The automized fracture edge detection and generation of three-dimensional fracture probability heat maps. (December 2022)

• Record Type: Journal Article
• Authors: Kahmann, Stephanie L.; Rausch, Valentin; Plümer, Jonathan; Müller, Lars P.; Pieper, Martin; Wegmann, Kilian
• Is Part Of: Medical engineering & physics. Volume 110(2022)
• Publication Date: 2022-12
• Subjects: Imaging -- Morphing -- Probability distribution mapping -- Shoulder -- Fracture classification; Biomedical engineering -- Periodicals; Biomedical Engineering -- Periodicals; Physics -- Periodicals; Génie biomédical -- Périodiques; Biomedical engineering; Electronic journals; Periodicals; 610.28
• DOI: 10.1016/j.medengphy.2022.103913
• Languages: English
• ISSNs: 1350-4533

Abstract

Highlights: A novel approach in retrospective statistical analysis of fracture patterns. Automation of parts of a method that formerly required high manual effort. Finds fracture edges precisely and quickly in simple geometries. Also applicable to complex shapes like acetabulum with higher time effort. Fracture probability heat maps show clinically relevant results. Abstract: With proven impact of statistical fracture analysis on fracture classifications, it is desirable to minimize the manual work and to maximize repeatability of this approach. We address this with an algorithm that reduces the manual effort to segmentation, fragment identification and reduction. The fracture edge detection and heat map generation are performed automatically. With the same input, the algorithm always delivers the same output. The tool transforms one intact template consecutively onto each fractured specimen by linear least square optimization, detects the fragment edges in the template and then superimposes them to generate a fracture probability heat map. We hypothesized that the algorithm runs faster than the manual evaluation and with low (< 5 mm) deviation. We tested the hypothesis in 10 fractured proximal humeri and found that it performs with good accuracy (2.5 mm ± 2.4 mm averaged Euclidean distance) and speed (23 times faster). When applied to a distal humerus, a tibia plateau, and a scaphoid fracture, the run times were low (1–2 min), and the detected edges correct by visual judgement. In the geometrically complex acetabulum, at a run time of 78 min some outliers were considered acceptable. An automatically generated fracture probability heat map based on 50 proximal humerus fractures matches the areas of high risk of fracture reported in medical literature. Such automation of the fracture analysis method is advantageous and could be extended to reduce the manual effort even further.