创新背景

对大脑进行成像是一种平衡行为。工具需要足够快才能捕获快速事件，例如神经元放电或血液流过毛细血管，并且它们需要显示不同尺度的活动，无论是在整个大脑中还是在单个动脉的水平上。
这些事项可以单独完成，但很难一起完成。

创新过程

在新研究中，Yao和他的团队揭示了他们如何通过开发超快光声显微镜（UFF-PAM）来解决这一长期存在的问题。

光声显微镜利用光和声音的特性来捕获全身器官、组织和细胞的详细图像。该技术使用激光将光发送到目标组织或细胞中。当激光击中细胞时，它会立即升温并膨胀，产生超声波，并传回传感器。

UFF-PAM依靠硬件进步和机器学习算法的结合来升级技术。在硬件方面，多边形扫描系统向更大的区域发送更多的激光脉冲，而新的扫描机制允许激光扫描仪和超声传感器同时工作。Yao说，这些变化使他们的设备速度翻了一番，使UFF-PAM成为光声领域最快的成像技术。

Yao和他的团队随后开发了一种机器学习算法，提高了图像的分辨率。他们利用之前实验中收集的400多张老鼠大脑图像，训练它识别大脑中的血管系统。尽管每个大脑都是独一无二的，但该算法学会了如何识别共同的结构，并利用这一知识填补之前缺失的像素。

研究人员通常以较慢的速度拍摄，得到的图像看起来和高分辨率图像一样详细，而且不需要牺牲整个视野。为了证明这一概念，研究小组使用了UFF-PAM来可视化小鼠大脑中的血管是如何对缺氧、药物引起的低血压和缺血性中风做出反应的。在缺氧挑战中，UFF-PAM追踪氧气如何在大脑中流动，并表明低水平的氧气会导致血管扩张。

在第二个挑战中，研究小组使用了硝普钠(SNP)，这是一种常用来治疗高血压的药物。此前，研究人员认为SNP会导致大脑中所有的血管扩张。但Yao和他的团队却证明，只有较大的血管会打开，而较小的血管会收缩。

在最后的挑战中，研究小组使用了UFF-PAM来观察大脑对中风的反应和开始恢复的情况。研究小组发现，中风后，受影响区域的血管会立即收缩。这导致它们相邻的血管也收缩，这种现象被称为扩散去极化波。由于大视场和高成像速度，该团队能够精确地确定波的起始位置，并在波在大脑中传播时跟踪其运动。

创新关键点

UFF-PAM依靠硬件进步和机器学习算法的结合来升级技术。在硬件方面，多边形扫描系统向更大的区域发送更多的激光脉冲，而新的扫描机制允许激光扫描仪和超声传感器同时工作。

创新价值

未来研究人员将使用UFF-PAM来探索其他大脑疾病模型，如痴呆、阿尔茨海默病甚至长冠肺炎。甚至可以将该工具的使用扩展到大脑以外的器官，如心脏、肝脏和胎盘。传统上，这些器官一直处于运动状态，因此成像工具需要以更快的速度操作。

A combination of hardware innovations and machine learning algorithms can reveal rapid brain activity

In the new study, Yao and his team revealed how they solved this long-standing problem by developing an ultra-fast photoacoustic microscope (UFF-PAM).

Photoacoustic microscopy uses the properties of light and sound to capture detailed images of organs, tissues and cells throughout the body. The technology uses lasers to send light into a target tissue or cell. When a laser hits a cell, it immediately heats up and expands, generating ultrasound waves that are sent back to sensors.

Uff-pam relies on a combination of hardware advances and machine learning algorithms to upgrade its technology. On the hardware side, the polygon scanning system sends more laser pulses to a larger area, while the new scanning mechanism allows the laser scanner and ultrasound sensor to work simultaneously. Yao says these changes have doubled the speed of their device, making UFF-PAM the fastest imaging technology in the photoacoustic field.

Yao and his team then developed a machine learning algorithm that improved the resolution of the images. Using more than 400 images of a mouse's brain collected in previous experiments, they trained it to identify the blood vessels in the brain. Although each brain is unique, the algorithm learns how to identify common structures and uses that knowledge to fill in previously missing pixels.

Researchers typically shoot at slower speeds, resulting in images that look just as detailed as high-resolution images without sacrificing the entire field of view. To demonstrate this concept, the team used UFF-PAM to visualize how blood vessels in the mouse brain respond to hypoxia, drug-induced hypotension, and ischemic stroke. In the hypoxia challenge, UFF-PAM tracks how oxygen flows through the brain and shows that low levels of oxygen cause blood vessels to dilate.

In the second challenge, the team used sodium nitroprusside (SNP), a drug commonly used to treat high blood pressure. Previously, researchers thought SNPS caused all the blood vessels in the brain to dilate. But Yao and his team have shown that only the larger vessels open, while the smaller ones contract.

For the final challenge, the team used UFF-PAM to see how the brain responds to the stroke and begins to recover. The team found that immediately after a stroke, blood vessels in the affected area constrict. This causes their adjacent blood vessels to contract as well, a phenomenon known as diffuse depolarization waves. Thanks to the large field of view and high imaging speed, the team was able to pinpoint where the waves started and track their motion as they traveled through the brain.