Funding source: Arthritis Research UK
Investigators: Tim Cootes, Claudia Lindner and Daniel Perry
Research Associate: Adrian K. Davison
What are the aims of this research?
This research aims to develop a system to accurately measure the bones in the hip joint using X-ray images. By studying the X-rays and the clinical symptoms of people with two common hip diseases, both during childhood and in later life, they will develop methods to predict likely outcomes of the disease, and to choose the most appropriate treatment.
Why is this research important?
Hip deformities are painful disorders that affect 1 in 500 children. Children with hip deformities are more likely to develop osteoarthritis and often need hip replacement surgeries as young adults. In some cases, this hip replacement surgery could be avoided by indextreatment in childhood. Unfortunately, very little is known about the relationship between childhood, adolescent and adult hip shapes, making it difficult for doctors to know which children will go on to develop osteoarthritis as adults. This study will develop a tool that allow doctors to recognise which children are likely to develop osteoarthritis and would most benefit from treatment as early as possible.
Introduction to Perthes Disease and SUFE
Legg-Calvé-Perthes disease (Perthes) is an idiopathic disease in children occuring between the ages of 2-14 years, with boys being affected 5 times more than girls. Perthes disease is usually analysed through radiographic images in the anterior-posterior (AP) or frog lateral views of the hip. The disease affects the femoral head by reducing the blood flow to the growth plate, causing bone necrosis that, if not treated, will lead to deformities in the femoral head as it grows and ossifies. It is unknown what exactly causes Perthes disease, however environmental, congenital and socio-economic issues have been associated with Perthes
Slipped Upper Femoral Epiphysis (SUFE) also affects boys more than girls with boys presenting later (10-17 years) than girls (8-15 years), contrasting with Perthes where the diagnosis occurs in earlier years. In SUFE, the growth plate between the upper femur and femoral head becomes unstable, with the bone breaking through the weakened growth plate. The femoral head then ‘slips’ backwards and downwards out of place. The disease can be described in two ways: stable and unstable, where the latter will be more difficult to treat and manage. Some risk factors of SUFE include obesity and hyperthyroidism, however the exact cause is still unknown.
We use the well established Random Forest Regression Voting Constrained Local Model (RF-CLM) applied to the X-ray images (in AP views) to locate landmark points in two age groups: Ages 2-11 and 12-18. We split into these two groups to address the large variations that occur in younger children compared with the older, who resemble closer to an adult femur. One of the biggest differences is the existence of a growth plate that slowly dissappears as the femur goes from multiple bones into one. In the images below, you can see that the younger group landmark points take into account the growth plate and have 58 points in total. The older group does not have a distinct growth plate line on X-rays, so these points are not included for the 42 points used. An object detector based on Random Forests (RF) is used to automatically initialise the RF-CLM model on each image. The RF-CLM returns the points related to the age group, where parameters for shape, texture and appearance of the hip are extracted. Each of these parameters can then be used as features to classify between Perthes disease and healthy, with the aim to predict long-term outcomes for surgeons to choose appropriate treatment.
How will the findings benefit patients?
This project will help researchers to get a better understanding of two childhood diseases affecting the hip. This research aims to create better ways to accurately measure the hip and the severity of these and other diseases affecting the hip joint. These measurements can be used to help predict future outcomes and allow doctors to select the best treatment for each child, reducing long-term pain and disability.
T.F.Cootes, M.Ionita, C.Lindner and P.Sauer, "Robust and Accurate Shape Model Fitting using Random Forest Regression Voting", ECCV 2012 (PDF)
T.F. Cootes, D. Cooper, C.J. Taylor and J. Graham, "Active Shape Models - Their Training and Application." Computer Vision and Image Understanding. Vol. 61, No. 1, Jan. 1995, pp. 38-59. (PDF)