China 
Africa 
Explore chapters and articles related to this topic
Imaging the Living Eye
Published in Margarida M. Barroso, Xavier Intes, In Vivo, 2020
Brian T. Soetikno, Lisa Beckmann, Hao F. Zhang
The fundus camera magnifies and photographs the retinal fundus, providing two-dimensional (2D) anatomical information. Images can be used to determine the progression of retinal diseases, especially those that involve retinal vasculature or retinal lesions. With monochromatic filters, the fundus camera can also enhance its contrast, for example, of hemoglobin in imaging vasculature or other pigmented chromophores in the retina, such as drusen in age-related macular degeneration (AMD). Fluorescent dyes can also be imaged by the fundus camera by using filters designed to pass specific wavelength ranges of fluorescence emission. This technique is primarily used for fluorescein angiography, a method to map the retinal vasculature after intravenous injection of the fluorescein dye. Fundus cameras can be integrated with other imaging modalities, such as photoacoustic ophthalmoscopy (Liu et al., 2013) or optical coherence tomography (Song et al., 2012) to guide these techniques by providing a large, real-time field-of-view for the operator. Multiwavelength and hyperspectral fundus cameras, which can obtain spectroscopic measurements of hemoglobin absorption inside the vasculature, have also been tested for quantifying the oxygen saturation in the retinal blood vessels (Hardarson et al., 2006; Li et al., 2017).
Overview: An Evolving State of the Art in Tissue Engineering
Published in Claudio Migliaresi, Antonella Motta, Scaffolds for Tissue Engineering, 2014
The creation of accurate 2D and 3D visual representations of tissue and organ structures for use in RP and other manufacturing techniques are generated by medical scanning devices such as magnetic resonance imaging (MRI) and computed tomography (CT). A combination of CT with positron emission tomography (PET), CT/PET, provides a more refined image that has the potential of measuring tissue functions such as blood flow.31 Electron beam CT (EBCT) scanners produce images in fractions of a second, even permitting the scanning of a beating heart,32 and diffraction-enhanced imaging (DEI) delivers more detailed images of soft tissue than is possible with other CT techniques.33 Imaging techniques are also employed in examination of diseased tissues and organs. Fluorescein angiography, for example, is used for imaging vascular patterns in the human retina for diagnosis of diabetic retinopathy and retino-vascular occlusive disease. Image resolution of the vascular patterns of vessels of various dimensions present a fractal pattern that can be quantified and used to diagnose early-stage vascular disease in the human retina.34 Development of Fourier domain optical coherence tomography (FD-OCT) has been proposed as an improvement over fluorescein angiography due to improved sensitivity and acquisition time.35 Synchrotron radiation micro-CT has also been used for both morphological characterization and quantification of cell distributions down to the micrometer level of tissues, thereby
Deep CNN based microaneurysm-haemorrhage classification in retinal images considering local neighbourhoods
Published in Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, 2022
Mahua Nandy Pal, Ankit Sarkar, Anindya Gupta, Minakshi Banerjee
Nayak et al. (2008) and Acharya et al. (2008) deal with different stages of diabetic retinopathy. Nayak et al. classified images of DR (normal, proliferative and non-proliferative) with the artificial neural network using 140 images graded by ophthalmologists. They reported image classification accuracy of 93%, sensitivity of 90% and specificity of 100%. Acharya et al. classified DR images (no DR, mild DR, moderate DR, severe DR and proliferative DR) with an accuracy of 82% and specificity of 88%. Pratt et al. also used CNN architecture with a high-end GPU on a big data set of 80,000 images and demonstrated 75% validation accuracy. Lam et al. (2018a)used transfer learning with pre-trained GoogLeNet and AlexNet models from ImageNet. They reported image level classification accuracies of 74.5%, 68.8%, and 57.2% in 2-class, 3-class and 4-class classification for the detection of diabetic retinopathy affected images. Pratt et al. (Pratt, et al., 2016b)also reported image-level sensitivity of 95% and accuracy of 75% on 5,000 validation images. They resized images to 512 × 512 resolution. Model training time complexity is quite high. Tavakoli et al. (2013) evaluated their work on a different modality, fluorescein angiography images. In fluorescein angiography, a dye is injected into the vein and images of the back surface of the eye are captured as the dye propagates through the vessels. This procedure facilitates the diagnosis of abnormalities in the vessels. However, in fundus photography, a special camera is used to capture the image of the back surface of the eye. They provided Image level DR detection sensitivities with two rural databases and ROC online database. Orlando et al. (2018) provided the ROC curve and AUC values for image-level evaluation with random forest classifier using both deep learned and handcrafted features. Sun (2019) is another DR image diagnostic system. All these works address image level retinal classification.