Clinical and statistical correlation of various lumbar pathological conditions

Abstract

Current clinical evaluations often rely on static anatomic imaging modalities for diagnosis of mechanical low back pain, which provide anatomic snapshots and a surrogate analysis of a functional disease. Three dimensional in vivo motion is available with the use of digital fluoroscopy, which was used to capture kinematic data of the lumbar spine in order to identify coefficients of motion that may assist the physician in differentiating patient pathology. Forty patients distributed among 4 classes of lumbar degeneration, from healthy to degenerative, underwent CT, MRI, and digital x-ray fluoroscopy. Each patient underwent diagnosis by a neurosurgeon. Fluoroscopy was taken as the patient performed lateral bending (LB), axial rotation (AR) and flexion-extension (FE). Patient specific models were registered with the fluoroscopy images to obtain in vivo kinematic data. Motion coefficients, C(LB), C(AR), C(FE), were calculated as the ratio of in-plane motion to total out-of-plane motion. Range of motion (ROM) was calculated about the axis of motion for each exercise. Inter- and Intra- group statistics were examined for each coefficient and a flexible Bayesian classifier was used to differentiate patients with degeneration. The motion coefficients C(LB) and C(FE) were significantly different (p<0.05) in 4 of 6 group comparisons. In plane motion, ROM(LB), was significantly different in only 1 of 6 group comparisons. The classifier achieved 95% sensitivity and specificity using (C(FE), C(LB), ROM(LB)) as input features, and 40% specificity and 80% sensitivity using ROM variables. The new coefficients were better correlated with patient pathology than ROM measures. The coefficients suggest a relationship between pathology and measured motion which has not been reported previously.