离体心脏小梁收缩过程中的明场、荧光与光学相干断层成像同步观察 Ex Vivo

摘要

本方案介绍了一组关于肌节、钙离子及宏观几何结构的 来自主动收缩的心室小梁的数据 离体这三种成像模式的整合实现了上述同步测量。

摘要

在心肌细胞中，细胞内Ca²⁺ 瞬变激活收缩性肌丝，引起收缩、宏观缩短和几何形变 变形。由于无法直接观察肌肉内部，也无法精确追踪兴奋-收缩耦联动力学的时空特性，我们对这些事件之间内在关系的理解一直受到限制。为解决这些问题，我们构建了一种集成了多种成像模态的装置。具体而言，该装置整合了一台明场显微镜，用于测量肌节长度和组织应变的局部变化；一台荧光显微镜，用于可视化钙离子²⁺ 瞬态，以及光学相干断层扫描仪用于捕捉组织的几何结构 心脏周期内随时间变化的成像基础设施及相关数据采集框架。数据采集自称为肉柱的离体条状组织结构。在我们的仪器中，一对位置可控的铂金钩固定组织两端 离体 在持续用富含营养的生理盐水溶液灌流肌肉样本的同时，通过两个独立控制的钩子实现对肌肉长度和张力的实时调控。沿长度方向的平移运动可对样本进行分段扫描，从而克服显微镜成像窗口（540 µm × 540 µm）与典型小梁肌长度之间相对尺寸不匹配所带来的限制。>2000 µm）。肌肉腔两端的铂电极以用户设定的频率刺激心肌小梁。我们利用该刺激信号作为触发信号，用于同步各个成像窗口的数据，从而重建稳态条件下整个样本的收缩过程。对这些明场成像数据应用图像处理技术，可获得组织位移和肌节长度图谱。此类数据集合若整合至实验-建模流程中，将有助于深入理解生理与病理生理状态下肌肉收缩的均一性与非均一性。

引言

灌流离体心脏肌肉组织样本是研究心肌离子激活与力学特性的标准且广泛应用的实验方法¹。特别是从心室壁分离出的条索状结构——小梁肌的制备，使得研究人员能够评估包括收缩的长度依赖性激活²、收缩的牵张依赖性反应^3,4以及心肌组织舒张期黏弹性⁵在内的多种生理现象。该离体小梁肌灌流技术的开创者Ter Keurs最初结合使用荧光成像技术测量Ca²⁺浓度，以及激光衍射技术测定肌节长度^2,5。自这些早期研究以来，利用基于二维快速傅里叶变换（FFT）的方法对明场显微图像进行分析，以更高空间分辨率提取肌节长度信息，已成为越来越普遍的做法⁶。这两种成像系统的结合应用，可部分揭示Ca²⁺释放与肌节长度依赖性张力产生之间的内在关系。

心肌具有横纹，其可见的条带结构与由粗肌丝和细肌丝组成的收缩单位的有序排列相关。构成肌节的这些肌丝之间的相互作用是力量生成的基础，该过程启动如下： a 去极化电信号，即动作电位，引起电压依赖性L型Ca²⁺ 细胞膜上的通道打开； 随之而来的细胞内Ca²⁺流入²⁺ 诱导 Ca 的释放²⁺ 来自肌质网（SR），一种细胞内钙²⁺ 储存，这一过程称为 Ca²⁺诱导的 Ca²⁺ 释放⁷；细胞内 Ca²⁺ 浓度的这种突然升高²⁺ 浓度在纳摩尔至微摩尔范围内即可启动力的产生； Ca²⁺ 泵持续挤出 Ca²⁺ 从胞质溶胶返回肌浆网和细胞外区室； 当细胞内 Ca²⁺ 当浓度恢复至纳摩尔范围时，肌肉停止产生力量并随之舒张。在力量产生过程中，粗细肌丝相互滑动。肌节长度决定了两者重叠的程度，从而宏观上决定了肌肉产生力量的潜力。

本文将这些荧光-明场成像技术进一步扩展，以结合光学相干断层扫描（optical coherence tomography, OCT）。OCT 利用光的干涉物理原理，能够获取组织的几何形变，从而研究肌肉收缩的异质性⁸。我们的设备（图1）采用的是谱域 OCT（spectral-domain OCT, SD-OCT）系统。在 SD-OCT 中，一个分束器将来自宽带短相干长度超辐射发光二极管的光分为参考臂和测量臂。参考臂包含一个固定镜面，测量臂则包含一个二维检流计用于引导光束。从样品返回的散射光被收集后，与参考臂中反射的光发生干涉，形成干涉图样。深度信息被编码在光谱条纹的频率中。为了提取这些信息，信号通过光谱仪，然后对结果进行逆快速傅里叶变换（inverse FFT）。相应的 1D 信号代表不同深度的结构，对应于折射率的变化⁹（A-scan）。通过沿单一轴线偏转激光，可以构建出感兴趣样品的横截面图像（B-scan）；类似地，通过在另一轴向上以步进方式重复该过程，可生成三维图像（C-scan）。进一步地，基于外部触发信号，可在同一切片位置连续采集一系列 B-scan，用于呈现具有时间变化特征的重复性样本，并重复该过程以生成三维扫描图像，从而表示一幅随时间变化的平面图像¹⁰。

在整合三种成像系统时，我们考虑了以下两个原则。第一，成像传感器不应检测到来自其他成像模式的光信号；第二，物理设计应至少为三个同时进行的成像平面预留自由空间。为满足第一个要求，明场显微镜采用波长为660 nm的LED光源，在倒置构型下对样品进行照明。荧光显微镜采用落射荧光构型，使用同一物镜进行激发光照射和发射光收集。激发光波长范围为340 nm至380 nm，而发射光由光电倍增管（PMT）在510 nm波长处检测。一对二向色镜使这两条光学路径能够在不干扰彼此测量的前提下共享相同的空间位置（图2）。最后，光学相干断层扫描（OCT）使用中心波长为840 nm、光谱宽度为100 nm的宽带光源，该波长与其他两种成像方式明显不同。由于OCT所用光源具有低相干性，来自明场-荧光光源的任何散射光均不会影响编码深度信息的干涉图样。为满足第二个要求，毛细管的外壳设计使得样品的前侧、下侧和上侧平面均具有可访问的光学通路。在实验过程中，两根铂金钩将一条小梁肌固定于充满含氧克雷布斯-亨塞尔莱特（KH）溶液的毛细管内。OCT的检流计扫描头被正交地安置于明场-荧光成像光路，以利用第三个正交的光学平面（图3）。

本文概述了构建一种能够同时成像钙离子、肌节长度和肌肉几何结构装置的设计考量。为展示这些测量能力，我们描述了心室小梁肌的分离过程、所需缓冲溶液的配制，以及组织处理和荧光染料加载过程中的关键步骤 离体 小梁。最后，本文概述了将数据集转化为更具实用性的可视化图像所需的过程。

方案

The University of Auckland's Animal Ethics Committee approved the handling of rats and the preparation of tissue samples.

1. Imaging calibration

1. Brightfield microscope pixel calibration
   1. Fill the measurement chamber with distilled water.
   2. Place a diffraction grating with known lines per µm into the measurement chamber.
   3. Press F1 to enable the capture and adjust the Frame Rate [Hz] until the diffraction grating is clearly visible (Figure 4A). Ensure that the diffraction grating runs parallel with the edge of the frame. Press F1 again to stop capture.
   4. Set the Total Images to Capture? to one, press Ctrl + Shift + S to stream data to disk, and press F1 to capture an image of the diffraction grating.
   5. Open ImageJ and import the diffraction grating image (File > Open > Select Calibration image). Holding shift, draw a line that encompasses 20 light and dark bands of the diffraction grating.
   6. Calibrate the image (Analyze > Set Scale). The length of the line from Step 1.1.5 sets the Distance in Pixels value. Set the Known Distance value to 20 times the lines per µm measurement and the Unit of Length to µm. The inverse of the scale is the number of micrometers represented by a pixel.
2. OCT depth resolution calibration
   1. Measure the thickness of a glass microscope slide using Vernier calipers.
   2. Turn on the OCT laser source.
   3. Cover the galvanometer head and click Get BG to measure the background interference pattern and subtract it from the measurement (Figure 4B).
   4. Clamp the measured glass microscope slide (from Step 1.2.1) in the measurement arm of the OCT.
   5. Click Live Stream to view the OCT image. Adjust the glass microscope slide until it is visible within the B-scan.
   6. To capture the B-scan image, set Range Y (Steps) to one, click Stream B-scan Data?, and click Acquire.
   7. Import the B-scan image into ImageJ (File > Open > Select B-scan). Holding shift, draw a line between the boundaries of the glass microscope slide.
   8. Set the scale (Analyze > Set Scale). Set Known Distance to the measurement collected in Step 1.2.1.
   9. To calculate the depth resolution in air, correct for the refractive index of the microscope slide (n_glass = 1.5175)¹¹ by multiplying the measurement per pixel value by n_glass.
      NOTE: The n_glass quoted is for borosilicate glass. Microscope slides can be made from different materials. Use the appropriate refractive index for the slide measured.
   10. To scale the depth resolution for myocardium, divide the value from Step 1.2.9. by n_myocardium = 1.38 (value reported previously¹²).

2. Muscle sample preparation

1. Prepare the dissection rig.
   1. Pour some of the dissection solution (outlined in Table 1) into a small metal bowl and place in the freezer about an hour before heart excision.
   2. Setup a dissection rig ensuring dissection solution is well-oxygenated (100% oxygen) and has flushed through each of the tubing lines. Fill the dissection chamber with oxygenated dissection solution and loosely tie 3/0 suture around the perfusion catheter.
2. Excise the heart.
   1. Anesthetize 8-10 weeks old Wistar rat using gaseous isoflurane (< 5% in oxygen). Confirm anesthesia by tail pinching.
   2. Position the anesthetized rat in a supine position and inject subcutaneously in the abdominal area with heparin solution (1000 IU/kg). Maintain anesthesia for five more minutes to allow the heparin to circulate.
   3. Retrieve the metal bowl containing dissection solution from the freezer and place it near the euthanization bench.
      NOTE: Avoid freezing the dissection solution completely to enable the complete submersion of the dissected heart.
   4. Transfer the anesthetized rat to the euthanization bench and euthanize by cervical dislocation.
   5. Open the rat chest with scissors, first cutting the body wall along the underside of the ribcage, then the diaphragm, before proceeding along the lateral borders of the ribcage. Lift the chest out of the way.
   6. Grab the heart with one hand while the other hand uses a pair of curved scissors to cut the connecting vessels (aorta, vena cava, etc.).
   7. Quickly submerse the heart in the cold dissection solution.
3. Isolate a trabecula.
   1. Identify the aorta while the heart is in the metal bowl, then transfer the heart to the dissection chamber. Using two curved forceps, pull the aorta over the perfusion cannula.
   2. Hold the aorta in place with one forceps. Meanwhile, open the tubing line to allow the dissection solution to flow through the perfusion cannula.
      NOTE: Aim to complete perfusion within the minute following the heart excision.
   3. Once the coronary vasculature is cleared of blood and the heart is completely perfused with the dissection solution, halt the perfusion flow, and secure the aorta in place using the suture. Turn the flow back on and perfuse the cannulated heart.
   4. Rotate the cannula so that the left coronary artery is visible on the superior surface. Pin the apex of the heart to the bottom of the dissection chamber (Figure 5A). Cut off both atria (Figure 5B).
   5. With a set of spring scissors, cut along the right side of the septum to the apex of the heart (as indicated on Figure 5B). Pin the opened left ventricle to the base of the dissection chamber. Then cut along the left side of the septum, open the right ventricle, and pin it to the base of the dissection chamber, too (Figure 5C).
      NOTE: In order to pin the ventricles in an open position, some papillary muscles will have to be cut. Identify a free-running trabecula in the right ventricle (Figure 5D-E).
   6. Using the spring scissor and a forceps, cut the wall tissue surrounding the trabecula, then cut the wall tissue in half orthogonally to the direction of the trabecula. Trim the wall tissue until its dimension is appropriate for the mounting configuration used. In this case, approximately half the size of a sesame seed (Figure 5F).
      NOTE: Trabeculae can be dissected from the right and left ventricles, but those from the left are typically more turbid and less applicable for sarcomere and geometry measurements.
   7. Leave the excised trabecula in the dissection chamber, continuously superfusing with the dissection solution.

3. Experimental protocol

NOTE: The device¹³ used for this experiment was built in-house and uses custom control code. The necessary considerations for the design of a device built to replicate these data are two independently actuated mounting hooks, a measurement chamber with three optically clear axes (Figure 3), and an external trigger line that synchronizes the brightfield and OCT cameras with the stimulator. The PMT voltage and force signal were collected using analog DAQ cards, the images from the OCT and brightfield microscope were collected using Camera Link frame-grabber cards, and the stimulus signal was collected using a digital I/O card. Data were stored offline using a set of producer consumer loops to maintain temporal alignment.

1. Prepare the cardiomyometer.
   1. Flush hot (~60 °C) water, distilled water (room temperature), and then superfusate solution through the measurement chamber. Continuously bubble the superfusate solution with carbogen.
   2. Turn on the brightfield microscope illumination source and press F1 to enable capture (Figure 4A). Manually adjust the downstream hook until it is centered in the brightfield image. Click Zero Downstream Axis, then Downstream Disabled to enable the motor (Figure 4C). Move the DS Setpoint [um] slider until the end of the hook aligns with the edge of the default region of interest.
      1. Re-zero the downstream axis, then move the DS Setpoint [um] slider to 1000. Repeat the process with the upstream hook, but do not move the US Setpoint [um] slider.
   3. Click Move to Mounting (Figure 4C).
   4. Start the fluorescence illumination system by toggling the Lamp switch before quickly turning on the controller subsystems by toggling the Main switch.
      NOTE: Some UV light sources produce large amounts of ozone. If this is the case, connect an ozone extractor to the outlet vent of the light source and ensure it is running prior to turning on the fluorescence illumination source.
   5. Switch the operational mode to Turbo-Blanking by pressing the Mode button on the front panel, followed by 2, then 1. Press the On-line button to allow the control code to inform operation.
2. Mount the trabecula.
   1. Pause superfusate flow through the measurement chamber. Fill the mounting chamber with the dissection solution.
   2. Using a 1 mL syringe, transport the trabecula from the dissection chamber to the mounting chamber (Figure 5G).
   3. To transfer the trabecula, place the syringe vertically and in contact with the surface of the mounting chamber solution. Allow the trabecula to descend into the mounting chamber via gravity (Figure 5H).
   4. Lower the fluid level in the mounting chamber so that it is in level with the midsection of the hooks.
   5. Adjust the distance between the hooks to reflect the slack length of the trabecula by moving the DS Setpoint [um] slider.
   6. Using a microscope to aid visualization, lightly grip one of the pieces of end tissue with forceps and mount it onto the upstream hook. Mount the other piece of end tissue onto the downstream hook (Figure 5I).
   7. Once securely mounted, move the trabecula back into the measurement chamber (Figure 5J) by pressing Move to Chamber (Figure 4C). Resume superfusate flow and fluid extraction.
   8. Set Stimulus frequency [Hz] to 1, Stimulus duration [ms] to 10, and Stimulus Voltage to 10. Start stimulation by pressing Stimulus On?.
3. Prepare the trabecula.
   1. After about 1 h of acclimatization, gradually decrease the stimulus voltage and stimulus duration in 1 V and 1 ms steps, respectively. A typical set of values is 3 V and 3 ms.
   2. Turn on the brightfield illumination system. Press F1 and select a region of interest that encloses a striated area on the user interface. Click Compute SL? to calculate the average sarcomere length in the highlighted region. Increase the muscle length until the average sarcomere length is 2.32 µm by increasing the Separation Setpoint [um] slider.
      NOTE: Compute SL? uses a 2D FFT outlined in Step 4.3. The region of interest used to calculate the average sarcomere length is typically a 100 µm to 150 µm square. Therefore, as the muscle approaches optimal sarcomere length, 43 to 65 sarcomeres are used to calculate the average sarcomere length.
   3. Move the muscle by adjusting the Centre Setpoint [um] slider on the "Centre and Separation Control tab (Figure 4C) so that the edge of the downstream hook is just visible within the brightfield image. Collect the fluorescence information for ten twitches.
   4. Increase the Centre Setpoint [um] value by 200 and collect another ten twitches worth of fluorescence information. Repeat this process until the brightfield image contains the upstream hook. Collect the final window's worth of fluorescence information.
   5. Return the trabecula to a central position by setting the Centre Setpoint [um] value to 0.
   6. Reduce the stimulus frequency to 0.2 Hz and switch from the KH superfusate to the Fura-2 loading solution (detailed in Table 1).
   7. Measure the fluorescence signal every 10 min by clicking Enable Fluorescence Source on the Stim and Data tab. Visualize the fluorescence signal on the PMT Signal tab.
   8. After the 360 nm signal has increased by a factor of 10 or the duration of the loading procedure has exceeded 2 h, return the stimulus frequency to 1 Hz and switch back to the KH superfusate solution.
   9. Check the ratio measurement every 10 min until the ratio measurement stabilizes, at which point data collection can begin.
4. Collect brightfield- and fluorescence-imaging data.
   1. Return the muscle to the position where the edge of the downstream hook is just present within the brightfield image. Start streaming hardware data by clicking Stream Data to Disk on the Stim and Data tab of the hardware-control user interface. Capture fluorescence information by clicking Enable Fluorescence Source.
   2. On the brightfield imaging user interface, set the capture mode to an external trigger, increase the frame rate to 100 Hz, and set the number of images to capture to 100. Press Ctrl + Shift + S followed by F1 to record the brightfield imaging data for this window.
   3. Increase Centre Setpoint [um] value by 200 and repeat Step 3.4.2. Continue with the scanning protocol until the imaging data has been collected for the final window from Step 3.3.4.
   4. Return the trabecula to a central position by setting the Centre Setpoint [um] value to 0.
5. Collect OCT imaging data.
   1. Turn on the OCT laser source by turning the master key to the | symbol, pressing the power button, followed by the SLDs button.
   2. Cover the galvanometer head and click Get BG to measure the background interference pattern and subtract it from the measurement (Figure 4B).
   3. Set the image capture mode to live-view.
   4. Adjust the y-position until the B-scan image contains the upstream hook only. Divide the muscle length displayed on the control front panel (Figure 4B) by two and subtract the current y-position. Enter this value into the "y-offset" input. Adjust the "x-offset" value until the cross-section of the trabecula is centered in the frame.
   5. With the trabecula centered, scan along the y-axis by adjusting the y-position to find the positions corresponding to the upstream and downstream hooks. Note these positions down. Set Range Y (steps) to the absolute difference between these values divided by ten.
   6. Set the image capture mode to Stimulus Triggered?, Range X (steps) to 100, and click the Set Active Parameters button.
   7. Click Stream B-Scan Data?, then Acquire.
      ​NOTE: The gated imaging protocol requires 200 twitches to capture the entire muscle geometry for a sample 2 mm in length, which corresponds to a capture time of ~3 min 20 s.

4. Process the brightfield image dataset

1. Prepare the images for analysis.
   1. Import images into ImageJ (File > Import > Image Sequence > Select image).
   2. Increase image contrast (Image > Adjust > Brightness/Contrast > Move minimum and Maximum sliders to centralize the image histogram).
   3. Sharpen the images (Process > Filters > Unsharp Mask > set Radius (Sigma) to 1.0 pixels and Mask Weight (0.1-0.9) to 0.6).
   4. Export the image sequence (Save As > Image Sequence > Set the Format to PNG, Start at to 0 and Digits (1-8) to 4).
2. Stitch the images, measure the localized displacement, and compute the local sarcomere lengths.
   1. Open "TrabeculaProcessing.m" (available on request) and set the FolderPath variable to the main folder containing all of the data, and ImagePath to the folder where the image sequence from Step 4.1.4 was saved. Set sections to the number of imaging windows and frames to the number of frames captured per window.
   2. Run the code.
      NOTE: The outputs will be present in the output folder path specified by the user. (By default, the path is set to FolderPath/Output).
3. FFT sarcomere length technique
   1. Use image-processing software to perform a FFT on a region of the image where sarcomeres are highly visible.
   2. Multiply the pixels per µm calibration results from Step 1.1.6 by 1.6 µm and 3.0 µm before calculating the inverse to get the range of spatial frequencies of interest.
   3. Fit an exponential to the FFT result, ignoring the frequency information in the frequency range calculated in Step 4.3.2, and subtract it from the transform result to remove the DC term.
   4. Fit a Gaussian curve to the frequency band of interest.
   5. Calculate the inverse of the peak of the Gaussian curve. This is the average sarcomere length for the region of interest.
      ​NOTE: The FFT calculation and the fitting of the exponential and Gaussian equations were performed using custom LabVIEW code.

5. Process the fluorescence data

1. Subtract the window-dependent autofluorescence from the respective window and compute the quotient of the signals associated with the 340 nm and 380 nm excitation wavelengths.

6. Process the OCT imaging data

1. Prepare the OCT image set for segmentation.
   1. Open ImageJ and import the images (File > Import > Image Sequence). On the file explorer window, this opens, locate the images, select one and click Open.
      NOTE: If the control code for the OCT does not store the images in a format that is readable by ImageJ, convert them to PNG.
   2. To ease visualization, organize the image sequence into a hyperstack (Image > Hyperstacks > Stack to Hyperstack). In the dialogue box that opens, set the number of slices to the number of B-scans per slice and the X to the number of slices along the length of the trabecula.
   3. Draw a rectangle that encloses the trabecula. Confirm that it encloses the entire volume through time by using the sliders on the hyperstack window. Crop the image to the window (Image > Crop).
   4. Remove slices that contain images of the mounting hooks (Stacks > Tools > Slice Keeper). Select the range of slices that contains trabecula information only.
2. Train WEKA segmentation.
   1. Open WEKA segmentation (Plugins > Segmentation > Trainable Weka Segmentation).
   2. Set the selection mode to Freehand.
   3. Click Settings and adjust the classifier and training settings. (For this model, the following Training features were used: Gaussian blur, Sobel filter, Hessian, Difference of Gaussians, membrane projections, Bilateral, and Lipschitz. Membrane thickness was set to 1, Membrane patch size to 8, Minimum sigma to 1, and Maximum sigma to 32. The classifier was set to FastRandomForest and the classifier options were set to: batchSize 100, maxDepth to 32, numFeatures to 32, numThreads to 0, and numTrees to 200.)
   4. Manually segment images until training results in satisfactory segmentations.
   5. Save the classifier.
3. Segment the processed B-scans
   1. Launch WEKA segmentation following Step 6.2.1.
   2. Load the classifier from Step 6.2.5.
   3. Click Create Result.
   4. Convert the images to 8-bit (Image > Type > 8-bit).
   5. Convert the images to binary (Process > Binary > Make Binary > Default method and default background).
   6. Save as image sequence (PNG).
4. Calculate the average CSA in the segmented B-scan images.
   1. Count the number of white pixels in a binary B-scan image.
   2. Multiply the pixel area by the calibrated depth resolution (from Step 1.2) and 10 µm (the distance between neighboring A-scans).
   3. Repeat for all of the B-scans between the hooks and average the measurements.
5. Convert segmented image into a mesh.
   1. Open "OCTmain.m" (available on request) and set imageDirectory to the folder containing the output from Step 6.3.6. Set outputPath as necessary.
   2. Set slices to the value of "Range Y (steps)" (Step 3.5.5) and frames to the value of "Repeat X" (Step 3.5.6), z_dim to the depth resolution (Step 1.2.10), and x_dim & y_dim to the value assigned to 10.
   3. Click Run.

结果

为了获取此处所示心室肌小梁全长的区域性 Ca²⁺ 和明场信息，共需要七个肌肉位置。图6 表明，这种移动并未干扰收缩力，说明主动收缩力的产生不存在位置依赖性。

使用光学相干断层扫描以100 Hz的速率采集的B扫描图像，通过ImageJ插件WEKA¹⁴进行分割（图7A）。由于横向分辨率（10 µm）与轴向分辨率（心肌中为1.73 µm）不同，每个横截面均呈现扭曲。该扭曲通过将图像的深度轴按横向分辨率与深度分辨率的比值进行缩放而得以校正。图7B、C显示，在对小梁肌原始C扫描图像进行缩放后，其几何形状近似为圆柱形。测量室壁的反射信号有时会与肌肉信号重叠（图7A、B），但分割软件可经训练以消除该干扰（图7D、E）。完成分割后，可在整个收缩过程中计算肌肉长度方向上的横截面积（图7F）。需注意，该特定小梁肌具有一小分支从主干伸出，分支的运动在距小梁肌约0.75 mm处清晰可见。最后，分割后的图像可转换为网格，以辅助构建几何模型（图7G）。

以100 fps的速率在小梁不同位置捕获的成像数据被拼接在一起，生成小梁的单个完整图像（图8A）。这些图像的分辨率为0.535 µm/像素。在相邻窗口的重叠区域使用线性加权函数有助于可视化，并最小化明场图像中晕影效应的影响。为测量荧光信号，采用波长分别为340 nm、365 nm和380 nm的光以600 Hz的频率循环照射小梁上一个540 µm × 540 µm的区域。由340 nm和380 nm激发光引起的发射荧光强度之比，反映了用Fura-2负载后小梁细胞内的钙离子浓度。由于该测量为比值形式，有效测量速率为200 Hz。每个窗口测得的细胞内Ca²⁺瞬变信号经平均（n = 10）后，与其成像区域对齐（图8B）。尽管瞬变信号的峰值表现相对一致，但沿小梁900 µm至1800 µm区域的舒张期[Ca²⁺]较低。类似地，位移追踪结果（图8C）和肌节长度计算结果（图8D）也表明存在区域性差异。所使用的无标记追踪技术能够在具有足够对比度的情况下处理每个像素的位移。在绘制肌节长度分布图时，采用128像素 × 128像素（约67 µm × 67 µm）的互相关区域来计算局部肌节长度。当样本接近最佳肌节长度时，该区域大约包含29个肌节。在处理这些数据时，每个互相关窗口质心之间的步长（在x-和y-方向上）设定为50像素（约26 µm）。肌节长度估计的可靠性基于对FFT信号进行高斯拟合的宽度和幅度进行检验。在小梁0 µm至500 µm的肌肉区域，这些条件未被满足，因此无法在此区域计算肌节长度信息。结合相关的位移数据，推测该区域的肌节在收缩期发生了拉长。与此推测一致的是，小梁右侧区域的平均肌节长度在该时期缩短。综合各图板提供的信息可见，横截面积最大的区域并未产生最显著的力。假设Ca²⁺瞬变的区域性变化具有近似平滑的梯度，则图8B表明，最大振幅的Ca²⁺瞬变发生在沿小梁1300 µm至1600 µm之间的某处。位移图显示，运动幅度最小的区域与Ca²⁺瞬变峰值位置高度吻合。然而，该区域的横截面积是样本中最小的。基于这些数据，可以推断该区域产生了最大的应力。

图1心肌细胞计的标注图像。 主要光学元件均已标出。插图为显微镜物镜的放大后视图 原位测量室下方 请点击此处以查看此图的放大版本。

图2：明场与荧光显微镜同步成像的光路示意图。 荧光显微镜的照明光源为氙弧灯，其输出光在340 nm、365 nm和380 nm波长之间循环切换。弧光灯的输出光路中包含一个截止波长为409 nm的二向色镜，该镜将紫外光反射至另一反射镜，再导入荧光显微镜物镜。物镜将激发光聚焦至样品，并收集波长更长（510 nm）的发射光。该发射光可透过第一个二向色镜，但无法透过第二个二向色镜（其截止波长为552 nm）。随后，场镜将反射光聚焦至光电倍增管（PMT）的传感器上。与此同时，明场显微镜的照明光源（660 nm LED）位于样品上方。透射光经聚光镜聚焦至样品，由20×荧光物镜捕获透射图像。明场照明所用波长超过两个二向色镜的截止波长，因此可同时透过两者，最终成像于CMOS相机传感器上。 请点击此处查看该图的放大版本。

图3测量室支架设计 (A) 测量室支架的等距视图，叠加了光路。明场照明来自上方表面（z-轴）；荧光照明来自下方表面（z-轴），而OCT测量臂信号垂直于另一照明轴（y-轴）。实验过程中，两根铂金钩将心室肌条固定于一根玻璃毛细管内，该毛细管作为测量腔室。每个铂金钩由音圈电机控制，其位置通过激光干涉仪测定。系统将实时位置与用户设定的目标值进行比较，并通过现场可编程门阵列（FPGA）中编码的PID控制器最小化误差。B测量室 原位 在明场照明开启的情况下。后视图如下所示 图1 插图. (C) 流经测量室支架的灌流液流动示意图。灌流液从模块后方进入，并沿箭头所示方向流动。上游和下游电极建立电场刺激，以诱发测量室内固定的心室小梁肌收缩。蓝色阴影表示实验过程中灌流液流经的区域。 请点击此处以查看此图的放大版本。

图4：图像采集与控制软件的前面板。（A）明场成像用户界面。（B）OCT成像用户界面。（C）硬件控制用户界面。请点击此处查看此图的放大版本。

图 5：小梁肌解剖与固定方案。（A）置于解剖室中的经Langendorff灌流的大鼠心脏。 （B） 同一心脏，心房已被切除。虚线表示切开心室的切割路径。（C）打开的心脏，以暴露两个心室的内部结构。虚线框标示出小梁肌通常所在区域。（D）已切除的右心室壁区域（与C图中虚线框所示区域相同）。虚线突出显示三条小梁肌。（E）从D图三条小梁肌中选取的一条。（F）E图中小梁肌，周围心室壁组织已被去除。（G）分离后的小梁肌置于1 mL注射器末端。（H）小梁肌置于固定室中。（I）小梁肌固定于两根铂金钩之间。（J）固定于钩上的小梁肌置于测量室中（图 3B）。绿色斑点为第一片二向色镜产生的伪影。（K）固定后小梁肌在测量室中的另一角度视图。小梁肌与显微镜物镜之间的距离约为1 mm。请点击此处查看该图的放大版本。

图6：力测量的位置依赖性。 来自各个成像位置的肌肉产生的力（n = 7）叠加显示。平均主动产力为 0.527 mN ± 0.003 mN，达到50%收缩的时间为 77.1 ms ± 0.3 ms，达到50%舒张的时间为 328.1 ms ± 0.9 ms（所有数据均以均值 ± 标准误表示）。请点击此处查看该图的放大版本。

图7：OCT成像分析。（A）WEKA分割示例。肌肉的分割横截面以红色突出显示，背景以绿色突出显示。（B）小梁原始C扫描数据的俯视图。图像顶部附近的亮斜线为测量室壁的反射。（C）小梁原始C扫描数据的侧视图。（D）分割后OCT数据的俯视图。（E）分割后OCT数据的侧视图。（F）沿小梁长度方向（x轴）随时间（y轴）变化的横截面积。沿肌肉长度的平均横截面积为0.0326 mm² ± 0.0005 mm²（均值 ± 标准误）。（G）小梁的网格模型。该网格大致与F图面板中的横截面积图对齐。请点击此处查看该图的放大版本。

图8：明场与荧光成像分析。（A）心室小梁的拼接图像（七个成像区域）。（B）沿心室小梁长度方向的 Ca²⁺ 瞬变信号。（C）每个成像区域的平均 x 方向位移。正值表示向右运动，负值表示向左运动。（D）具有足够图像对比度的每个成像区域的平均肌节长度。请点击此处查看该图的放大版本。

表1：溶液配制表 请点击此处下载该表格。

讨论

在本研究中，我们提出一种配置方案，可集成包含明场、荧光和光学相干断层扫描（OCT）成像的三个光学系统，以采集正在主动收缩组织的多模态数据 离体 心室小梁（图1 和 图2）。这种协调一致的整合之所以能够实现，得益于测量室的设计（图3）以实现OCT与明场-荧光轴的正交排列。肌肉固定系统在同步量化表征心肌兴奋-收缩动态关键指标的成功中同样起着至关重要的作用。该系统的创新之处在于能够在对肌肉机械性能无明显干扰的情况下进行肌肉扫描操作。图6）。结合成像配置与用于力测量的电机钩系统，该系统可评估心肌细胞内钙信号的区域性异质性²⁺ 瞬时位移与肌节长度，以及在整个收缩时程中收缩小梁的宏观几何信息图7 和 图8).

鉴于明场-落射荧光成像系统在心脏研究实验室中的广泛应用，只需进行一些简单的硬件调整即可重现本研究结果。本文介绍了用于整合明场-落射荧光成像与光学相干断层扫描（OCT）的图像处理工具包，这对于分析潜在的收缩异质性至关重要。OCT 的集成需要一条无遮挡的光路，而门控成像则需要在刺激装置与 OCT 及明场成像相机之间连接外部触发信号线，同时还需要能够在整个测量腔室内移动样品的肌肉固定钩。所需的所有后期处理软件和方法均可免费获取。特别是所使用的分割软件 WEKA¹⁴ 为开源软件。无标记材料点追踪技术⁸、肌节长度分析、门控三维成像¹⁰ 以及网格生成代码同样可获取，并可根据要求向通讯作者申请获得。

肌肉活性、Fura-2 的最佳负载以及成像聚焦是成功实验的三大基础。使用含有 BDM 的解剖溶液以防止肌肉挛缩，通过注射器运输肌肉组织，持续对溶液进行氧合，以及在实验当天配制新鲜的实验溶液，这些措施均有助于实现较高的肌肉活性。在用 Fura-2AM 对小梁肌进行负载之前，必须针对所研究的每种条件采集自发荧光信号，因为自发荧光可能对测得的 Ca²⁺ 瞬变信号产生显著影响¹⁵。由于 Fura-2AM 负载溶液中需加入表面活性剂 Pluronic-F127 以促进染料负载，这使得溶液的氧合过程变得复杂。为应对该表面活性剂引起的过度起泡问题，在负载溶液中加入少量消泡剂可使操作者提高氧合速率，从而增加小梁肌在整个负载过程中维持功能活性的可能性。最后，成像聚焦必须沿肌肉全长保持均匀，以最大化明场和荧光信号的信噪比。

本研究所述方法存在两个局限性。首先是荧光显微镜的空间分辨率问题。尽管光学相干断层扫描（OCT）和明场成像的空间分辨率较高，但荧光显微镜的分辨率受限于在540 µm × 540 µm成像窗口内捕获的荧光信号体积的积分。可通过使用高增益电荷耦合器件（CCD）相机替代光电倍增管（PMT）来捕获荧光信号，从而提高荧光显微镜的空间分辨率，但代价是信噪比降低¹⁶。第二个局限性在于可研究的小梁直径，这关系到可测量的肌节长度和几何 深度。用于计算肌节长度的加窗快速傅里叶变换（windowed-FFT）方法虽具有提高空间分辨率的优势，但其鲁棒性有所降低（图8D）。当研究浑浊或大直径的小梁时，由于较大组织样本中肌节条带对比度降低，FFT的可分辨能力将显著下降。同样，在OCT成像中，成像深度超过300 µm时的背向反射信号在分割阶段将过于微弱而无法分辨。因此，本技术仅适用于直径小于300 µm的小梁。然而，不建议研究大直径样本，因为在高频刺激条件下，肌肉核心区域可能存在氧气扩散不足的问题¹⁷。

我们的方法能够评估健康和病变肌肉中与肌肉几何结构相关的离子力学功能，为深入理解心肌生理学、病理生理学和药理学提供了有力手段。本文所述的图像处理流程可提取关键数据，有助于更深入地理解收缩异质性。充分发挥此类丰富数据集潜力的一个途径是构建数学模型，以整合和解读这些数据，并利用我们的设备通过实验验证模型所做出的预测。

材料

本文使用的材料清单

姓名	公司	目录编号	评论
2,3-丁二酮单肟	Acros Organics	150375000
20× 显微镜物镜	Nikon	CFI Super Fluor 20×	NA 0.75
二维振镜	Thorlabs	GVSM002/M
50-50 分束器	Thorlabs	FC850-40-50-APC
90-10 分束器	Thorlabs	TW850R2A2
模拟输入模块	National Instruments	NI-9205	以 200 kHz 采样 PMT 信号
明场成像光源	CoolLED	PE-2	660 nm LAM
宽带光源	Superlum	Broadlighter-840
CaCl2	Sigma-Aldrich	C4901
Cameralink 采集卡	National Instruments	NI-1429	明场成像图像采集卡
碳酰气（5% CO₂ + 95% O₂）	BOC	气体编号：181
聚光镜	Nikon	LWD 0.52
D-(+)-葡萄糖	Merck	108337
数据采集设备（DAQ）	National Instruments	NI-6259	触发振镜运动
二向色镜 1	Semrock	FF409-Di03
二向色镜 2	Semrock	FF552-Di02
衍射光栅	Wasatch Photonics	1200 线/mm @840 nm
二甲基亚砜	Sigma-Aldrich	276855
Direct-Q 3 UV 系统	Merck Millipore	ZRQSVR3WW	纯水制备系统
干式恒温器	Corning	6875-SB	LSE 数字干浴器
FIJI	ImageJ		开源图像处理软件
Fura-2AM 五钾盐	Thermofisher	F14186
硬件 FPGA 卡	National Instruments	NI-7813R	同时控制明场图像采集的触发
肝素	Pfizer	61024
HEPES	PanReac AppliChem	A1069
倒置显微镜	Nikon	TI-DH 照明支柱
异氟烷	MedSource	VAPDRUGISO250
KCl	Sigma-Aldrich	P9541
KH2PO4	Sigma-Aldrich	P5655
线扫描相机	Basler	spL2048-70km	光谱仪相机
磁力搅拌器	IKA	3810000	RCT basic
Matlab	Mathworks		数据处理代码
MgCl2	Sigma-Aldrich	M8266
MgSO4.7H2O	Sigma-Aldrich	M1880
NaCl	Sigma-Aldrich	71376
NaH2PO4.2H2O	Sigma-Aldrich	71505
NaHCO3	Sigma-Aldrich	S6014
OCT FPGA 卡	National Instruments	NI-1483R
氧气瓶	BOC	气体编号：100D
pH 计	Mettler Toledo	MP220
光电倍增管	Hamamatsu	H7422-20
Powerload	Thermofisher	P10020
超发光二极管	Broadlighter	D-840
跨阻放大器	定制
三(羟甲基)氨基甲烷	Sigma-Aldrich	252859
Wistar 大鼠	Vernon Jansen Unit		8 – 10 周龄
氙弧灯	Sutter Instrument	DG-4	Lambda DG-4