次世代智慧型加護病房照護系統-子計畫八：基於次世代智慧型ICU之生醫訊號處理器與多重生醫資訊顯示整合設計( I )

Abstract

在次世代智慧型加護病房(ICU)之電子資訊工程設計上重點包含多重生醫資訊整合顯示、空間域生醫訊號顯示、可程式化生醫訊號處理器、低功耗、即時處理。故本子計畫旨在設計開發1) 可調式高能源效率獨立成分分析演算法及處理器，2) 可程式化生醫用高能源效率繪圖處理器，3)低複雜度腦映射演算法設計，4)終端多重生醫訊號顯示介面。在本子計畫中，我們會將上述新型設計整合於嵌入式平台上展示，以達成智慧型加護病房照護系統。中控端包含可調式化獨立成分分析處理器設計提供後級/使用者針對任意通道之生醫訊號均能分析處理，經由獨立成分分析處理器計算完成之EEG訊號權重矩陣及DOT訊號，將透過生醫用低功率三維繪圖處理器計算三維腦殼映射結果並完成空間域之三維生醫訊號顯示。最後，將所有二維與一維生醫資訊透過整合顯示介面與三維場景共同顯示在嵌入式系統上。第一年之執行摘要如下所述：本子計畫第一年計畫預計開發可調式高能源效率獨立成分分析處理器架構及可程式化生醫繪圖處理單元架構之設計。透過演算法的設計將會知道該獨立成分分析所需要運算種類集合，並依此分析設計該獨立成分分析處理器之指令集。同時生醫用高能源效率三維繪圖處理器之生醫繪圖處理單元之指令與架構會在該年度完成設計。

In the next-generation intelligent ICU care system, the electronic-information design features include multiple biomedical information display, spatial-domain biomedical signal display, programmable biomedical signal processor, low power, and real time processing. This sub-project will develop the following the advanced designs: 1) flexible energy-efficient independent component analysis (ICA) algorithm and processor, 2) programmable biomedical low-power graphics processing processor, 3) low-complexity brain mapping algorithm, 4) terminal-side multiple biomedical information display interface. In this sub-project, we will integrate and demo the above designs in the embedded system to show the intelligent ICU care system. In the console side, the energy-efficient ICA processor can provide the flexibility to the users and accelerate the EEG signal processing for the arbitrary channel. Next, the biomedical low-power 3D graphics processing processor can show EEG weighting matrix and DOT in 3D brain scene via the low-complexity brain mapping. At last, all 2D and 1D biomedical information together with 3D information will be displayed in the embedded systems. The brief summary of the first year is as follows:In the first year of the sub-project, we expect to explore the flexible ICA processor architecture and biomedical shader unit architecture. The instruction set of the flexible ICA processor will be determined according to the ICA operations and algorithm. The instruction set and architecture of the biomedical shader unit will be done for the biomedical low-power 3D graphic processor in this year.