| Summary: | Low back pain is one of the most frequent medical problems in the western world and its consequent cost is enormous. However, despite the high occurrence of low back pain, diagnosis ofthe causes is still a major problem. Research has indicated that low back pain is often related to mechanical disorders of the spinal or holding elements. Therefore, it could be very helpful for clinical diagnosis to study the motion of lumbar spine in order to determine where abnormal motion exists and hence any sources of mechanical instability. Digital videofluoroscopy (DVF) is currently the only practical medical imaging technique to obtain real-time, continuous motion sequences of the lumbar spine. However, DVF images suffer from th~Jresence of noise, poor contrast and adjacent structures near the vertebrae due to the low radiation dosage. Recently, wavelet-based approaches have been applied in edge detection to acquire multiscale gradient images. In multi-scale detection, the edges are more accurately located with low scales but some false edges are produced; with large scales, fewer false edges are identified but traded against a reduced accuracy in the edge location. This project presents a scale multiplication in the identification of spinal vertebrae as a basis for quantifying kinematics. The scale multiplication is defined as the product of the response of the detection filter at different scales so that it combines the advantages of the low and large scales. Once edges are determined as the local maxima in scale multiplication, more robust detection results are obtained after thresholding. The threshold values are decided by the standard. deviation of the noise in the images. With the edge information of the lumbar spine vertebrae, biomechanical parameters, such as rotation and intervertebral angles can be measured via manual landmarking. Another development of this project is the automated tracking technique by using the Generalized Hough transform algorithm. With the mathematical description of the vertebral edges detected by the wavelet scale multiplication method; the vertebral movements in spine motion are tracked. This approach is applied to the DVF image sequences from a calibration model and from ten human subjects to demonstrate its reliability and robustness. This.research would benefit the diagnosis of low back pain and providea platform for the further development of other clinical analysis, such as the cervical spine study.
|
|---|
