基於三種基底函數之自適應濾波器應用於雙軸測試平台上之高強度聚焦超聲波探頭運動控制

Abstract

高強度聚焦超聲波（High-intensity focused ultrasound, HIFU）為一具發展潛力之醫療技術，具備非侵入性、無電離輻射及可縮短住院時間等優點，因而在臨床治療中日益受到重視與採用。然而，治療成效高度仰賴超音波探頭之定位精度與安全性，而該定位精度常受到呼吸所導致之器官移動的影響。為解決此問題，目前臨床上常採用「「呼吸中止法」以降低呼吸干擾，惟此法可能對患者造成額外負擔與副作用。因此，允許病患在自然呼吸狀態下進行 HIFU 治療將更為理想。為達成此目標，本研究將此問題視為一實時位置追蹤控制問題，旨在於病患自然呼吸期間，精確追蹤因呼吸所造成的腫瘤位移，進而實現超音波探頭之精確定位。為此，本論文建構一模擬肝腫瘤消融手術之實驗平台，平台包含線性滑軌、馬達、編碼器、馬達驅動器與微控制器等硬體元件。本研究採用適應性逆控制（Adaptive inverse control, AIC）架構進行追蹤控制，以因應呼吸引起之腫瘤運動。其中，為降低遞迴最小平方（Recursive least squares, RLS）演算法之時間複雜度，本文以拉格爾濾波器（Laguerre filter）取代傳統有限脈衝響應（FIR）濾波器。此外，為補償單一拉格爾濾波器於均方誤差（MSE）上的劣勢，本研究亦探討結合 Laguerre 與 FIR 濾波器之混合型自適應控制策略。實驗結果顯示，AIC 架構相較於基準 PID 控制器具有更佳之追蹤效能，能有效降低腫瘤與超音波探頭間之位置誤差。本論文最後將對不同濾波器與控制架構之性能差異進行比較與討論。

High-Intensity Focused Ultrasound (HIFU) is a promising medical modality possessing advantages, including non-invasiveness, non-ionization, and reduced hospitalization, that make it receive increasing acceptance in treatments. Since treatment safety and accuracy of the position of the ultrasound probe are strongly affected by respiration-induced organ motions, a “breath-hold” strategy is applied during the treatment to overcome the inefficacies. However, this strategy may lead to side effects, and thus, allowing breathing during HIFU treatment is preferable. To allow breathing and provide optimal positioning of the ultrasound probe, we consider this issue as a real-time position-tracking control problem. The goal is to achieve accurate positioning by predicting and precisely tracking the displacement of the target tissue induced by respiration during HIFU treatment. In this thesis, we construct a platform consisting of linear guides, motors, encoders, motor driver, and microcontroller to simulate the surgery process of liver tumor ablation conducted by HIFU. Adaptive inverse control (AIC) is applied to track the tumor motion caused by respiration. A Laguerre filter replaces the FIR filter to reduce the time complexity of the recursive least square (RLS) algorithm. Moreover, to compensate for the MSE of the Laguerre filter, we explore the combined Laguerre-FIR structured AIC. The results revealed that AIC effectively improved the tracking performance compared to the baseline PID controller, and therefore, the position error between the target tumor and the ultrasound probe was reduced. A comparison of the performance of different filters and structures is also discussed.