The importance of processing procedures and threshold values in CT scan segmentation of skeletal elements: An example using the immature os coxa. (April 2020)

• Record Type: Journal Article
• Authors: Stock, Michala K.; Garvin, Heather M.; Corron, Louise K.; Hulse, Cortney N.; Cirillo, Laura E.; Klales, Alexandra R.; Colman, Kerri L.; Stull, Kyra E.
• Is Part Of: Forensic science international. Volume 309(2020)
• Publication Date: 2020-04
• Subjects: Forensic anthropology -- Virtual anthropology -- Biological anthropology -- Subadult -- Precision -- Amira™ -- Geomagic®; Medical jurisprudence -- Periodicals; Chemistry, Forensic -- Periodicals; Forensic Medicine -- Periodicals; Médecine légale -- Périodiques; Chimie légale -- Périodiques; Gerechtelijke geneeskunde; Gerechtelijke chemie; Gerechtelijke psychiatrie; Chemistry, Forensic; Medical jurisprudence; Electronic journals; Periodicals; Electronic journals; 614.1
• DOI: 10.1016/j.forsciint.2020.110232
• Languages: English
• ISSNs: 0379-0738

Abstract

Highlights: CT scan processing protocols should remain consistent for accurate results. Slight variation ( e.g ., ∼50 HU) in thresholding does not substantially alter resultant surfaces. Error generated by scan processing is less than allowable measurement error (1–2 mm). Abstract: As the accessibility and utility of virtual databases of skeletal collections continues to grow, the impact that scan processing procedures has on the accuracy of data obtained from virtual databases remains relatively unknown. This study quantifies the intra- and inter-observer error generated from varying computed tomography (CT) scan processing protocols, including re-segmentation, incrementally varying thresholding value, and data collectors' selection of the threshold value on a set of virtual subadult pelves. Four observers segmented the subadult ossa coxarum from postmortem CT scans of the fully-fleshed bodies of eleven individuals of varying ages. Segmentation protocol was set, with the exception of each observer selecting their own thresholding value for each scan. The resulting smoothed pelvic surfaces were then compared using deviation analyses. Root mean square error (RMSE), average distance deviation, and maximum deviation distances demonstrated that thresholding values of ∼50 HU (Hounsfield units) are easily tolerated, the surfaces generated are robust to error, and threshold value selection does not systematically vary with user experience. The importance of consistent methodology during segmentation protocol is highlighted here, especially with regards to consistency in both selected thresholding value as well as smoothing protocol, as these variables can affect subsequent measurements of the resultant surfaces.