智慧型模糊推論血液透析品質分類輔助系統

本研究開發「智慧型模糊推論血液透析品質分類輔助系統」，由長期收集的血液透析檢驗數據去進行分析，分析結果顯示提出的方法對血液透析品質有分辨的能力。本文首先依文獻去設計專家問卷，並訪談相關的專家去萃取血液透析品質指標：Kt/V、URR、Albumin、Hct，再利用修正的MEPA將四個指標數值離散化建立其歸屬函數。最後運用模糊理論，建立法則庫，提供醫護人員臨床照護參考，對病患狀況作預先處置，達到適化透析的境界。本研究以區域醫院血液透析中心128位長期血液透析病患，取得病患臨床檢驗數據作．驗證，使用開發之輔助系統後所得結果：當品質指標值分為兩個語意值及α=0.7 β=0.3時正確率為98.30%；品質指標值分成三個語意時α=0.8 β=0.1時正確率為95.47%。研究中提出以區別分析與開發的系統作比較，亦顯示本系統之適用性及正確率均優於統計方法。歸納本研究最主要之貢獻：建立血液透析品質指標隸屬函數、將過去三個月醫師臨床診斷資料與四個品質指標值，事先歸類與轉換成隸屬值，作模糊相似性運算與門檻值取捨產生推論法則，並重複進行整個資料的學習與測試，使其辨識率達到滿意效果。及開發智慧型模糊推論血液透析品質分類輔助系統，將病患檢驗資料作品質分類，提供醫師臨床診斷治療的預警，解決病患隱藏的危機。

關鍵字

血液透析；品質指標；模糊推論

並列摘要

In this paper, we develop ”Intelligent Hemodialysis Quality Classification Support System based on Fuzzy Inference method”. From long-term accumulated experiment values for dialysis, the analytic results show that the proposed method can discriminate effectively the dialysis quality. We design an expert questionnaire by studying the medical research, and interview the related experts to extract the dialysis-quality indexes, which is: Kt/V, URR, Albumin, and Hct. Then the membership functions of dialysis-quality indexes are established. Further, we use fuzzy theory to set up the rule-base to provide the useful information for the medical workers serving in clinical care. It will be helpful for the patient if we can predict the patient's condition and take the proper treatment; we may prevent his illness getting worse and reach the state of the appropriate dialysis treatment. In this paper, we collect 128 patients who have been accepted the long-term dialysis at a dialysis-center of an area-scale hospital as the study objects, and then verify their history clinical data diagnosed by the doctors. This paper has three contributions as follows: 1. Establish the membership function of dialysis quality indexes. 2. Previously, taking the three-months-ago clinical data diagnosed by the doctors and four quality indexes to classify and transform into membership function, we produce the inference rule by fuzzy-similarity operation and threshold selecting. Repeat the whole cycle of learning and testing the data and make the satisfying right identification rate. 3. Developing the support system of evaluating the dialysis quality, the system can classify the patients' laboratory data to different kind of treatment quality, and provide the alarm of the clinical treatment for the doctors, and solve the potential risk of the patients.