Texture analysis of articular cartilage applied on magnetic resonance relaxation time maps. Gray level co-occurrence matrices and local binary patterns
Thesis event information
Date and time of the thesis defence
Place of the thesis defence
Oulu University Hospital, Auditorium 12
Topic of the dissertation
Texture analysis of articular cartilage applied on magnetic resonance relaxation time maps. Gray level co-occurrence matrices and local binary patterns
Doctoral candidate
Master of Science Arttu Peuna
Faculty and unit
University of Oulu Graduate School, Faculty of Medicine, Research Unit of Medical Imaging, Physics and Technology
Subject of study
Medical physics
Opponent
Professor Hannu Eskola, Tampere University
Custos
Docent Eveliina Lammentausta, Oulu University Hospital
Texture analysis of magnetic resonance images reveals additional information for knee osteoarthritis
Analysis of an image texture (TA) is an efficient way to increase the information obtained from medical images. Mathematical computations are performed to evaluate and quantitatively characterize the structure and pathology of the target tissue. Resulting variables may provide complementary information regarding the subtle macromolecular changes occurring in various diseases, such as knee osteoarthritis (OA).
OA is a common chronic disease that is characterized by disabling pain, joint dysfunction, and morphological alterations in multiple synovial joint structures. One of the structures affected in OA is the articular cartilage (AC) that provides frictionless movement and load-dampening properties for the joint articulation.
Current diagnostic tools lack sensitivity for the early stage of OA. TA combined with quantitative magnetic resonance imaging techniques, such as T2 relaxation time mapping, allow sensitive evaluation of AC degeneration and disease onset, and may lead to more accurate diagnostics, individualized treatment planning, and better patient outcomes.
The aim of this thesis was to assess two texture analysis methods, gray level co-occurrence matrix (GLCM) and local binary pattern (LBP), on the Oulu knee osteoarthritis (OKOA) dataset which comprises 80 confirmed OA patients and an equal number of healthy controls. The TA method’s ability to discern OA subjects from healthy subjects was evaluated on the whole dataset and on an early OA -stage simulating subset. Texture analyzed data included T2, T1 and T2 relaxation time maps and DESS images. Developed techniques were compared against the current mean relaxation time analysis scheme. Furthermore, the parametric outcomes of TA were subjected to machine learning classification algorithms and tested for automatic segregation of the study groups.
TA demonstrated excellent performance in discerning the study groups and appears to be more sensitive to the early changes than the current mean relaxation time focused analysis methods. TA can be applied on various quantitative contrasts and resulting outcomes can be utilized in automated classification pipelines for OA detection. TA demonstrates great potential for further research evaluations and clinical applications and these findings warrant further inquiries into the topic.
OA is a common chronic disease that is characterized by disabling pain, joint dysfunction, and morphological alterations in multiple synovial joint structures. One of the structures affected in OA is the articular cartilage (AC) that provides frictionless movement and load-dampening properties for the joint articulation.
Current diagnostic tools lack sensitivity for the early stage of OA. TA combined with quantitative magnetic resonance imaging techniques, such as T2 relaxation time mapping, allow sensitive evaluation of AC degeneration and disease onset, and may lead to more accurate diagnostics, individualized treatment planning, and better patient outcomes.
The aim of this thesis was to assess two texture analysis methods, gray level co-occurrence matrix (GLCM) and local binary pattern (LBP), on the Oulu knee osteoarthritis (OKOA) dataset which comprises 80 confirmed OA patients and an equal number of healthy controls. The TA method’s ability to discern OA subjects from healthy subjects was evaluated on the whole dataset and on an early OA -stage simulating subset. Texture analyzed data included T2, T1 and T2 relaxation time maps and DESS images. Developed techniques were compared against the current mean relaxation time analysis scheme. Furthermore, the parametric outcomes of TA were subjected to machine learning classification algorithms and tested for automatic segregation of the study groups.
TA demonstrated excellent performance in discerning the study groups and appears to be more sensitive to the early changes than the current mean relaxation time focused analysis methods. TA can be applied on various quantitative contrasts and resulting outcomes can be utilized in automated classification pipelines for OA detection. TA demonstrates great potential for further research evaluations and clinical applications and these findings warrant further inquiries into the topic.
Last updated: 23.1.2024