BoneMetrics (US)

BoneMetrics (US) is a fully automated radiological image processing software that uses machine learning to measure Cobb angles on frontal spine radiographs in patients aged 4 years and older with suspected or present spinal deformities such as scoliosis. The software receives DICOM images automatically, processes them without user input, and returns annotated result images with angle measurements overlaid, intended as a concurrent reading aid for radiologists.

BoneMetrics uses convolutional neural networks for image classification and vertebral landmark detection, followed by classical methods to compute Cobb angles. Unlike the predicate device (IB Lab LAMA), it does not perform segmentation, compute auxiliary points, or calculate distances—only angle measurements. The processing architecture is simplified to spine-specific measurements rather than the multi-limb capabilities of the predicate.

Not stated in this summary.

Both devices are fully-automated radiological image processing software using convolutional neural networks and classical computational methods to generate automated anatomical measurements on radiographs for use by trained radiologists. Although BoneMetrics targets spine/Cobb angles (versus the predicate's limb-length and knee alignment measurements) and younger patients (4+ years versus 22+ years), these differences do not raise new safety or effectiveness questions because the fundamental technology and intended use—providing accurate, reproducible automated measurements as a reading aid—remain identical. The same mitigation controls apply: dedicated algorithm training for the specific indication and population, plus clinical performance testing demonstrating Mean Absolute Errors of 2.56°–2.78° across all patient groups, meeting the pre-specified acceptance criteria.