Development of Instrumentation for Fluorescence-Guided Surgery

用于体内的分子靶向探针的快速发展使得人们对结合这些探针信息的临床应用越来越感兴趣。特别是，需要在手术过程中可视化这些靶向探针的技术，特别是识别要切除或保存的组织。当观察组织中的荧光探针时，主要的困难通常是将探针荧光与组织自身荧光分离。由于自体荧光与荧光团具有不同的光谱特性，因此可以很容易地使用多光谱成像进行分离，其中可以为每个像素取几个完整的发射光谱。虽然声光可调滤波器的实现大大提高了这些技术的速度，但在商用系统中获取图像立方体所需的最短时间仍然是几秒钟。当使用效率较低或剂量较低的荧光探针成像时，图像采集时间可能长达几分钟。这种数据采集速率使得在外科手术过程中使用多光谱成像提供实时反馈变得不切实际。即使用于诊断目的，在获取图像立方体所需的时间内，被摄体不可避免的运动也会导致不可接受的空间分辨率损失。

The rapid development of molecularly targeted probes for use in vivo has led to growing interest in clinical applications that incorporate information from these probes. In particular, there is a need for techniques to visualize these targeted probes during surgery, particularly to identify tissue either for removal or preservation. When looking at fluorescent probes in tissue, the major difficulty is usually separating the probe fluorescence from the tissue autofluorescence. Because the autofluorescence has different spectral properties than the fluorophore, it can easily be separated using multispectral imaging, in which several full emission spectra can be taken for each pixel. Although the implementation of acousto-optic tunable filters has greatly increased the speed of these techniques, the minimum time required to acquire an image cube is still several seconds with commercially available systems. When imaging using fluorescent probes with lower efficiency or when used at lower doses, the image acquisition time can be as long as several minutes. This data acquisition rate makes the use of multispectral imaging to provide real-time feedback during a surgical procedure impractical. Even when used for diagnostic purposes, the unavoidable motion of the subject during the time required to acquire an image cube can lead to an unacceptable loss of spatial resolution. This project focuses on the development of instrumentation for incorporating fluorescent and multi-spectral imaging into surgical applications, using two major approaches. The first approach is to develop a system for real-time visualization of fluorescent probes to guide surgery. The second approach is to develop a system for diagnostic applications, to provide a white-light stack of images for spatial registration of the multispectral image cube. As an alternative to multispectral imaging, aggressive filtering of both emission and excitation light can help to minimize the effects of tissue autofluorescence. Although the signal level drops as the spectral window is narrowed, this can be overcome by using a sensitive camera, such as a cooled CCD, ICCD, or even an EMCCD. Because a single spectral window is used, substantially faster data acquisition is possible. The immediate application for this instrument was the identification of peritoneal metastases in ovarian cancer, using a GSA-Rhodamine Green probe developed in NCI. The first prototype instrument, used for mock cytoreductive surgery in a murine model of ovarian cancer, had sufficient sensitivity to identify labeled metastases at an image acquisition rate of five frames per second with specificity comparable to that achieved with a multispectral system. An improved prototype, which permitted image acquisition at fifteen frames per second with enhanced sensitivity, was used for cytoreductive surgery on anesthetized animals with similar success to the previous mock surgery. This year, we also continued development on a separate multispectral system to provide a white light image stack for physical registration of the multispectral image cube, using a 92:8 beam splitter and a low-cost monochrome CCD camera in parallel with the multispectral instrumenation. Preliminary experiments on a cervical cancer animal model are underway. The hardware developed for this application could be easily adapted to incorporate a second camera into the instrument for fluorescence guided surgery; this second camera could be used either for a second fluorescence image, in order to provide a simple autofluorescence correction, or for a pseudo-white light image of the surgical field. Finally, some initial proof-of-principle in vivo imaging experiments were performed using upconverters, in this case upconverting nanocrystals that absorb 980 nm light and emit in the visible to near-infrared. These novel compounds could allow single-wavelength imaging with no detectable autofluorescence; the tissue background is so low that a white light image will likely be required to provide anatomical reference. Although considerable work remains in developing these nanoparticles for medical applications, this is a promising avenue for future efforts.