Medical Image Processing and Analysis Techniques for Detecting Giant Cell Arteritis

Medical image segmentation and classification algorithms are commonly used in clinical applications. Several automatic and semiautomatic segmentation methods were used for extracting veins and arteries on transverse and longitudinal medical images. Recently, the use of medical image processing and analysis tools improved giant cell arteries (GCA) detection and diagnosis using patient specific medical imaging. In this chapter, we proposed several image processing and analysis algorithms for detecting and quantifying the GCA from patient medical images. The chapter introduced the connected threshold and region growing segmentation approaches on two case studies with temporal arteritis using ultrasound (US) and magnetic resonance imaging (MRI) imaging modalities extracted from the Radiopedia Dataset. The GCA detection procedure was developed using the 3D Slicer Medical Imaging Interaction software as a fast prototyping open-source framework. GCA detection passes through two main procedures: The pre-processing phase, in which we improve and enhances the quality of an image after removing the noise, irrelevant and unwanted parts of the scanned image by the use of filtering techniques, and contrast enhancement methods; and the processing phase which includes all the steps of processing, which are used for identification, segmentation, measurement, and quantification of GCA. The semi-automatic interaction is involved in the entire segmentation process for finding the segmentation parameters. The results of the two case studies show that the proposed approach managed to detect and quantify the GCA region of interest. Hence, the proposed algorithm is efficient to perform complete, and accurate extraction of temporal arteries. The proposed semi-automatic segmentation method can be used for studies focusing on three-dimensional visualization and volumetric quantification of Giant Cell Arteritis.

Tessarine and Quaternion-Valued Deep Neural Networks for Image Classification

Many image processing and analysis tasks are performed with deep neural networks. Although the vast majority of advances have been made with real numbers, recent works have shown that complex and hypercomplex-valued networks may achieve better results. In this paper, we address quaternion-valued and introduce tessarine-valued deep neural networks, including tessarine-valued 2D convolutions. We also address initialization schemes and hypercomplex batch normalization. Finally, a tessarine-valued ResNet model with hypercomplex batch normalization outperformed the corresponding real and quaternion-valued networks on the CIFAR dataset.