E-News Exclusive

Radiologists Reliably Differentiate Rickets and CMLs

Rickets and classic metaphyseal lesions (CMLs) exhibit distinct radiographic signs, and radiologists can reliably differentiate these two entities, according to a study in the American Journal of Roentgenology. Noting both high interobserver agreement and diagnostic performance for differentiating the two entities in this seven-center study, “recognition that CMLs mostly occur in children younger than 6 months and are unusual in children older than one year may assist interpretations,” notes corresponding author Boaz Karmazyn from the Riley Hospital for Children in Indianapolis.

Karmazyn and colleagues’ retrospective study included children younger than two years old who underwent knee radiographs from January 2017 to December 2018 and either had rickets (25-hydroxy vitamin D <20 ng/mL and abnormal knee radiographs) or knee CMLs and a diagnosis of child abuse from a pediatrician. Eight radiologists independently interpreted radiographs for rickets or CML diagnoses, rating confidence levels and logging associated radiographic signs.

Ultimately, children with CML were younger than children with rickets (3.9% vs 65.7% >1 year old). The rate of false-positive moderate or high-confidence interpretations was 0.6% for CML and 1.6% for rickets. Only a single child with CML and low vitamin D received an interpretation of combined CML and rickets.

Reiterating that less- and more-experienced pediatric and nonpediatric radiologists had high diagnostic performance in differentiating rickets and CML—regardless of the presence of vitamin D deficiency, with few false-positive interpretations for these diagnoses—the authors conclude, "findings suggestive of both rickets and CML should be viewed as indeterminate.”

— Source: American Roentgen Ray Society

AI Diagnoses Osteoporosis on Hip X-rays

A new method that combines imaging information with AI can diagnose osteoporosis from hip X-rays, according to a study in Radiology: Artificial Intelligence. Researchers say the approach could help speed treatment to patients before fractures occur.

People with osteoporosis, a skeletal disease that thins and weakens bones, are susceptible to fracture associated with bone fragility, resulting in poor quality of life and increased mortality. According to statistics from the International Osteoporosis Foundation, one in three women worldwide over the age of 50 years and one in five men will experience osteoporotic fractures in their lifetime.

Early screening for osteoporosis with dual-energy X-ray absorptiometry (DXA) to assess bone mineral density is an important tool for timely treatment that can reduce the risk of fractures. However, the low availability of the scanners and the relatively high cost has limited its use for screening and posttreatment follow-up.

In contrast, standard X-ray is widely available and used frequently for various clinical indications in daily practice. Despite these attributes, it has been relatively underutilized in the management of osteoporosis because diagnosing osteoporosis using only X-rays is challenging, even for experienced radiologists.

“For patients with hip pain, radiologists often evaluate only image findings that may cause pain, such as fractures, osteonecrosis, and osteoarthritis,” says study author Hee-Dong Chae, MD, from the department of radiology at Seoul National University Hospital in Korea. “Although X-ray images contain more information about the healthiness of the patient’s bones and muscles, this information is often overlooked or considered less important.”

Chae and colleagues developed a model that can automatically diagnose osteoporosis from hip X-rays. The method combines radiomics, a series of image processing and analysis methods to obtain information from the image, with deep learning, an advanced type of AI. Deep learning can be trained to find patterns in images associated with disease.

The researchers developed the deep-radiomics model using almost 5,000 hip X-rays from 4,308 patients obtained over more than 10 years. They developed the models with a variety of deep, clinical, and texture features and then tested them externally on 444 hip X-rays from another institution. The deep-radiomics model with deep, clinical, and texture features was able to diagnose osteoporosis on hip X-rays with superior diagnostic performance than the models using either texture or deep features alone, enabling opportunistic diagnosis of osteoporosis.

“Our study shows that opportunistic detection of osteoporosis using these X-ray images is advantageous, and our model can serve as a triage tool recommending DXA in patients with highly suspected osteoporosis,” Chae says.

The researchers are planning a larger study that combines the clinical information from Korea’s National Health Insurance Service database with the imaging data of the Seoul University Hospital.

— Source: RSNA