posterior abdominal wall on both sides of the spinal column. The main functions of the kidneys vital organs, including metabolism, excretion waste, toxins, regulation of blood pressure and maintain fluid balance. The renal failure or abnormal functioning is caused by the body produces numerous toxins, organ damage or even mortality. In order to extend or save lives of kidney patients, doctors usually use kidney replacement therapy, including Kidney Transplantation, Hemodialysis(HD) and Peritoneal Dialysis(PD). Hemodialysis is also known as dialysis, which uses of artificial kidney to reduce uremic of discharge excess uremia and water. The doctor will recommend the patients to perform the dialysis when Blood Urea Nitrogen(BUN) is greater than 90 mg/dL, Blood Creatinine(CRE-Blood) is greater than 9 mg/dL and Creatinine Clearance(CC) is less than 0.17 ml/s, or Uric Acid Creatinine(CRE-UA) is greater than 707.2 mg/dL. However, kidney is fragile when those test items is getting high. In other words, the kidney damages greater than 1/3 when the above-mentioned indexes are up to dialysis threshold. Thus, this thesis wants to find some other key items, such as Protein ratio(A/G ratio) of kidney function, Red Blood Cells(RBC) in blood examination or White Blood Cells(WBC) in urinalysis and so on, that can be used to predict the probability of hemodialysis. We use entropy function to find out the key features which have high relationship with hemodialysis, and apply k-means clustering algorithm with these key features to group the patients. Furthermore, the proposed scheme applies data mining technique to find the association rules from each cluster. The rules can be used to warn patients who might need hemodialysis.