Abstract
Introduction
Morphological parameters are used to describe curve characters in AIS like curve location, curve magnitude, stiffness etc. Like all other morphological metrics the accuracy is more when digital imaging, archiving and extraction of features is used rather than manual measurements. The content Based Image Retrieval system is anew software that allows rapid, accurate documentation of AIS images and their retrieval by visual content.
Classification systems and their shortcomings
Traditional classifications only looked at curve location (Ponsetti/Friedman); this was enhanced to add curve flexibility (to include or exclude secondary curves in fusion) (PUMC, King/Moe etc). Newer classifications like the Lenke have added sagittal profile into the decision making equation. From 5 basic curve types the subtypes have increased to 42 potential curve patterns by the addition of one parameter!! In future as we understand the 3-D geometry of these curves better we may want to add more measureable items (like degree of rotation) and by adding one term the subtypes would be 128!!! This suggests that we need to have a simple easy to remember way of classifying or eliminate classifications altogether.
Experimental evidence
Several experiments were conducted with the new CBIR software which showed that similar images of scoliosis cases could be retrieved without resorting to a classification scheme. Even surgical planning can be made by downloading all similar cases operated before. The variability can be set to any level of precision desired.
Significance
In future we may eliminate classifications to decide on curve types and for surgical planning and recall from a large multicentre database similar curves and their surgical plan.
