Evaluation of Squared Signal Processing for Enhancing Low-Contrast Resolution in CT Imaging

Purpose In hyperacute stroke, early CT signs are key indicators for diagnosis. However, the earlier the imaging is performed after the onset, the more subtle the blurring of the corticomedullary and insular ribbon boundaries, making visual identification extremely difficult. This study investigated whether applying a nonlinear square transformation to computed tomography (CT) images could improve low-contrast resolution. Methods Using the low-contrast resolution evaluation phantom Catphan 700, four types of images were generated from the acquired CT data: a standard CT image (Normal), an image processed with a subtraction of 5 Hounsfield Units (HU) from all pixels followed by squaring (Square-5), a squared image without offset (Square), and an image processed with an addition of 5 HU to all pixels followed by squaring (Square + 5). Physical evaluations were performed using a task transfer function (TTF), noise power spectrum (NPS), and contrast-to-noise ratio (CNR _LO ). Visual assessment was also conducted. Results No significant differences were observed in the TTF or low-contrast object-specific CNR _LO . The NPS was markedly increased in the squared images compared with that in the original CT images. However, the influence of noise was limited by tone mapping, resulting in an improvement in low-contrast resolution. Conclusion The nonlinear square transformation improved low-contrast resolution and demonstrated potential effectiveness as a diagnostic aid for acute cerebral infarction.