Procrustes analysis—a form of statistical shape analysis—could be used to predict postoperative results in advance of surgery for adolescent idiopathic scoliosis (AIS), a study presented by Adrian Gardner (Royal Orthopaedic Hospital, Birmingham, United Kingdom) at the British Scoliosis Society (BSS) 2019 annual meeting (BSS 2019; 21–22 November; Cardiff, United Kingdom) has concluded.
The study, which was selected as the best clinical paper at the meeting, describes a technique to provide patients and clinicians with a clearer, understandable image of the resulting visual effect of surgery.
Procrustes analysis is a type of statistical analysis used to analyse the distribution of a set of shapes. Gardner suggests that using this technique can provide an answer to the question ‘What will I look like after I have had surgery?’, which he says is a common query from AIS patients prior to treatment. Literature suggests that using pedicle screws the mean correction is approximately 65%, however it is difficult to present this in a way that is comprehensible to patients, Gardner reasoned in the paper.
The Procrustes methodology was applied to AIS surgery by Gardner and colleagues by identifying ten posterior torso ‘landmarks’ on ISIS2 (Integrated Shape Imaging system 2) images of 288 individuals treated for AIS at the pre- and postoperative stage. The landmarks used by Gardner and colleagues included the vertebra prominens, sacrum, shoulders, axilla, waist and the two most prominent points over the posterior torso. The Procrustes analysis finds the mean change in shape from the procedures. One pre-operative image was then compared in turn to every other image using Procrustes techniques to find the best fit. The matching postoperative images for the best fit pair were then also identified and compared using similar techniques—ultimately resulting in a pair of clinical pictures that could be used in the outpatient setting to be able to demonstrate to patients the likely end result of their particular curve and asymmetry pattern.
Commenting on the model, Gardner said: “If successful it will give patients an idea of their external image post-op before they have had the surgery and you will be able to say, ‘this is what you will look like based on our model’.”
However, Gardner acknowledged that a larger database of patients is likely to be needed to improve the effectiveness of the model, while more posterior torso landmarks could be incorporated to further improve the accuracy of the technique—albeit with a need for greater processing power to produce the images.
“Further research is now required to pursue this technique prospectively along with a patient-derived assessment of their own shape. This technique may answer ‘What will I look like after the surgery?’ on a more patient-specific fashion,” Gardner concludes.