Measurement of the Detector response function

 To obtain more reliable results that could facilitate a distinct quantitative analysis, the Detector response function (DRF) was evaluated according to the EUREF Fourth edition Supplements. [2]

A 2-mm thick Al plate was placed on the x-ray tube and the compression paddle was then removed. Furthermore, a 2-mm thick stainless steel was used to protect the x-ray detector, and the x-ray dosimeter (center of measurement) was placed at a distance of 60-mm from the chest wall edge on the centerline of the breast support table (Fig. 1).

Fig. 1: Experimental set up to evaluate the Detector response function.

 The air kerma was measured using the x-ray tube voltage of 29kV (W target / Rh filter), whereas the tube current time products of the x-ray tube were set to 2.0, 4.0, 8.0, 16.0, 32.0, 63.0, 125.0, and 250.0 mAs. The air kerma on the surface of the x-ray detector was calculated using the x-ray dosimeter in accordance with the inverse-square law. Under these circumstances, there was no correction performed to incorporate attenuation from the table, grid, and detector cover in our measurements. 

 Five images were obtained using the same parameters as in the previous measurement.  The image format menu was set to the “QC Test” image format. The pixel value of the RAW image exhibits a linear relationship with dose logarithm. Therefore, images were obtained by converting the pixel value of RAW image into antilogarithm of dose (hereafter referred as “linearized images”).

 The linearized images were then generated by the workstation, and the mean pixel value was measured on ImageJ (NIH, USA) using a 10 mm × 10 mm of the region of interest (ROI) that coincided with the center of the x-ray dosimeter. The DRF was calculated using the mean pixel value and the air kerma of the surface of the x-ray detector. The  DRF demonstrated excellent linearity (Fig. 2).

Fig. 2: Detector response function (pixel value versus detector entrance air kerma).

Geometric sharpness

 Breast compression decreases breast thickness and improves the geometric sharpness because the distance between the breast tissue and the x-ray detector is reduced. This is verified through the application of the System Contrast Transfer Function (SCTF) according to the IEC 61223-3-2/Ed.2:2007. [3]

 A 2-mm thick Al plate was placed on the x-ray tube and a 1-mm diameter steel ball was attached on the periodic resolution pattern. Subsequently, this periodic resolution pattern was accommodated on the compression paddle that was tilted by 45° based on the center line (Fig. 3, 4).

Fig. 3: Set up for the SCTF measurement. The periodic resolution pattern was placed onto the compression paddle in which bar patterns of 4 lp/mm were tilted by 45° based on the center line.

Fig. 4: Arrangement of the predict resolution bar pattern. The 1 mm steel ball was placed on the line parallel to the chest wall edge through the intersection of the center line and the 4 lp/mm bar patterns.

 The height of the compression paddle ranged from 10 mm to 90 mm in increment of 20 mm from the breast support table. Periodic resolution pattern images were then obtained in five consecutive images. The image format menu was set to the “QC Test” image format. The x-ray conditions to obtain the Predict resolution bar pattern were 29 kV (W target / Rh filter) of x-ray tube voltage and 32 mAs of tube current time product. These parameters were obtained without the predict resolution bar pattern on the compression paddle using AEC.

 The linearized images were subsequently generated by the workstation, and an ROI was placed to measure the SCTF: M(f) of 4 lp/mm bar pattern. The shape of the ROI size was in the form of a circumscribed square with a steel ball (Fig. 5).

Fig. 5: Region of interest (ROI) to calculate the SCTF: M(f). The shape of the ROI size was in the form of a circumscribed square with a steel ball. m: average pixel value of ROI. σ: standard deviation of ROI.

The following formula was used to calculate the SCTF: M(f)

Fig. 6: The formulas for calculating the SCTF: M(f).

The SCTF: M(4) increased as the distance (height) between the compression paddle (periodic resolution pattern) and the breast support table decreased (Fig. 7).

Fig. 7: SCTF results at different compression paddle heights (Periodic resolution pattern).

Detectability of low contrast

 Thinner breast thickness facilitated by compression improved the image contrast, enabling the use a lower peak kilovoltage beam that can in turn reduce x-ray scattering due to breast thickness. These effects were verified using a combination of a CDMAM phantom and PMMA. We used CDMAM type 3.4 and the CDMAM analyzer Ver. 1.5.5 software to perform our analysis, and the CDMAM phantom is usually inserted between two 20-mm thick PMMA plates. In this study, the initial setting dictated the positioning of a CDMAM phantom on a 20-mm thick PMMA. Consequently, additional PMMAs with 20 and 40-mm thickness were placed on the CDMAM phantom. This process allowed avoidance of potential geometric blurring while maintaining the optimal height from the breast support table. The image format menu was set to the “QC Test” image format. Exposure conditions are shown in Fig. 8.  The kV and mAs were chosen to match as closely as possible those selected by AEC, using AEC mode at dose setting N, when imaging a 30-mm, 50-mm and 70-mm thickness of PMMA. To obtain these measurements, the CDMAM phantom was replaced by a 10-mm thick PMMA.

Fig. 8: CDMAM and PMMA phantoms set up and exposure conditions. The x-ray exposure was acquired with the compression paddle adjacent to the phantom.

 Following the acquisition of 16 images using each parameter, linearized images were generated by the workstation. These images were then analyzed as a contrast detail curve on the software. This curve demonstrated that the lower left side of the graph exhibited smaller detection ability (low contrast). The thinner PMMA allowed for a distinct improvement in both image quality and detectability (Fig. 9).

Fig. 9: Automated threshold contrast measurements with different PMMA thickness.

Depiction ability of simulated breast tissue

 The depiction ability of simulated breast tissue was evaluated by means of the AutoPIA 3.7.2 software using the TORMAM phantom that had an embedded filament, particle, and circular detail. The simulated breast tissue characteristics were represented as filament diameter (mm), particle size range (μm), and circular detail under nominal contrast (% at 28kV). All characteristics comprised six distinct steps. The acquired image data were automatically detected as the simulated breast tissue on the AutoPIA (Fig. 10). Individual Image Quality Indice (IQI) was calculated by equations 1–3 . [4]

Fig. 10: AutoPIA automatically recognizes the simulated pathological features extracted from the X-ray image.

• IQI value of filaments.

Fig. 11: Image quality indices of filaments.

•  IQIs values of particles.

Fig. 12: Image quality indices of particles.

•  IQI values of circular detail.

Fig. 13: Image quality indices of circular detail.

  The TORMAM phantom was placed between two PMMAs of 10, 20, and 30 mm thickness (used separately). Each phantom generated 10 images under the “LCC image” format and the exposure conditions are shown in Fig. 14. The kV and mAs were chosen to match as closely as possible those selected by AEC, using AEC mode at dose setting N, when imaging a 30-mm, 50-mm and 70-mm thickness of PMMA. To obtain these measurements, the TORMAM phantom was replaced by a 10-mm thick PMMA.

Fig. 14: TORMAM and PMMA phantom set up and exposure conditions. The x-ray exposure was acquired with the compression paddle adjacent to the phantom.

 The processed images (DICOM: “for presentation”) obtained from three parameters were outputted and calculated as IQIs on the analysis software. The Jonckheere-Terpstra test was applied to analyze the trend between these three parameters. The significance level was set at α=0.05.

 IQI values decreased significantly (P < 0.05 for trend) with increasing thickness of the PMMA plates except for the 106-63 μm particles (Fig. 15–17).

Fig. 15: Image quality indices of filaments.

Fig. 16: Image quality indices of circular details.

Fig. 17: Image quality indices of particles.