應用於雙波段紅外線之乳癌檢測盲源分離演算法

Abstract

本論文提出一個針對雙波段紅外線頻譜(Dual-band IR Spectrogram)，應用改良的盲源分離演算法(Blind Source Separation Algorithm)進行分析，來判定追蹤乳癌的長期化療之成效。本文採用雙波段紅外線頻譜的原始資料(RAW Data)為輸入檔，進行演算法分析，並針對盲源分離演算法的特性，在過程中運用資料分析進行改良，最後將該演算法加載至本實驗室開發的醫療控管處理器設計進行整合，提供我們一個更便利的管道來體現雙波段紅外線頻譜分析應用於醫療的成果。在標定率方面，該改良演算法設計比其他演算法平均高15%，在癌細胞判斷的正確率方面，該演算法設計比其他演算法平均高10%。

This work presents an application of Blind Source Separation (BSS) Algorithms on Dual-band IR Spectrogram for breast cancer detection, which is used to trace the effect of long-term chemotherapy for breast-cancer patients. We take Dual-band IR Spectrogram’s RAW Data as an input to the BSS algorithms. Also, we plan to integrate this analytical algorithm into the back-end processor of our designed Dual-band IR Sensor and Readout Circuit Platform. This work will provide a more convenient medical application of our Improved Neighbor-based BSS algorithm on Dual-band IR Spectrogram for breast cancer detection. For Demarcating Degree, our Improved Neighbor-based BSS algorithm is approximately 15% better than other algorithms. For Correctness Rate, our improved algorithm approximately increases 10% compared with other algorithms.