Computer aided detection of lumbar spine landmarks for ultrasound guided lumbar punctures and epidurals

Palpation-based techniques are the most common approaches for facilitating lumbar spine bedside procedures such as lumbar puncture and spinal and/or epidural placement. However, the efficacy of these approaches is diminished in patients whose surface landmarks are difficult to palpate due to obesity or spinal deformities. In this work, we present the imaging results of a handheld ultrasound system that was developed to image the lumbar spine and facilitate lumbar punctures in patients whose landmarks are not easily palpable. The system features two novel technologies that permit improved imaging of the lumbar spine: 1) a bone-specific ultrasound imaging mode that enhances the contrast of bone surfaces relative to nearby soft-tissue structures and 2) a real-time computer-aided detection (CAD) algorithm that automatically identifies the spine midline, spinous processes, interlaminar spaces, and the depths to key landmarks. The imaging performance of these algorithms was assessed in 80 volunteers (BMI between 18.5 and 48 kg/m
²
) and was benchmarked against image interpretations provided by three radiologists. The CAD algorithm detected the interlaminar space with a sensitivity of 94.2% and a specificity of 85.5% and the spine midline with a sensitivity of 93.9% and a specificity of 91.3%. Overall, the results of this study demonstrate the feasibility of using a lumbar spine CAD algorithm and bone enhancement imaging technology to assist with image interpretation of ultrasound images of the lumbar spine.