Ensemble倒傳遞類神經網路應用於嚴重頭部創傷病患預後評估指標模型之建立

Ensemble Back-propagation Neural Network Applied to brain death index prediction for the patients with severe head injury

鄭凱元(Kai-Yuan Cheng)

指導教授 ： 謝建興

元智大學/工程學院/機械工程學系/碩士(2008年)

https://doi.org/10.6838/YZU.2008.00093

摘要

近年來由於器官捐贈的觀念已漸漸被國人所接受，對於腦傷病患的腦死判定也因此顯得格外重要。而目前的法定程序僅針對已簽署器官捐贈同意書的患者進行腦死的判定，但對於同樣屬於嚴重腦傷卻未簽署器官捐贈同意書的不可逆之呼吸停止昏迷（irreversible apnoeic coma, IAC）病患，依法卻無法進行標準的腦死判定程序來宣告病人的死亡。有鑑於此，本研究計畫提出一種較為簡易的方法來建立嚴重頭部創傷病患的預後評估指標（brain death index, BDI）模型，以提供醫護人員更多的資訊作為評估病患腦部創傷嚴重度的參考。
　　資深的醫師，可依據本身臨床診療的經驗判讀生理訊號的變化來診斷並預測病換的預後狀況。本研究希望建立一套智慧型系統，藉由模擬醫師的決策流程來達到預測的工作。此外心律變異度與腦傷嚴重度的關係亦是本研究欲探討的另一項主題。心律訊號是一種最為常見的非侵入式訊號，除了蘊含許多生理資訊之外，臨床上它亦是最常見被應用於量測自主神經功能的生理參數。有鑑於此，本研究設計了兩階段的實驗來達成目標，第一階段的實驗是針對護理人員的臨床護理紀錄來取得欲分析的10組重要生理參數，並找出其重要性的排序。第二階段的實驗則是個別針對心律訊號做分析，並利用類神經網路（artificial neural network, ANN）中的多層式感知器（multi-layer perception, MLP）架構作為系統開發的模型，配合經驗模態分解法（empirical mode decomposition, EMD）的技術來研發腦死病患的預後評估指標模型，以提供醫護人員於標準腦死判定程序外，另一個較為方便及快速的參考指標，以便後續積極的進行器官移植或是安寧緩和照顧的工作。

關鍵字

不可逆之呼吸停止昏迷 ； 心律變異度 ； 類神經網路 ； 經驗模態分解法 ； 嚴重頭部外傷病患預後評估指標 ； 多層式感知器

並列摘要

The concept of organ donation has already been accepted by people gradually in recent years so the judicial brain death determination process becomes very important. Clinically, patients with irreversible apnoeic coma (IAC) will be considered legally as brain death based on a judicial process, but this process can only be applied to people who had already signed the letter of consent to donate organs. Besides, the judicial process is also very lengthy. This study tried to find out an easier way to diagnose the prognosis of the patients with severe head injury, and offer the medical staffs other information to determine brain death.
　　Medical doctors depend on their experience to diagnose patients’ prognosis by observing the variation of different physiological signals. We tried to find out this relationship between the signals’ variation and doctors’ determination process. Heart rate variability (HRV) is the most common noninvasive physiological signal. It’s also a rich signal which contains a lot of information of human body, and probably the most investigated and readily assessable measure of autonomic nerves’ function. Therefore, we designed a two-stage experiment to achieve our purpose. For the first stage, the model is an expert system designed to mimic the determining process of medical doctors. We chose ten most important physiological signals to be the analyzing data in the first stage. But in the second stage, we only focused on analyzing the heart rate signal. The technique of artificial neural networks (ANN) and empirical mode decomposition (EMD) has been applied to construct the prediction model of brain death index (BDI). The multi-layer perception (MLP) and ensemble neural networks are chosen to be the network type of BDI model. This model can provide medical staffs a reference index to evaluate the status of IAC and brain death patients.

並列關鍵字

irreversible apnoeic coma (IAC) ； heart rate variability (HRV) ； artificial neural networks (ANN) ； empirical mode decomposition (EMD) ； brain death index (BDI) ； multi-layer perception (MLP)

參考文獻

[2].Wijdicks EF, “Brain death worldwide: accepted fact but no global consensus in diagnostic criteria,” Neurology, vol.58, pp:20-25, 2002.

[3].Reimers CD, “The definition and determination of brain death,” Bailliere’s Clinical Anaesthesiology, vol.13, no.2, pp; 211-225, 1999.

[4].Shivalkar B, Van LJ, Wieland W, “Variable effects of explosive or gradual increase of intracranial pressure on myocardial structure and function,” Circulation, vol.87, pp:230-239, 1993.

[6].Smith M., “Physiologic changes during brain stem death-Lessons for management of the organ donor,” J Heart Lung Transplant, vol.23, pp: 217-222, 2004.

[7].Wijdicks EF, “The diagnosis of brain death,” N Engl J Med, vol.344, no.16, pp:1215-1221, 2001.