A SINGLE-PIXEL HYPERSPECTRAL IMAGING SYSTEM FOR CANCER MARGIN DETECTION
We have developed a hyperspectral imaging (HSI) system based on a single pixel camera design to detect differences in tissue properties based on the optical reflectance and autofluorescence spectra of the tissue. The long-term goal of this project is to develop an HSI system to be used as a surgical navigation aid. Surgical excision of malignant tissue continues to be the foundation of treatment for most solid mass tumors. A significant remaining challenge for cancer surgery is ensuring that no residual malignant tissue is left behind, as recurrent tumors lead to high mortality rates. Unfortunately, cancerous tissue is often indistinguishable from healthy tissue under visual inspection during surgery and there are few diagnostic imaging tools to aid the surgeon in the determination of the tumor margins in real time. Cancer surgeons are in need of additional intraoperative imaging modalities for use during surgery to clearly delineate tumor margins and identify areas of residual disease. Recent research has demonstrated that optical spectroscopy can be used to distinguish between healthy and diseased tissue. HSI is a hybrid imaging modality that combines imaging and spectroscopy and provides a 2D image that contains spectral information in each pixel. Because HSI captures both spatial and spectral information, this technique has potential applications for noninvasive disease diagnosis and surgical guidance. Conventional HSI systems employ spatial or spectral scanning techniques that reconstruct the spectral image after scanning is complete. We are developing a different type of HSI system based on a single pixel camera design. The single pixel design provides an imaging architecture that is more flexible than traditional HSI scanning techniques and can provide better performance in several key areas including improved operation at low light levels and enhanced dynamic range.A single pixel camera uses a single detector to create a 2-D image of a scene rather than using an array of detectors. The design of a single-pixel camera relies on the mathematical theory and algorithms of compressive sampling (CS), which is based on the idea that a small number of linear projections of a compressible image contain enough information for reconstruction. A single-pixel camera uses a digital micromirror device (DMD) as a spatial light modulator to optically calculate linear projections of a scene onto pseudo-random binary patterns. Hadamard matrices are used as the binary patterns. A single-pixel imaging system produces an image by rapidly obtaining many measurements of the intensity of a scene using different Hadamard matrices. The signal and the corresponding Hadamard code are saved for reconstruction, which occurred after all codes had been displayed on the DMD. This method of single-pixel imaging based on CS can be made hyperspectral by replacing the single detector with a spectrometer. The HSI system was built and images of known objects were acquired to test the spectral and spatial resolution of the camera and to determine operating parameters for future studies. The software necessary to acquire and reconstruct images was developed in LabView and Matlab. The test images were imaged in fluorescence and reflectance modes simulating future study conditions. The compressive error was tested and relations between image reconstruction quality and the number of Hadamard codes sent to the DMD were quantified. The ability of the single-pixel HSI system to distinguish between healthy and unhealthy tissue was initially tested using an ex vivo porcine tissue model. The autofluorescence emission from collagen (400 nm) and NAD(P)H (475 nm) along with differences in the optical reflectance spectra were used to differentiate between healthy and thermally damaged tissue. Thermal lesions were created in porcine skin (n = 12) and liver (n=15) samples using an IR laser. The damaged regions were clearly visible in the hyperspectral images. Sizes of the thermally damaged regions as measured via hyperspectral imaging were compared to sizes of these regions as measured in white light images and via physical measurement. Good agreement between the sizes measured in the hyperspectral images, white light imaging and physical measurements was found. The HSI system can differentiate between healthy and damaged tissue. The ability of the HSI system to distinguish between healthy and cancerous tissue was evaluated by imaging human pancreatic tissue samples ex vivo. Differences in the optical reflectance spectra were used to identify healthy and malignant pancreatic tissue. Tissue samples from 20 patients were imaged with the HSI system and these images were compared to white light and histological analysis of these samples. An overall sensitivity of 74.80±9.18% and a specificity of 68.59±10.43% as measured from reflectance HSI was found which confirms the system is sensitive to the changes in tissue caused by the presence pancreatic cancer. Differences in the optical autofluorescence emission from collagen (400 nm) and NAD(P)H (475 nm) were also used to identify healthy and malignant pancreatic tissue in a small subset of samples (n=2).Finally, polarization imaging capability was added to the HSI system. Thermal lesions were created in porcine skin (n = 8) samples using an IR laser. Sizes of the thermally damaged regions as measured via hyperspectral polarization imaging were compared to sizes of these regions as measured in reflectance HSI and white light imaging of the samples. Good agreement between the sizes measured in the polarization HSI, reflectance HSI and white light images was found. These results confirmed the sensitivity of the camera to changes in the tissues polarization properties.