Investigation of Extended HU Mode Using a Semi-Custom CT Phantom
Abstract
Purpose
Metals in computed tomography (CT) scans often saturate image pixel values, making low- and high-density metals indistinguishable. The purpose of this work was to test the effectiveness of the extended HU mode on a GE Discovery RT16 CT scanner to distinguish between metals in the range of 2.7-18.1 g/cm3.
Methods
Extended HU CT scans (single helical slice, 300 mA, 120 kVp) were taken of an acrylic phantom containing single cylindrical (9- and 30-mm diameter) inserts of aluminum (2.7 g/cm3), titanium (4.5 g/cm3), stainless steel (7.9 g/cm3), brass (8.5 g/cm3), or tungsten (18.1 g/cm3). Profiles across each insert were acquired to quantify HU changes with material size and density. Metal regions in the images (ROI) showed cupping artifacts that suppress the central mean HU value within the ROI. Alternative ROI analysis methods were compared to the mean including: maximum pixel value and edge painting (averaging the brightest pixels along the edge of the insert). HU metrics were plotted as a function of mass density.
Results
The mean pixel value method observed for 9 mm (30 mm) inserts values of 1910 (1966), 7203 (5806), 5478 (7433), — (7442), and 17370 (7922) HU for aluminum, titanium, stainless steel, brass, and tungsten, respectively. It did not distinguish titanium from stainless steel for 9-mm inserts or titanium, stainless, brass, and tungsten for 30-mm inserts. The maximum pixel value method better distinguished between metal types, observing 2030 (2105), 7732 (7106), 8771 (10688), — (11267), 21417 (12028) HU for aluminum, titanium, stainless steel, brass, and tungsten, respectively. It distinguishes all metal types for 9-mm inserts but fails to clearly distinguish brass from tungsten for 30-mm inserts.
Conclusion
The extended HU range distinguished metal types independently of insert size for objects ≤ 4.5 g/cm3, above which the HU value of the metal becomes size-dependent.