長波長干渉拡散相関分光法 (LW) によって可能になるポータブルな高速血流測定

Scientific Reports volume 13、記事番号: 8803 (2023) この記事を引用

825 アクセス

1 引用

メトリクスの詳細

拡散相関分光法 (DCS) は、組織内の血流を特徴付けるために使用できる光学技術です。脳血行動態の測定は、DCS の有望なユースケースとして浮上していますが、従来の DCS の実装では、成人の脳血流の堅牢な測定を行うには最適ではない信号対雑音比 (SNR) と脳の感度が示されています。この研究では、脳の感度と SNR の両方を向上させるために、より長い照明波長 (1064 nm)、マルチスペックル、および干渉検出の使用を組み合わせた長波長干渉 DCS (LW-iDCS) を紹介します。超伝導ナノワイヤ単一光子検出器に基づく長波長 DCS との直接比較により、人間の被験者で測定された血流信号において、LW-DCS の単一チャネルと比較して SNR が約 5 倍向上することを実証しました。我々は、LW-DCS と LW-iDCS の間で抽出された血流の同等性を示し、3.5 cm の線源と検出器の距離で 100 Hz で測定された LW-iDCS の実現可能性を実証します。この性能の向上により、脳血行動態の堅牢な測定が可能になり、拡散相関分光法の新たな使用例が可能になる可能性があります。

拡散相関分光法 (DCS) は、組織血流の非侵襲測定を可能にする確立された光学技術です1。 DCS は、拡散後方散乱光の測定を通じて、収集された信号の時間的変動を血管系を通る血球の動きに関連付けます。ベッドサイドでの臨床血流モニタリング 2、特に脳血流モニタリング 3 は、DCS のユースケースとして爆発的に増加しており、DCS は外科手術中の脳灌流の指標を推定するために使用されています 4,5,6,7,8、脳自己調節 9,10、脳血管反応性11、頭蓋内圧12、13、14、臨界閉鎖圧15、16。 DCS モニタリングを含む多くの研究が成人集団で実証されていますが、脳の感度と信号対雑音比の制限により 17、標準的な DCS 技術は、脳外組織（頭皮と頭蓋骨）は成人よりも大幅に薄い18、19。成人集団における DCS のパフォーマンスを向上させるために、多くのグループが脳の感度、信号対雑音比、またはその両方を改善する DCS の改良を開発しました。これらの方法には、干渉検出法20、21、22、23、24、25、並列スペックル検出法26、27、28、音響光学変調法29、30、31、経路長分解法32、33、34、35、36、37、スペックルコントラスト法38が含まれます。、39、40、そして長波長は41、42に近づきます。私たちのグループの最近の研究では、1064 nmで適用される長波長DCSの使用の有用性が示されていますが、実際の臨床測定では、現在入手可能な市販の検出器は、深層流（InGaAs/InP単一）に敏感な測定に対して妥当なノイズ性能を備えていません。 -光子なだれダイオード（SPAD））43、または臨床応用するには大きすぎる（超伝導ナノワイヤ単一光子検出器（SNSPD））。検出器技術におけるこのギャップに対処するために、当社は長波長干渉計 DCS (LW-iDCS) を開発しました。これは、1064 nm で動作する利点をすべて活用し、干渉計を使用して 1064 nm の光に敏感な検出器技術のマイナス面を回避します。高度に平行なラインスキャンカメラセンサーと組み合わせて検出します (Zhou らによるより短い波長で行われた研究に触発されました 21,44)。この研究では、新しい LW-iDCS 技術による血流推定の同等性を検証し、測定された信号の品質を比較するために、パイロット被験者研究で LW-DCS と LW-iDCS のパフォーマンスを直接比較します。

3.5 mm center-to-center distance), 1 single mode fiber for short-separation DCS (5 mm) and several co-localized long-separation detection fibers: 4 single mode fibers (LW-DCS), and 7 multimode detection fibers (LW-iDCS). A high coherence (lc > 10 km), fiber (MFD 6.6 µm) laser source emitting ~ 125 mW at 1064 nm (RFLM-125-0-1064, NP Photonics) was fusion spliced (S185HS Fusion Splicer, Fitel) to a 90:10, polarization maintaining fused fiber coupler (MFD 6.6 µm, PN1064R2A1, Thorlabs). The 10% arm of the coupler was used as the input for a fiber amplifier (MAKO-AMP1064, Cybel), and was connected via an FC/APC connector. The amplifier output fiber (MFD 10 µm) was fusion spliced to the input of a 50:50, 105 µm, multimode fused fiber coupler (TW1064R5A1B, Thorlabs). The two outputs of the fiber coupler were spliced to two 105 µm multimode source fibers connected to the probe. The light was amplified to allow for two MPE limited spots54 (1 W/cm2 at 1064 nm, 3.6 mm spot size diameter, 102 mW each spot) to increase the achievable signal-to-noise ratio. The 90% output arm of the polarization maintaining coupler was connected to the reference arm input of the LW-iDCS interferometer. All spliced connections were confirmed by the fusion splicer to have losses less than 0.03 dB./p> 50%. (B) For this maneuver, as expected, the systemic physiology was not significantly affected by the tightening of the tourniquet on the forehead./p> 3.5 mm apart could be used, allowing for an even higher SNR for high quality pulsatile blood flow measurements. The SNR of the LW-iDCS measurement seen in the high-speed pulsatile measurements was 4.5× the SNR of the SNSPD LW-DCS measurement when making single channel comparisons, representing an enabling improvement to the quality of blood flow measured. In the context of the DCS systems currently used for translational research, this improvement is especially significant considering that even the single illumination SNSPD LW-DCS has an SNR gain of 16× over conventional DCS42, and that measurements at 3.5 cm are not feasible with conventional NIR DCS. The use of a camera which is sensitive to light at 1064 nm takes advantage of both the higher number of photons per mode as compared to traditional NIR wavelengths as well as the slower decay of the autocorrelation function. For cerebral blood flow measurements made at long source-detector separations, the autocorrelation decay for traditional NIR DCS can happen in 1–10 s of microseconds, and a significant portion of the decay could be missed if not sampled quickly enough. The use of both heterodyne detection, measuring the slower decaying ${g}_{1}\left(\tau \right)$ as opposed to ${g}_{2}\left(\tau \right)$, and 1064 nm relaxes the sampling rate needed to effectively sample the correlation function. The longer source-detector separation achievable with these advanced DCS systems enables measurements with reduced sensitivity to the upper tissue layers relative to the sensitivity of currently applied DCS systems in the traditional NIR wavelength range (explored in the supplement). The decreased sensitivity to extracerebral signals is greatly beneficial to DCS measurements, especially in clinical applications where systemic physiological fluctuations are more likely to occur and the timing of relevant cerebral hemodynamic changes is not as well defined. We also see good agreement with the estimated noise performance given by Monte Carlo simulation (Figure S3). Additionally, the cost of the system is greatly reduced compared to LW-DCS based on SNSPDs. For this implementation of the LW-iDCS system, the detector used is ~ 7× less expensive (~ $25 k total, camera + frame grabber: ~ $20 k, assorted lenses, opto-mechanics, and fibers: ~ $5 k) as compared to the SNSPDs (~ $180 k total, cryostat: ~ $100 k, individual nanowire detectors: ~ $20 k each). The LW-iDCS cart-based system is also more mobile than the SNSPD based LW-DCS system. These improvements in cost, SNR, and mobility are promising for the clinical usability of LW-iDCS measurements of CBF in adults. The signal processing approach used to extract the correlation function from the raw data stream points to potential pitfalls in the development of iDCS instruments using multimode fiber and free space interferometers though. The motion of fibers and vibrations in the environment have the potential to corrupt the iDCS signals, however, these challenges are manageable, and the use of the custom data analysis pipeline, described in supplementary information, was successful in removing artifacts from the data. The use of a weighted fitting approach allowed for equivalent blood flow indices to be fit from both the LW-DCS and LW-iDCS correlation functions, evidenced by the results shown in Fig. 3C and D. While the results presented matched well, investigation of the generalizability of the weighting factor selected in this study is warranted given the influence that tissue layer thicknesses, optical properties, and ratios of scalp and brain blood flow are known to have on fitting autocorrelation functions67,68. Another challenge posed by the implementation of massively parallel multi-speckle detection is the raw data rate of the instruments. Recent publications on massively parallelized detection have quoted raw data rates between 0.24 GB/s (0.864 TB/hr) and 9.0 GB/s (32.4 TB/hr)22,25,26,27,28,44,69. For clinical blood flow measurements, these data rates could result in untenably large data files, though real time processing utilizing GPUs or FPGAs have been explored as a solution to address this challenge28,69. The increased SNR provided by the LW-iDCS instrument presented here enabled high sensitivity to the cerebral blood flow signal as well as a high rate of BFi calculation. These factors will be highly enabling for the clinical translation of DCS as a noninvasive cerebral blood flow monitor./p>