"As 3D printing programs grow, it is important to ensure the models we produce are consistently accurate," Dr. Anish Ghodadra from the University of Pittsburgh Medical Center told AuntMinnie.com.
Most 3D printers generate highly accurate 3D-printed models based on the dimensions of virtual 3D models, he noted. However, the process of segmenting medical images to create these virtual 3D models in the first place is subject to considerable variation, depending on the experience and methodology of individual clinicians.
To prevent excessive variation in the production of 3D-printed models, Ghodadra and colleagues used the Gage repeatability and reproducibility statistical tool to assess the accuracy of virtual 3D models generated by different image segmenters.
Several segmenters with at least one year of experience in 3D printing -- including a radiologist, technologist, and biomedical engineer -- individually performed image segmentations of the femur in the hip CT scans of five different patients.
Statistical analysis of the resulting virtual 3D models revealed a 16.5% variability among the models, indicating that the overall accuracy of the models was somewhat acceptable but not ideal for high-quality 3D printing.
"Using a Gage repeatability and reproducibility analysis allows us to assess staff members' reliability and repeatability in producing high-quality, accurate image segmentations and ensure that they can create consistent, precise segmentations before we have them work on actual cases for 3D printing," Ghodadra said.