Semiautomated multiplexed quantum dot-based in situ hybridization and spectral deconvolution.

Gene expression profiling has identified several potentially useful gene signatures for predicting outcome or for selecting targeted therapy. However, these signatures have been developed in fresh or frozen tissue, and there is a need to apply them to routinely processed samples. Here, we demonstrate the feasibility of a potentially high-throughput methodology combining automated in situ hybridization with quantum dot-labeled oligonucleotide probes followed by spectral imaging for the detection and subsequent deconvolution of multiple signals. This method is semiautomated and quantitative and can be applied to formalin-fixed, paraffin-embedded tissues. We have combined dual in situ hybridization with immunohistochemistry, enabling simultaneous measurement of gene expression and cell lineage determination. The technique achieves levels of sensitivity and specificity sufficient for the potential application of known expression signatures to biopsy specimens in a semiquantitative way, and the semiautomated nature of the method enables application to high-throughput studies.

Texture analysis of speckle in optical coherence tomography images of tissue phantoms.

Optical coherence tomography (OCT) is an imaging modality capable of acquiring cross-sectional images of tissue using back-reflected light. Conventional OCT images have a resolution of 10-15 microm, and are thus best suited for visualizing tissue layers and structures. OCT images of collagen (with and without endothelial cells) have no resolvable features and may appear to simply show an exponential decrease in intensity with depth. However, examination of these images reveals that they display a characteristic repetitive structure due to speckle. The purpose of this study is to evaluate the application of statistical and spectral texture analysis techniques for differentiating living and non-living tissue phantoms containing various sizes and distributions of scatterers based on speckle content in OCT images. Statistically significant differences between texture parameters and excellent classification rates were obtained when comparing various endothelial cell concentrations ranging from 0 cells/ml to 25 million cells/ml. Statistically significant results and excellent classification rates were also obtained using various sizes of microspheres with concentrations ranging from 0 microspheres/ml to 500 million microspheres/ml. This study has shown that texture analysis of OCT images may be capable of differentiating tissue phantoms containing various sizes and distributions of scatterers.

Autofluorescence removal, multiplexing, and automated analysis methods for in-vivo fluorescence imaging.

The ability to image and quantitate fluorescently labeled markers in vivo has generally been limited by autofluorescence of the tissue. Skin, in particular, has a strong autofluorescence signal, particularly when excited in the blue or green wavelengths. Fluorescence labels with emission wavelengths in the near-infrared are more amenable to deep-tissue imaging, because both scattering and autofluorescence are reduced as wavelengths are increased, but even in these spectral regions, autofluorescence can still limit sensitivity. Multispectral imaging (MSI), however, can remove the signal degradation caused by autofluorescence while adding enhanced multiplexing capabilities. While the availability of spectral "libraries" makes multispectral analysis routine for well-characterized samples, new software tools have been developed that greatly simplify the application of MSI to novel specimens.

Texture analysis of optical coherence tomography images: feasibility for tissue classification.

Optical coherence tomography (OCT) acquires cross-sectional images of tissue by measuring back-reflected light. Images from in vivo OCT systems typically have a resolution of 10 to 15 mm, and are thus best suited for visualizing structures in the range of tens to hundreds of microns, such as tissue layers or glands. Many normal and abnormal tissues lack visible structures in this size range, so it may appear that OCT is unsuitable for identification of these tissues. However, examination of structure-poor OCT images reveals that they frequently display a characteristic texture that is due to speckle. We evaluated the application of statistical and spectral texture analysis techniques for differentiating tissue types based on the structural and speckle content in OCT images. Excellent correct classification rates were obtained when images had slight visual differences (mouse skin and fat, correct classification rates of 98.5 and 97.3%, respectively), and reasonable rates were obtained with nearly identical-appearing images (normal versus abnormal mouse lung, correct classification rates of 64.0 and 88.6%, respectively). This study shows that texture analysis of OCT images may be capable of differentiating tissue types without reliance on visible structures.

Semi-automated Algorithm for Localization of Dermal/ Epidermal Junction in Reflectance Confocal Microscopy Images of Human Skin.

The examination of the dermis/epidermis junction (DEJ) is clinically important for skin cancer diagnosis. Reflectance confocal microscopy (RCM) is an emerging tool for detection of skin cancers in vivo. However, visual localization of the DEJ in RCM images, with high accuracy and repeatability, is challenging, especially in fair skin, due to low contrast, heterogeneous structure and high inter- and intra-subject variability. We recently proposed a semi-automated algorithm to localize the DEJ in z-stacks of RCM images of fair skin, based on feature segmentation and classification. Here we extend the algorithm to dark skin. The extended algorithm first decides the skin type and then applies the appropriate DEJ localization method. In dark skin, strong backscatter from the pigment melanin causes the basal cells above the DEJ to appear with high contrast. To locate those high contrast regions, the algorithm operates on small tiles (regions) and finds the peaks of the smoothed average intensity depth profile of each tile. However, for some tiles, due to heterogeneity, multiple peaks in the depth profile exist and the strongest peak might not be the basal layer peak. To select the correct peak, basal cells are represented with a vector of texture features. The peak with most similar features to this feature vector is selected. The results show that the algorithm detected the skin types correctly for all 17 stacks tested (8 fair, 9 dark). The DEJ detection algorithm achieved an average distance from the ground truth DEJ surface of around 4.7µm for dark skin and around 7-14µm for fair skin.

Fine-needle aspiration of follicular adenoma versus parathyroid adenoma: the utility of multispectral imaging in differentiating lesions with subtle cytomorphologic differences.

BACKGROUND: Multispectral image analysis is an emerging tool that utilizes both spatial and spectral image information to classify images that can be used for the differentiation between benign versus malignant cells. The aim of the current study was to analyze the ability of this tool in differentiating subtle cytologic differences that cannot be appreciated by the human eye. Herein, the authors used fine-needle aspirations (FNAs) of follicular adenoma (FA) and parathyroid adenoma (PA) as a test case. METHODS: The Nuance platform was used to collect image stacks that were subsequently analyzed with CRI-MLS software, a neural network-based artificial intelligence system that can classify images using automatically "learned" spatial-spectral features. CRI-MLS was trained on random, well-preserved FA cells and PA cells from the training set (n = 45 cells each). An algorithmic solution was developed and then validated on an independent series comprised of 1904 FA cells from 5 FA cases and 690 PA cells from 5 PA cases. RESULTS: The solution from the CRI-MLS classifier showed 1876 FA cells (98.5%) as true FA and 28 FA cells (1.5%) as false PA, whereas 663 PA cells (96%) were true PA and 27 PA cells (4%) were false FA. The summary result of this solution was a sensitivity of 98.5%, a specificity of 96.1%, and a positive predictive value of 98.6%. CONCLUSIONS: The best spatial-spectral imaging solution was able to correctly classify 2534 of 2594 cells (98%) and misclassified only 55 of 2594 cells (2%). These data suggest that this technology may be valuable in a clinical setting to help differentiate and classify morphologically similar lesions.