MSc Thesis Defence by Jintao Ren


Toward automatic detection and localization of bone erosion in HR-pQCT of hand


Rheumatoid arthritis (RA) is a disease affecting primarily joins, hand and wrists. Inflammation provokes the destruction of cortical surface of bones and creates cavities, the process is called bone erosion. Erosion quantification is necessary in monitoring the progress of RA. This is usually performed by semi automatic analysis of high-resolution peripheral quantitative computed tomography

(HR-pQCT) of hands. In this thesis, the first, to our knowledge, fully automatic framework for detection and localization of bone erosion from HR-pQCT is proposed. It combines a set of traditional image processing methods and deep learning neural network algorithms to achieve automatic detection and position of hand bone erosion areas. It achieves a high detection accuracy and a low false positive rate in the test data.

François Lauze, DIKU
Arash Moaddel Haghighi, DIKU
Kresten Krarup Keller, AUH