registration of 50 3D scans in order to create an average scan

vmatyas · June 5, 2018, 4:47pm

Hi All,

I will have 6x50 3D scans of feet. I need to register the 50 scans in each group so the corresponding vertices align so I can calculate average foot scan for each group.
Any ideas how I can do non-rigid transformations to achieve this. I know this has been done before in the below paper but the author is not writing back to me.

ncbi.nlm.nih.gov

Comprehensive surface-based morphometry reveals the association of fracture risk and bone geometry.

D Wang, L Shi, JF Griffith, L Qin, DT Yew and CM Riggs, Journal of orthopaedic research : official publication of the Orthopaedic Research Society, Aug 2012

Surface-based morphometry method is advantageous in its objectivity and increased capability in detecting focal morphological changes, but has not been applied in bone-related research. Orthopedics research in human has confirmed the association of the bone geometry in proximal femur and its fracture. In this study, surface-based morphometry is used to test the hypothesis that there is relationship between bone geometry and fracture risk of the proximal sesamoid bone (PSB) in forelimbs of Thoroughbred racehorses. The PSB surfaces were extracted from CT images of nonfractured forelegs (i.e., right foreleg in this study) of 6 racehorses with fractures in the contralateral (i.e., left) foreleg, and the right forelegs of 6 matched controls. Significant differences were detected at the abaxial margin of the medial PSB base which was found to be up to 3.5 mm more prominent in the fracture-group compared to the control-group. This study demonstrated a successful application of computational morphometry in bone. The detected anatomical differences may lead to a larger moment arm generated via the medial branch of the suspensory apparatus, increasing pressure on the sesamoid surface, and thus potentially predisposing to fracture. Findings from this pilot study not only increase the likelihood of accurate PSB fracture risk assessment, but also shed light on investigating the influence of sports and exercise on human athletes.

Thank you.

Matyas Varga PhD student

tim-evain · June 5, 2018, 5:14pm

Hello @vmatyas

I can recommand this Kitware blog post. It explains how to create an active shape model (in 2D, but 3D is quite straightforward) which will give exactly what you want (a shape average), given that you can have the delineation of the foot as it works on binary images.

HTH,

Tim

vmatyas · June 6, 2018, 9:05am

Thank you… i think the average is not necessarily the problem, I can make that work in Matlab. I am mainly trying to find a way to effectively do the high number of mesh (I guess “similarity”) registrations (non-rigid) and optimising vertex correspondence between scans.

Thank you.
Mat