It seems like a simple question. As old as you are, right?
Dana Duren, Ph.D., director of orthopaedic research for the Department of Orthopaedic Surgery, Sports Medicine, and Rehabilitation, and associate professor in the Division of Morphological Sciences and Biostatistics at Wright State University’s Lifespan Health Research Center, said while skeletons develop in the same way, the timing and tempo can vary from person to person. One child may develop early, another later; one may develop rapidly and another at a slower pace. On the outside, it’s not easy to tell if an 8-year-old has six years or 12 to finish growing.
For some children, finding out exactly how mature their skeleton is and whether they’re done growing is critical. Short stature, for example, can indicate a number of health issues, from hormonal disorders to genetic diseases. Diagnosis and treatment can depend on accurately calculating the age of the child’s bones and how much time they have left to grow. In an age when medical care has seen so many technological advances, Duren said physicians and researchers still evaluate bone age by looking at X-rays, using information gathered from the great-grandparents of today’s children.
$1.6 million grant to update data
Duren and her colleagues hope to change that. The National Institutes of Health’s National Institute of Arthritis, Musculoskeletal and Skin Diseases recently awarded a $1.6 million grant to Duren’s team to update the data used to determine bone age, expand the data to include a racially and ethnically diverse population, and more accurately predict whether a child’s skeletal maturation will speed up or slow down. A second grant of $50,000 from the Boonshoft School of Medicine’s Translational Research Development Grant program will allow them to create an open-source, semi-automated program to calculate bone age.
Duren’s co-investigators include Richard Sherwood, Ph.D., director of the Division of Morphology and Biostatistics; Ramzi Nahhas, Ph.D., a biostatistician with the division; Travis Doom, Ph.D., and Thomas Wischgoll, Ph.D., both associate professors of computer science and engineering; Elizabeth Ey, M.D., a pediatric radiologist and medical director of medical imaging at Dayton Children’s Hospital; and Babette Zemel, Ph.D., of the University of Pennsylvania School of Medicine Department of Pediatrics.
Their work will focus on the Fels Method for determining bone age, Duren said, which was originally based on data from Wright State’s Fels Longitudinal Study.
The study, started in Yellow Springs, Ohio, in 1929, collected a myriad of information on the growth and development of children. Participants were enrolled before birth and tracked throughout their lives. Many of the original participants had children, grandchildren, and even great-grandchildren followed in the study, as well. The data from the Fels study was used to create the pediatric growth charts used from 1978 to 2000 to monitor a child’s growth compared to his or her peers.
The Fels Method
In the 1980s, researchers led by Dr. Alex Roche used Fels data to create the Fels Method for determining skeletal maturity, or bone age. Roche’s new method used 98 indicators to evaluate a child’s left hand and wrist X-ray. The method looks at individual bones to see how many appear on the X-ray, their size, and their shape. Each indicator is assigned a value, and those values go into an algorithm to calculate the bone age. The Fels Method also reports a standard error, an important consideration when bone age is used to determine the course of medical treatment.
“It’s a great standard,” Duren said, “but is underutilized in clinical practice.”
Duren said even Ey, one of her research partners for the project, doesn’t use the Fels Method in her clinical practice. Researchers have tried to automate the process, but with limited success. Radiologists instead tend to use the older method, Greulich-Pyle, based on a series of hand and wrist X-rays done in the 1930s. The X-rays were compiled into a book, or atlas, based on age. Physicians compare a patient’s image to the ones in the book to determine bone age.
The hand and wrist have 27 bones, plus the radius and ulna in the forearm. They begin as mostly cartilage. As the bones undergo ossification, more become visible on the X-ray. Images from older teens and young children don’t take as much time to score as images from children from about 10 years old into the early teens.
Ey said most radiologists are trained in the Greulich-Pyle method. While X-rays don’t always exactly match the ones in the atlas, she said it comes pretty close and only takes about five minutes. The Fels Method, by comparison, would take her much longer. “The problem is that each assessor’s eye goes to a different thing,” Duren said.
With Gruelich-Pyle, two experts could look at the same X-ray from a patient and arrive at two different ages, especially if they focus on different markers.
Duren hopes their work will make the Fels Method as fast and easy for physicians as the Greulich-Pyle.
“We want it to be as accurate as having a human eye on the X-ray, but with the speed of the simpler atlas method,” she said.
Diversifying the data
The issues with current bone age calculation go beyond just speed. The children used to set the Greulich-Pyle and Fels standards were Caucasian, and even the ones used in the Fels Method would be well into their 40s and 50s by now.
“That’s not what our population looks like today,” Ey said. Physicians realize that affects the accuracy of the bone age calculations.
Sherwood said secular trends show children are maturing earlier and undergoing puberty at younger ages. Puberty and bone development are closely linked, making it possible the way the bones mature are different, as well. Using a 40- or 80-year-old standard may not provide an accurate bone age for today’s children.
The lack of diversity in the initial samples raises issues as well. Researchers don’t know if the bones of an African-American, Asian, or Hispanic child develop at the same rate as a Caucasian child, but still use X-rays of Caucasian children as the standard. Using a Caucasian standard for children of other races could produce an inaccurate bone age.
“We know there are a lot of skeletal differences between African-American and Caucasian children, for example,” Sherwood said. “If you don’t look at (bone growth), you’re never going to know.”
When Ey does a bone age evaluation for a healthy child of normal height, it tends to measure one or more standard deviations above the current standards, a clear sign bone age has shifted, due in part to nutrition improving growth rates.
Their work will do more than just update the data, said Nahhas. They hope to be able to track the variations in how the growth process has changed in the decades since the first images were evaluated.
The Fels Method’s 98 indicators can provide the medical community with a tremendous amount of information about how children grow and mature. Sherwood said maturation of some indicators will likely occur earlier, while others may have slowed down. Pinpointing those changes could provide insight into why children are reaching puberty earlier. In girls, especially, that can stunt their growth, since the hormones released by the onset of menstruation will shut down bone growth soon after. In children with short stature, it limits the window for effective diagnosis and treatment.
Duren and her team will be looking at more than 8,000 hand and wrist X-rays collected from racially diverse groups of children throughout the country between 2002 and 2007. Many of the images will come from the Bone Mineral Density in Childhood Study conducted by the National Institute of Child Health and Human Development.
Duren’s three research assistants, Carol Cottom, Kimberly Lever, and Sharon Lawrence, will evaluate each X-ray by hand using the Fels Method.
FelsXpress will provide faster diagnosis
The Translational Grant from the School of Medicine will allow Doom and Wischgoll to take the information gathered and develop a beta version of a program for computers, phones, and tablets called FelsXpress. The program would provide a faster diagnosis using updated information, something they hope doctors will embrace.
Doom said the goal is to have a physician upload the X-ray, enter the child’s chronological age, and have the computer do all of the work, but still allow the physician to review the information that led to the bone age calculation.
“It can’t be a black box,” Doom said. “They have to be able to understand how it was done.”
Using pattern recognition technology, the program will be able to ask for any additional information and then provide a visual result and an indication of how confident it is in the answer. Doom said that allows the physician to concentrate his or her efforts on the most important indicators for each patient.
“We’re not changing the science,” Doom said. “We want to make (the Fels Method) more accessible.”
Physicians know the data used in the Greulich-Pyle Atlas needs to be updated Ey said, and they also understand the Fels Method is superior. She believes her colleagues will embrace the new data and, hopefully, FelsXpress.
Duren and her team will need to work to get FelsXpress into the right hands.
“People think the Fels Method is too difficult; they don’t want to learn a new method,” Sherwood said. “We have to find a way to show the people who train the radiologists this method is the best.”
— Shannon Neal