Dataset for "Smartphone scanning is a reliable and accurate alternative to contemporary residual limb measurement techniques"

Data collection method:

Ten residual limbs across seven participants were scanned using a criterion Artec EVA scanner, and an Apple iPhone 12 using a range of scanning applications. These scanning applications were Polycam, Luma, and Meshroom. Both Polycam and Luma also had web-based equivalents, for which a separate set of photos were created to be uploaded to their servers following scanning. Each scan was conducted three times so the repeatability of each scanning application could be evaluated. Each participant was scanned at least ten minutes following the doffing of their prosthesis, to allow swelling to plateau prior to commencement of scanning. A set of fiducial markers were stuck to the residual limb to aid tracking and alignment, as well as a reference object to aid in correct scaling of the limb. The scans were conducted in random order, and the participant was asked to remain as motionless as possible whilst each scan took place. On average, scans took 2-4 minutes each, and a total of 18 scans per participant across repetitions and applications were conducted. Between 80-150 frames were captured for each of the smartphone scans, the range of which was proven to have a negligible impact on the outcomes of the scans.

Data processing and preparation activities:

Collected scans were post-processed in Blender (v 3.3.1), to remove irrelevant data such as the participants torso, to remove miscellaneous/broken geometry, and scale meshes. Two distinct scaling methods were used, splitting the dataset into two halves. These were the Naive set, in which the meshes were scaled according solely to the reference geometry captured in the scan, and the Optimal set, in which the meshes were scaled according to the Artec criterion scans, resulting in a more accurate scaling operation. Once complete, the scans were compiled and aligned in Artec Studio 12, and split into ten sections along their lengths. The data collected from these sections included perimeter, cross-sectional area, and total volume. A separate measurement operation was performed to gather RMSE data, for which a search distance of 15 mm was used.

Technical details and requirements:

To capture the mobile phone scans, Polycam, Luma and Meshroom were used. Polycam (3.2.15) and Luma (0.9.9) have both mobile application and web versions, so both were used to evaluate any potential differences, whereas Meshroom (2023.3.0) was solely a desktop application. The desktop application Blender (3.3.1) was used to post-process all of the scans.

Additional information:

All data has been sorted into Excel spreadsheets, with each sheet detailing the repeated scans for a single limb, for each section along the length of the limb. Each limb has two sheets, which are the naive (N) and optiomal (O) datasets.