Adaptive dual-energy algorithm based on pre-calibrated weighting factors for chest radiography

Medical imaging often requires distinguishing between soft tissues and bones to improve diagnostic accuracy. Prior research has shown that dual-energy (DE) imaging can help separate materials based on their x-ray attenuation properties. However, conventional DE methods face challenges in balancing material selection and noise suppression. This gap motivated the development of adaptive algorithms that adjust weighting factors spatially. Existing techniques typically use uniform weighting, which may not account for spatial variations in image composition. No prior work had resolved how to dynamically calibrate weighting factors for different regions. The need for real-time, high-quality DE imaging in clinical settings remains unmet. This paper contributes by proposing a novel DE algorithm that adapts to spatial variations in tissue composition.

The aim of this study was to develop a dual-energy algorithm that improves material selection and noise suppression in chest radiography. The specific problem addressed is the limitation of conventional DE imaging, which uses static weighting factors and may not adapt to spatial variations in tissue composition. The motivation stems from the need for real-time, high-quality imaging in clinical environments. The proposed method uses pre-calibrated weighting factors derived from calibration phantoms. These factors are spatially adapted to the image content, allowing for better material separation and noise reduction. The study tested the algorithm's performance in different orientations and phantom sizes. The goal was to evaluate whether adaptive weighting could enhance image quality compared to conventional methods. The significance lies in the potential for improved diagnostic accuracy through better material differentiation.

The study employed calibration step-phantoms containing overlapping slabs of solid water and bone. These phantoms were used to determine material selection and noise suppression weighting factors. Material selection weighting factors were calculated by finding the zero of contrast-to-noise ratio (CNR) between overlapping and non-overlapping regions. Noise suppression weighting factors were determined by maximizing signal-to-noise ratio in overlapping regions. The pre-calibrated factors were fitted to low and high energy radiographs of Rando phantoms. These phantoms were used to generate maps of weighting factors for material selection and noise suppression. The DE algorithm was combined with anti-correlated noise reduction (ACNR) to produce final images. Three implementations were tested: different phantom sizes and orientations (oblique and anterior-posterior). Image quality was evaluated using CNR, contrast, and noise values in regions of interest.

The adaptive DE (aDE) algorithm showed improved image quality compared to simple log subtraction (SLS) in all tested orientations. Contrast-to-noise ratio (CNR) increased for both soft-tissue and bone images, indicating better material separation. The aDE algorithm provided better contrast than SLS, with no significant loss in noise levels. Implementation of the aACNR algorithm further reduced image noise but slightly decreased contrast compared to aDE alone. The aACNR method outperformed the uniform ACNR approach in terms of contrast preservation. Soft-tissue and bone-only images generated with aDE and aACNR showed higher diagnostic value. The performance was consistent across different phantom sizes and orientations. These results suggest that adaptive weighting improves both material selection and noise suppression in DE imaging.

The authors concluded that the novel adaptive DE algorithm improves material selection and noise suppression compared to conventional methods. The aDE algorithm demonstrated higher contrast and better CNR in soft-tissue and bone images. The aACNR method further reduced noise but at the cost of some contrast reduction. These findings suggest that adaptive weighting can enhance image quality in DE radiography. The proposed method can be implemented in clinical settings for real-time imaging. The results support the use of spatially varying weighting factors for improved diagnostic accuracy. The study does not propose new clinical applications but validates the algorithm's performance in controlled settings. The significance lies in the algorithm's potential to improve diagnostic imaging through better material differentiation.