Fig 1.
QRM phantom variations and bovine femur biological sample.
The image of a 100-mm-diameter QRM phantom without any metal inserts for reference data (a), with an aluminium (AL) insert (b), and with a stainless steel insert (S)(c). The phantom has two calibration rods of HA (201.4 and 406.9 mg/cm3), and three rods of iodine (4.83, 9.66, and 14.56 mg/cm3) along with adipose and water. The biological sample is a bovine femur with a stainless steel volar fixation plate and cortical screw (d).
Fig 2.
Spectral images for three datasets.
Illustration of the phantom image (a) and its CT images acquired at 118 kVp and 80 μA across five energy bins (b,c,d,e,f). The phantom without any metal insert (a1), with aluminum (AL) (a2), and with stainless steel (a3), has two calibration rods of HA (201.4 and 406.9 mg/cm3), three rods of iodine (4.83, 9.66, and 14.56 mg/cm3) along with adipose and water rods each 20-mm in diameter. Energy images of metal inserts reveal the metal artefacts reduction in higher energy bins. The grayscale bar represents the Hounsfield units (HU) range from -1000 to 2000.
Fig 3.
The top row shows the Voxel-wise spectral response of HA (201.4 and 406.9 mg/cm3) without metal insert (a), with aluminium (AL) (b), and with stainless steel (c). The bottom row shows the spectral response of iodine (4.83, 9.66, and 14.56 mg/cm3) without metal insert (d), with aluminium (AL) (e), and with stainless steel (f). The horizontal line inside each box represents the median value (50% percentile of the data). The top and bottom boundaries of the box indicate the lower and upper quartile values of HU representing the 25% and 75% percentiles respectively.
Fig 4.
Linearity response of x-ray attenuation.
Linear regression response of x-ray attenuation (HU) in each energy bin as a function of known concentrations of hydroxyapatite (HA) (a–c) and iodine (d–f) in the absence and presence of aluminium (AL) and steel. The standard error ranges between 1 to 3 HU.
Table 1.
The linear regression (R2) and Root Mean Square Error (RMSE) values for Hydroxyapatite (HA) and iodine across each dataset.
Fig 5.
Pairwise assessment of signal-to-noise-ratio (SNR).
Bland-Altman plots in the presence of aluminium (AL) and steel inserts as compared to reference data (without any metal insert) across four ROI (yellow circles) in each category: (a) inside the materials, (b) the immediate vicinity of metal, (c) and the outer proximity. Bland-Altman plots show the difference between (b/w) SNR (SNR without metal—SNR with metal) as a function of the mean SNR for each category (a,b,c). The solid blue and black lines represent the SNR difference for aluminium (AL) and steel, respectively. The corresponding dashed lines represent their upper and lower limits (confidence limits ± 1.96 × standard deviation.).
Fig 6.
Material density images in the absence and presence of metal inserts.
Energy images (top row; grayscale shows HU) and density images (second and third row; scale bar shows mg/cm3). Material decomposition eliminates the effect of aluminium (AL) and steel from the HA density profile (e, f). However, aluminium (AL) is misidentified as iodine (g) compared to steel (h).
Fig 7.
Voxel-wise comparison of measured concentrations as compared to known concentration.
Voxel-wise comparison showing measured concentrations with known concentrations of hydroxyapatite (HA) (a) and Iodine (b). Dotted vertical grid lines separate datasets between those without metal inserts (blue), with AL (green), and with steel (yellow). In the HA density profile (a), the misidentification of iodine concentration and in the iodine density profile (b), the misidentification of HA is separated by solid vertical lines. Whiskers and boxes are described in the same manner as in Fig 3.
Table 2.
Characterization of material identification metrics (Sensitivity, specificity, Area under the curve (AUC 3), accuracy, Negative predictive value (NPV), and Positive predictive value (PPV)) and quantification analysis (root mean square error (RMSE)) for Hydroxyapatite (HA) and iodine, both with and without the presence of aluminium (AL) and steel.
Fig 8.
Illustration of biological sample.
(a) Image of a biological sample (bovine femur bone) with a precisely positioned volar plate (wrist fixation plate) implant, along with a cortical bone screw made from stainless steel; (b) Sagittal view of a bone with an implant for the lowest energy bin (7–40 keV), (c) and higher energy bin (79–118 keV) with significantly reduced artefacts. (d) Normalized line profile for the lowest and highest energy bin showing a noticeable reduction in the artefacts. (e) The material density image demonstrates the differentiation of tissues: bone (white), fat (yellow), and soft tissue (orange-red), (f) accompanied by a combined image of material and energy, highlighting the implant in green. (g) Additional zoom image focuses on the bone density showing the capability of the SPCCT in detailing the structure of the bone.
Fig 9.
Three-dimensional rendering of the bone sample using Mars visualization software. Green color showing the metal implant as volar fixation plate, and long stainless steel cortical screw.