[Objective] Spinal and spinal cord diseases, such as spinal canal stenosis, cause numbness, pain, and paralysis due to the compression of the nervous system within the spinal canal, directly impairing activities of daily living and quality of life. Diagnostic tests are essential for understanding these conditions. However, such tests require large equipment, such as X-ray CT and MRI, which impose burdens on both patients and physicians due to radiation exposure and long examination times. To address these issues, research has been conducted to develop diagnostic methods using non-invasive and convenient ultrasound devices. Previous studies focused on the pulsation of the dural sac and developed speckle tracking tailored for the dural sac by combining the SAD-IPI-KLT method with a learning model called SpeckleDNet. However, the accuracy of this tracking remains unclear. Therefore, this study aims to verify the accuracy of speckle tracking developed for dural sac pulsation analysis.
[Methods] Using the ultrasound simulation program Field II, speckle images simulating outputs from ultrasound diagnostic devices were generated, and videos showing the displacement of these generated images were created. Two types of contrast images were prepared, and two types of videos were created using them. Speckle tracking was applied to the created videos, and coordinates were extracted from the tracking results. These coordinates were then compared with the ground truth displacements, and accuracy was evaluated and expressed as the area matching rate.
[Results] In this study, the speckle tracking developed for the dural sac achieved an area matching rate of 81% or higher for 100 displacements corresponding to an actual displacement of 0.03 mm in both types of contrast videos. Additionally, it was confirmed that higher accuracy was maintained when the PPR value was low. It was also observed that changes in displacement magnitude and direction affected the accuracy.
[Discussion] Even when a learning model was added to the tracking method tailored for the dural sac, the speckle tracking maintained a high level of accuracy, achieving an area matching rate of 81% for displacements of 0.03 mm. The observation that higher accuracy was maintained when the PPR value was low suggests that speckle tracking for the dural sac is more suitable for low-contrast images. While large errors in dural sac tracking have been previously observed, this study suggests that these errors may have been caused by differences in accuracy depending on displacement magnitude and direction.