Hierarchical Cluster Analysis to Aid Diagnostic Image Data Visualization of MS and Other Medical Imaging Modalities

Abstract

Perceiving abnormal regions in the images of different medical modalities plays a crucial role in diagnosis and subsequent treatment planning. In medical images to visually perceive abnormalities' extent and boundaries requires substantial experience. Consequently, manually drawn region of interest (ROI) to outline boundaries of abnormalities suffers from limitations of human perception leading to inter-observer variability. As an alternative to human drawn ROI, it is proposed the use of a computer-based segmentation algorithm to segment digital medical image data.Hierarchical Clustering-based Segmentation (HCS) process is a generic unsupervised segmentation process that can be used to segment dissimilar regions in digital images. HCS process generates a hierarchy of segmented images by partitioning an image into its constituent regions at hierarchical levels of allowable dissimilarity between its different regions. The hierarchy represents the continuous merging of similar, spatially adjacent, and/or disjoint regions as the allowable threshold value of dissimilarity between regions, for merging, is gradually increased.This chapter discusses in detail first the implementation of the HCS process, second the implementation details of how the HCS process is used for the presentation of multi-modal imaging data (MALDI and MRI) of a biological sample, third the implementation details of how the process is used as a perception aid for X-ray mammogram readers, and finally the implementation details of how it is used as an interpretation aid for the interpretation of Multi-parametric Magnetic Resonance Imaging (mpMRI) of the Prostate.

Machine learning utilising spectral derivative data improves cellular health classification through hyperspectral infra-red spectroscopy

“…For a binary diagnostic problem of healthy versus death as used for this study, k = 2, where k is the number of required classes or clusters, creating a two-cluster problem designed for the separation of the two different cell states. k-means is often used in medical image processing to provide an improvement in image segmentation of different tissue types or classes, where ML tool is specifically applied to increase the accuracy of segmentation, as well as the total throughput of images as compared to human image analysis [ 20 ]. k-means has also been applied in the field of spectral analysis for clinical diagnosis, with endoscopic imaging producing a large spectral data-set, which were sampled and analysed using a pre-determined number of clusters to represent the different clinical diagnoses [ 21 ].…”

Hierarchical segmentation using equivalence test (HiSET): Application to DCE image sequences

“…Segmentation is known to be a useful tool for common image analysis to get SNR improvement, it is an even more important preprocessing steps for DCE-imaging analysis. Indeed, in addition to the expected SNR improvement, it allows to: 1/ reveal the functional anatomy that is hardly visible on static images [Hanson et al, 2017], therefore allowing the correct selection of different tissues [Irving et al, 2016b;Heye et al, 2013]; 2/ summarize the functional information, hence reducing the amount of parameter extractions; 3/ increase the conspicuity of the images and facilitate the fit of the TCs [Hou et al, 2014;Selvan et al, 2017;Giannini et al, 2010]; 4/ analyze genuine tissues and obtain correct parameters; 5/ provide and guarantee the detection, measurement and characterization of lesions in clinical practice [Baumgartner et al, 2005]; 6/ improve the communication between clinicians by providing synthetic pictures.…”

The ImageJ ecosystem: Open‐source software for image visualization, processing, and analysis

“…This is important for a wide array of biological studies that include quantifying the proximity of fluorescent‐labeled proteins, 9 tracking cell fates over time, 10 automating cell counting, 11 tracking invading cancer cells, 12 collecting whole‐slide information, 13 quantifying and characterizing cells, such as microglia, within the brain, 14 or registering multiview light sheet fluorescence microscopy datasets to study development 15,16 . Image analysis also serves a crucial biomedical role for diagnostic interpretation 17–20 . As the prevalence of large multidimensional datasets continues to grow, the ability to manually take measurements not only becomes impractically time‐consuming, but the sensitivity, accuracy, objectivity, and reproducibility of doing so can become greatly inhibited 21,22 .…”