Phantom Calibration's Impact on Measuring Radiodensity Changes: Implications for Evaluating Osteosarcoma Response

Improve radiodensity measurements through a controlled, cadaveric study with the primary hypothesis that using high-resolution digital radiographs and image calibration with an aluminum step phantom would be more accurate than in the absence of a phantom. Ultimately, the goal is to improve evaluation of osteosarcoma response to chemotherapy prior to resection using radiographs.

Five fresh-frozen cadaveric lower extremities were imaged with computed tomography (CT) and radiography (anteroposterior (AP) and lateral) with varying mAs/kVp before and after distal femoral instrumentation with 4 cm² rods of known radiodensities, simulating tumors of varying density, and a commercial radiographic calibration step phantom. Using Materials Laboratory for Digital Imaging and Communications in Medicine (DICOM) image analysis, grayscale measurements were obtained for: rods in/out of bone, normal bone proximal to instrumentation, air, and all 16 calibration phantom steps. Radiodensity subtraction accuracy was compared for calibration with all phantom steps, +/- bone or air subtraction. Accuracy was compared using Student T-test and one-way analysis of variance with Tukey HSD post-hoc test.

Mean accuracy for Cohort 1 (no calibration) (.12, standard deviation [SD] 0.11, N = 120) was better than that of Cohorts 2-4 (calibration) (0.39, SD 0.54-0.60, N = 1920) (p < 0.001, t = 5.4689). Lateral knee radiographs, specific imager settings (4 mAs-70 kVP, 8 mAs-70 kVp, and 4 mAs-80 kV), and variations in bone density/BMI improved accuracy (p < .001).

When measuring radiodensity changes using high-quality, digital images with identical settings, image calibration worsened accuracy, likely by introducing an additional source of potential error. Further work is required to determine if calibration will improve accuracy for images obtained with varying technique.