1. MRI-based radiomics analysis of sacroiliitis 2. Detecting sacroiliitis with the use MRI-based radiomics and machine learning.

Diagnosing active inflammatory sacroiliitis (IS) at an early stage is crucial to prescribe drugs that modulate the disease progression and significantly delay or prevent the debilitating forms of axial spondyloarthropathy. Conventional radiography and Computed tomography are not sensitive to acute inflammatory findings and mostly detect chronic structural lesions. MRI, on the other hand, is the modality of choice to detect bone marrow edema or osteitis. However, this is a laborious process that takes years to master. Moreover, differentiating the causality of these lesions can be challenging, even for experienced doctors. Machine Learning (ML) has proven its ability in various fields, by figuring out patterns not easy to spot in multi-dimensional datasets. The goal of this thesis is to create a radiomic signature that executes this task, to serve as an ancillary tool for medical doctors to diagnose active sacroiliitis. Sacroiliac joints from 565 individuals were segmented from axial fluid-sensitive MRI images and radiomic features extracted with the 3D Slicer software. After selecting the most informative of them, we tried certain ML methods to come up with the optimal model