STUDY DESIGN: Description of a novel application. OBJECTIVE: Proposal of a new system for content-based image retrieval (CBIR) of scoliosis images for the retrieval of clinically similar images to a query image on the basis of automatically derived features. BACKGROUND: Variability exists in the selection of strategic vertebrae, measurement of Cobb angle, and assignment of a curve type in a classification scheme for scoliosis images. Besides, many classification schemes are in use today, creating ambiguity in selecting a particular classification scheme. METHODS: A rule-based algorithm for strategic vertebrae selection and Cobb angle measurement was developed. A set of automatically derived features necessary for indexing the scoliosis image for CBIR was formulated. A hybrid CBIR system is proposed in which scoliosis features and treatment procedure along with preoperative and postoperative images are integrated. The system was evaluated using 30 curves on standing anteroposterior scoliosis images. The measurement was carried out by 3 independent observers who measured twice at an interval of 7 days. RESULTS: The average difference in Cobb angle was 2.0 degrees with a standard deviation of 3.0 degrees. High values for the correlation coefficient were obtained for both interobserver and intraobserver assessment, proving the repeatability of the system. The CBIR system was validated by query-by-example test method and the system could retrieve correct sets of clinically matching images in the increasing order of distance. CONCLUSIONS: The newly developed system is an accurate and reliable system for search and retrieval of scoliosis images. On querying with a new image, the therapeutic strategy for surgical planning can be assigned on the basis of the outcome assessment of the most similar images retrieved, thereby reducing the importance of any classification scheme.
STUDY DESIGN: Description of a novel application. OBJECTIVE: Proposal of a new system for content-based image retrieval (CBIR) of scoliosis images for the retrieval of clinically similar images to a query image on the basis of automatically derived features. BACKGROUND: Variability exists in the selection of strategic vertebrae, measurement of Cobb angle, and assignment of a curve type in a classification scheme for scoliosis images. Besides, many classification schemes are in use today, creating ambiguity in selecting a particular classification scheme. METHODS: A rule-based algorithm for strategic vertebrae selection and Cobb angle measurement was developed. A set of automatically derived features necessary for indexing the scoliosis image for CBIR was formulated. A hybrid CBIR system is proposed in which scoliosis features and treatment procedure along with preoperative and postoperative images are integrated. The system was evaluated using 30 curves on standing anteroposterior scoliosis images. The measurement was carried out by 3 independent observers who measured twice at an interval of 7 days. RESULTS: The average difference in Cobb angle was 2.0 degrees with a standard deviation of 3.0 degrees. High values for the correlation coefficient were obtained for both interobserver and intraobserver assessment, proving the repeatability of the system. The CBIR system was validated by query-by-example test method and the system could retrieve correct sets of clinically matching images in the increasing order of distance. CONCLUSIONS: The newly developed system is an accurate and reliable system for search and retrieval of scoliosis images. On querying with a new image, the therapeutic strategy for surgical planning can be assigned on the basis of the outcome assessment of the most similar images retrieved, thereby reducing the importance of any classification scheme.
Authors: Dean N Papaliodis; Pierino G Bonanni; Timothy T Roberts; Khalid Hesham; Nicholas Richardson; Robert A Cheney; James P Lawrence; Allen L Carl; William F Lavelle Journal: Int J Spine Surg Date: 2017-06-30