Optimization of a calibration phantom for quantitative radiography

Objective: Dual energy radiography (DER) makes it possible to obtain separate images for soft-tissue and bony structures (tissue maps) based on the acquisition of two radiographs at different source peak-kilovoltage values. Current DER studies are based on the weighted subtraction method, which requires either manual tuning or the use of precomputed tables, or on decomposition methods, which make use of a calibration to model soft-tissue and bone components. In this study, we examined in depth the optimum method to perform this calibration. Methods: We used simulations to optimize the calibration protocol and evaluated the effect of the material and size of a calibration phantom composed of two wedges and its positioning in the system. Evaluated materials were water, PMMA and A-150 as soft-tissue equivalent, and Teflon, B-100 and aluminum as bone equivalent, with sizes from 5 to 30 cm. Each material combination was compared with an ideal phantom composed of soft tissue and bone. Our simulation results enabled us to propose four designs that were tested with the NOVA FA X-ray system with a realistic thorax phantom. Results: Calibration based on a very simple and inexpensive phantom with no strict requirements in its placement results in appropriate separation of the spine (a common focus in densitometry studies) and the identification of nodules as small as 6 mm, which have been reported to have a low rate of detection in radiography. Conclusion: The proposed method is completely automatic, avoiding the need for a radiology technician with expert knowledge of the protocol, as is the case in densitometry exams. The method provides real mass thickness values, enabling quantitative planar studies instead of relative comparisons. (c) 2020 American Association of Physicists in Medicine

Automated summary

This study optimized the calibration protocol through simulations and evaluated the placement effects of the calibration phantom's materials and sizes in the system. The proposed method, based on a simple and cost-effective phantom, provides appropriate separation of the spine and identification of nodules as small as 6 mm in size.