以寬頻雙短通道功能性近紅外光譜系統分析訊號品質與組織血液動力學之關聯性

Abstract

功能性近紅外光光譜(Functional Near Infrared Spectroscopy, fNIRS)技術透過非侵入式光學方式在頭皮表面接收組織漫散射的漫反射光光譜，並透過修正比爾-朗伯定律(MBLL)定量大腦活動時的血液動力學反應。使用寬頻的光源除了可以定量含氧血紅素(Oxygenated Hemoglobin, HbO)、缺氧血紅素(Deoxygenated Hemoglobin, HHb)的濃度變化量，也可以定量與大腦能量代謝直接相關的氧化態細胞色素c氧化酶(oxCCO)的濃度變化。 然而，fNIRS依靠漫反射光的原理使其易受淺層干擾，且淺層干擾也不容易透過傳統的訊號處理流程去除。因此研究者使用短通道獨立量測行經淺層組織的漫反射光譜，並運用不同的方法處理短通道資訊。然而，淺層組織的血液動力學反應可能具有空間異質性，因此傳統上僅使用靠近長通道光源的短通道方法可能有其侷限性。 本研究使用寬頻的鹵素光源，並在長通道的光源與偵測端各設置一個短通道，對五位受試者進行延遲匹配樣本任務(Delayed Matching to Sample, DMS)、Stroop 色彩－詞彙測驗(Stroop Color and Word Test, Stroop)、n-back 任務等認知測驗，以定量出大腦在活化時三種吸收物質的濃度變化，並計算HbO, HHb及oxCCO濃度變化的訊號品質指標。在兩種及三種吸收物質的假設下，驗證殘差光譜(Residual Difference)形狀是否與 oxCCO 的莫爾消光係數(Molar Extinction Coefficient)光譜相似並具跨通道的一致性，藉此驗證三種吸收物質的假設能有效提取oxCCO的濃度變化。除了使用不分層的MBLL模型定量出的血液動力學變化之外，也使用雙層MBLL 的模型中搭配長通道光源附近的短通道偵測器或是長通道偵測器旁的短通道光源，驗證使用短通道能提升擬合成效及訊號品質的假設。並比較使用兩種不同的短通道以及同時使用雙短通道進行MBLL在擬合成效及訊號品質的變化是否與淺層組織血液動力學的異質性相關。並透過長短通道血液動力學之間的相關性試圖找出最佳的短通道選擇。 實驗結果顯示殘差光譜與oxCCO吸收光譜形狀相似且殘差光譜形狀具有跨通道的一致性。使用短通道搭配雙層組織模型相較於不使用短通道的單層組織模型的訊號對比度較高。在雙短通道的淺層組織血液動力學異質性較高時，選用其中一種短通道的訊號品質較佳。而雙短通道的淺層組織血液動力學異質性較低時，三種短通道方法的訊號品質差異不大。最後，基準光譜期間長短通道的血液動力學相關性對於訊號品質並無明顯的影響。

Functional near-infrared spectroscopy (fNIRS) is a non-invasive optical technique that measures diffusely reflected spectra from biological tissues at the scalp surface to quantify cerebral hemodynamic responses associated with brain activity using the Modified Beer–Lambert Law (MBLL). The use of broadband light sources enables not only the quantification of concentration changes in oxygenated hemoglobin (HbO) and deoxygenated hemoglobin (HHb), but also the assessment of concentration changes in oxidized cytochrome-c-oxidase (oxCCO), which is directly related to cerebral oxygen metabolism. However, due to its reliance on diffuse reflectance, fNIRS measurements are highly susceptible to contamination from superficial tissues, and such interference is difficult to remove using conventional signal processing techniques. To address this issue, shortseparation channels are commonly used to independently measure diffuse reflectance originating from superficial tissues, and various strategies have been proposed to incorporate short-channel information. However, superficial hemodynamic responses may exhibit spatial heterogeneity, suggesting that conventional approaches relying on a single short channel located near the source may have inherent limitations. In this study, a broadband light source was employed, and two short-separation channels were implemented, with one placed near the source and the other near the detector of the long channel. Five healthy participants performed a series of cognitive tasks, including the Delayed Matching-to-Sample, Stroop Color and Word Test, and nback tasks. Concentration changes of HbO, HHb, and oxCCO during brain activation were quantified, and signal quality metrics for each chromophore were evaluated. Under both two- and three-chromophore assumptions, the similarity between the residual difference spectra and the molar extinction coefficient spectrum of oxCCO, as well as their consistency across channels, were examined to validate the effectiveness of the three-chromophore model in extracting oxCCO concentration changes. In addition to hemodynamic quantification using a conventional single-layer MBLL model, a two-layer MBLL model was applied with short-separation channels. These configurations were used to evaluate the hypothesis that incorporating short-separation channels improves MBLL fitting performance and signal quality. Furthermore, the effects of using two distinct short-separation channels individually or jointly on fitting performance and signal quality were compared, and their relationships with the heterogeneity of superficial tissue hemodynamics were investigated. Finally, correlations between hemodynamic signals measured from long and short channels were analyzed to identify optimal short-channel selection strategies. The experimental results demonstrate that the residual difference spectra are similar in shape to the molar extinction coefficient spectrum of oxCCO, and that the shape of the residual difference spectra shows consistency across channels. Compared with a single-layer head model without short-channel correction, the use of short channels in combination with a two-layer head model yields higher contrast. When superficial tissue hemodynamic heterogeneity between the two short channels is high, selecting one of the short channels results in better signal quality. In contrast, when superficial tissue hemodynamic heterogeneity between the two short channels is low, the differences in signal quality among the three short-channel approaches are minimal. Finally, the hemodynamic correlation between long- and short-separation channels during the baseline period does not have a significant impact on signal quality.