Filtering by
- All Subjects: Medical Imaging
- Creators: Graves, William
- Creators: Vernon, Brent
- Status: Published
Handheld, portable confocal laser endomicroscopy (CLE) is being explored in neurosurgery for its ability to image histopathological features of tissue at cellular resolution in real time during brain tumor surgery. Over the course of examination of the surgical tumor resection, hundreds to thousands of images may be collected. The high number of images requires significant time and storage load for subsequent reviewing, which motivated several research groups to employ deep convolutional neural networks (DCNNs) to improve its utility during surgery. DCNNs have proven to be useful in natural and medical image analysis tasks such as classification, object detection, and image segmentation.
This thesis proposes using DCNNs for analyzing CLE images of brain tumors. Particularly, it explores the practicality of DCNNs in three main tasks. First, off-the shelf DCNNs were used to classify images into diagnostic and non-diagnostic. Further experiments showed that both ensemble modeling and transfer learning improved the classifier’s accuracy in evaluating the diagnostic quality of new images at test stage. Second, a weakly-supervised learning pipeline was developed for localizing key features of diagnostic CLE images from gliomas. Third, image style transfer was used to improve the diagnostic quality of CLE images from glioma tumors by transforming the histology patterns in CLE images of fluorescein sodium-stained tissue into the ones in conventional hematoxylin and eosin-stained tissue slides.
These studies suggest that DCNNs are opted for analysis of CLE images. They may assist surgeons in sorting out the non-diagnostic images, highlighting the key regions and enhancing their appearance through pattern transformation in real time. With recent advances in deep learning such as generative adversarial networks and semi-supervised learning, new research directions need to be followed to discover more promises of DCNNs in CLE image analysis.
This thesis scrutinizes CLE technology for its ability to provide real-time intraoperative in vivo and ex vivo visualization of histopathological features of the normal and tumor brain tissues. First, the optimal settings for CLE imaging are studied in an animal model along with a generational comparison of CLE performance. Second, the ability of CLE to discriminate uninjured normal brain, injured normal brain and tumor tissues is demonstrated. Third, CLE was used to investigate cerebral microvasculature and blood flow in normal and pathological conditions. Fourth, the feasibility of CLE for providing optical biopsies of brain tumors was established during the fluorescence-guided neurosurgical procedures. This study established the optimal workflow and confirmed the high specificity of the CLE optical biopsies. Fifth, the feasibility of CLE was established for endoscopic endonasal approaches and interrogation of pituitary tumor tissue. Finally, improved and prolonged near wide-field fluorescent visualization of brain tumor margins was demonstrated with a scanning fiber endoscopy and 5-aminolevulinic acid.
These studies suggested a novel paradigm for neurosurgery-pathology workflow when the noninvasive intraoperative optical biopsies are used to interrogate the tissue and augment intraoperative decision making. Such optical biopsies could shorten the time for obtaining preliminary information on the histological composition of the tissue of interest and may lead to improved diagnostics and tumor resection. This work establishes a basis for future in vivo optical biopsy use in neurosurgery and planning of patient-related outcome studies. Future studies would lead to refinement and development of new confocal scanning technologies making noninvasive optical biopsy faster, convenient and more accurate.
X-ray phase contrast imaging (XPCI) is a novel imaging method that utilizes phase information of X-rays in order to produce images. XPCI allows for highly resolved features that traditional X-ray imaging modalities cannot discern. The objective of this experiment was to model initial simulations predicting the output signal of the future compact x-ray free electron laser (CXFEL) XPCI source. The signal was reported in tonal values (“counts”), where MATLAB and MATLAB App Designer were the computing environments used to develop the simulations. The experimental setup’s components included a yttrium aluminum garnet (YAG) scintillating screen, mirror, and Mako G-507C camera with a Sony IMX264 sensor. The main function of the setup was to aim the X-rays at the YAG screen, then measure its scintillation through the photons emitted that hit the camera sensor. The resulting quantity used to assess the signal strength was tonal values (“counts”) per pixel on the sensor. Data for X-ray transmission through water, air, and polyimide was sourced from The Center for X-ray Optics’s simulations website, after which the data was interpolated and referenced in MATLAB. Matrices were an integral part of the saturation calculations; field-of-view (FOV), magnification and photon energies were also necessary. All the calculations were compiled into a graphical user interface (GUI) using App Designer. The code used to build this GUI can be used as a template for later, more complex GUIs and is a great starting point for future work in XPCI research at CXFEL.
X-ray phase contrast imaging (XPCI) is a novel imaging method that utilizes phase information of X-rays in order to produce images. XPCI allows for highly resolved features that traditional X-ray imaging modalities cannot discern. The objective of this experiment was to model initial simulations predicting the output signal of the future compact x-ray free electron laser (CXFEL) XPCI source. The signal was reported in tonal values (“counts”), where MATLAB and MATLAB App Designer were the computing environments used to develop the simulations. The experimental setup’s components included a yttrium aluminum garnet (YAG) scintillating screen, mirror, and Mako G-507C camera with a Sony IMX264 sensor. The main function of the setup was to aim the X-rays at the YAG screen, then measure its scintillation through the photons emitted that hit the camera sensor. The resulting quantity used to assess the signal strength was tonal values (“counts”) per pixel on the sensor. Data for X-ray transmission through water, air, and polyimide was sourced from The Center for X-ray Optics’s simulations website, after which the data was interpolated and referenced in MATLAB. Matrices were an integral part of the saturation calculations; field-of-view (FOV), magnification and photon energies were also necessary. All the calculations were compiled into a graphical user interface (GUI) using App Designer. The code used to build this GUI can be used as a template for later, more complex GUIs and is a great starting point for future work in XPCI research at CXFEL.