Geometric validation of a computer simulator used in radiography education

Objectives: The radiographical process of projection of a complex human form onto a two-dimensional image plane gives rise to distortions and magnifications. It is important that any simulation used for educational purposes should correctly reproduce these. Images generated using a commercially available computer simulation widely used in radiography education (ProjectionVRTM) were tested for geometric accuracy of projection in all planes. Methods: An anthropomorphic skull phantom was imaged using standard projection radiography techniques and also scanned using axial CT acquisition. The data from the CT was then loaded into the simulator and the same projection radiography techniques simulated. Bony points were identified on both the real radiographs and the digitally reconstructed radiographs (DRRs). Measurements sensitive to rotation and magnification were chosen to check for rotation and distortion errors. Results: The real radiographs and the DRRs were compared by four experienced observers and measurements taken between the identified bony points on each of the images obtained. Analysis of the mean observations shows that the measurement from the DRR matches the real radiograph +1.5 mm/−1.5 mm. The Bland Altman bias was 0.55 (1.26 STD), with 95% limits of agreement 3.01 to −1.91. Conclusions: Agreement between the empirical measurements is within the reported error of cephalometric analysis in all three anatomical planes. The image appearances of both the real radiographs and DRRs compared favourably. Advances in knowledge: The commercial computer simulator under test (ProjectionVRTM) was able to faithfully recreate the image appearances of real radiography techniques, including magnification and distortion. Students using this simulation for training will obtain feedback likely to be useful when lessons are applied to real-world situations.

DRRGenerator: A Three-dimensional Slicer Extension for the Rapid and Easy Development of Digitally Reconstructed Radiographs

As the interest in image-guided medical interventions has increased, so too has the necessity for open-source software tools to provide the required capabilities without exorbitant costs. A common issue encountered in these procedures is the need to compare computed tomography (CT) data with X-ray data, for example, to compare pre-operative CT imaging with intraoperative X-rays. A software approach to solve this dilemma is the production of digitally reconstructed radiographs (DRRs) which computationally simulate an X-ray-type image from CT data. The resultant image can be easily compared to an X-ray image and can provide valuable clinical information, such as small anatomical changes that have occurred between the pre-operative and operative imaging (i.e., vertebral positioning). To provide an easy way for clinicians to make their own DRRs, we propose DRR generator, a customizable extension for the open-source medical imaging application three-dimensional (3D) Slicer. DRR generator provides rapid computation of DRRs through a highly customizable user interface. This extension provides end-users a free, open-source, and reliable way of generating DRRs. This program is integrated within 3D Slicer and thus can utilize its powerful imaging tools to provide a comprehensive segmentation and registration application for clinicians and researchers. DRR generator is available for download through 3D Slicer’s in-app extension manager and requires no additional software.

Estimation of Pelvic Sagital Inclanation from Anteroposterior Radiograph Using Convolutional Neural Networks: Proof-of-Concept Study

Alignment of the bones in standing position provides useful information in surgical planning. In total hip arthroplasty (THA), pelvic sagittal inclination (PSI) angle in the standing position is an important factor in planning of cup alignment [1] and has been estimated mainly from radiographs. Previous methods for PSI estimation [2], [3] used a patient-specific CT to create digitally reconstructed radiographs (DRRs) and compare them with the radiograph to estimate relative position between the pelvis and the x-ray detector. In this study, we developed a method that estimates PSI angle from a single anteroposterior radiograph using two convolutional neural networks (CNNs) without requiring the patient-specific CT, which reduces radiation exposure of the patient and opens up the possibility of application in a larger number of hospitals where CT is not acquired in a routine protocol.

Influence of Foot Position on the Measurement of First Metatarsal Pronation Using the First Metatarsal Axial Radiographs

Category: Other Introduction/Purpose: In hallux valgus (HV) surgery, the importance of correcting first metatarsal (1MT) pronation, as well as correcting adduction, has been increasingly recognized. A 1MT axial radiograph is a simple method to quantify 1MT pronation (see figure). However, this view does not provide the exact axial projection, and 1MT is angled on the image. Furthermore, the 1MT angle is dependent on the foot position and alignment. Therefore the measured pronation angle on the radiograph may not be the same as the “true” pronation along the 1MT long axis. The purposes of this study were to 1) quantify the difference between the measured 1MT pronation on the axial radiograph and the true pronation angle, and 2) determine the influence of foot position on the measurement. Methods: CT images of 10 feet from HV patients (HV group; age, 58 years; HV angle, 44°) and 10 feet from those without HV (non-HV group; age, 47years; HV angle, 12°) were obtained. Digitally reconstructed radiographs of the 1MT were generated from the CT images in a three-dimensional virtual space (see figure). 1MT was pronated along the long axis (true pronation angle). Then, images with different plantarflexion (25° to 35° in 5° increments) and adduction (-10° to 10°) angles were created. This procedure was repeated for different pronation angles (-10° to 30°), and 135 simulated 1MT axial radiographs were obtained from each bone. We measured the 1MT pronation angles in all images. Differences in the measured pronation angle and true pronation angle were calculated for each group. Correlations between the measurement difference and 1MT plantarflexion/adduction angles were assessed using the Spearman correlation coefficient. Results: The mean measurement differences were 0.7° and 1.0° for the HV group and non-HV group, respectively. The standard errors of the measurement were 0.5° and 0.6° for the HV group and non-HV group, respectively. There was no significant correlation between the measurement difference and plantarflexion angle (P = 0.92 and P = 0.92 for the HV group and non-HV group, respectively), nor between the measurement difference and adduction angle (P = 0.82 and P = 0.74). Conclusion: The measurement differences were low in both HV and non-HV feet, indicating that the 1MT pronation angle measured on the axial radiograph represented the true pronation angle along the long axis of 1MT. The measurement differences were consistent regardless of plantarflexion and adduction angles. Therefore, a variation of 1MT angle on the image, which can be caused by the misalignment of foot position while taking the radiograph and difference in foot alignment (such as flatfeet and cavus feet), did not affect the measurement. 1MT axial radiograph could be used as a valid and robust method to quantify 1MT pronation.