A Supervoxel-Based Approach for Unsupervised Abnormal Asymmetry Detection in Mr Images of the Brain

Unsupervised brain anomaly detection in MR images

“…problem and sample type. Texture [6,[37][38][39][40], shape features [41][42][43], and, more recently, deep-learning-based features [35,[44][45][46] are common feature examples adopted in medical image analysis problems. Overall, machine learning can be either supervised or unsupervised.…”

“…Medical image analysis commonly uses outlier detection mainly for detecting anomalies (lesions). One-class classication (OCC) -also called unary classication -is a class of techniques commonly used for this purpose [40,[50][51][52][53]. Consider a training dataset with only medical images of healthy subjects -also known as control images.…”

“…In this section, we describe typical preprocessing steps applied to raw MR images [11,40,51,78,[78][79][80][81][82] with a focus on the techniques used throughout this thesis. The combination of these steps is problem dependent and empirically estimated [82].…”

“…The image operation resumes to a post-processing of the forest attributes, such as the root labels, optimum paths, and connectivity values. IFT has been successfully applied in dierent domains, such as image ltering [116], image descriptor [117,118], segmentation [41,[119][120][121][122][123], superpixel segmentation [40,61,124], representation [125], (semi) supervised classication [126][127][128], and data clustering [129,130].…”

“…One could rst segment the structure by using any automated segmentation tool, such as FreeSurfer [144], to dene the VOI. However, it might be essential to analyze VOIs that do not include any specic object segmented by these available tools [40,51]. We then propose a patch-based model (PBM) to localize VOIs without any segmentation, as illustrated in Fig.…”