From a clinical point of view, the threshold image contrast is an extremely important physical parameter. High image contrast guarantees that all the smallest changes that may be present in the examined breast (microcalcifications) are recorded in the image. It should be remembered that the threshold image contrast improves with the increase in the dose emitted by the X-ray tube during each projection of the mammographic examination. The increase in the dose of ionizing radiation contradicts the basic principles of radiological protection of the patient. Since mammographic examinations are performed mainly in women as part of a screening program, special care should be taken to ensure the level of exposure to ionizing radiation of potentially healthy women. When calibrating mammography devices with a digital image detector, regardless of whether they are CR or DR mammography devices, special attention should be paid to setting the optimal exposure parameters that will provide an appropriately low dose of ionizing radiation or a high threshold image contrast. Such a setting will provide a high-quality image while maintaining the low dose of ionizing radiation that women receive during a mammography examination. It is therefore extremely important that the user of the mammography device ensures periodic measurement of the threshold image contrast on their device and measurement of the dose in a correct manner and in accordance with available scientific knowledge.

Underestimating the threshold image contrast due to analyzing too few images may result in falsely assuming the imaging system is performing within acceptable limits. In practical terms, this can lead to failure in detecting subtle but clinically important features such as microcalcifications, especially at early stages of breast cancer. Furthermore, inadequate image sampling may reduce the sensitivity of quality control protocols, masking gradual degradation in detector performance or exposure system calibration. This, in turn, could affect compliance with accreditation standards (e.g., EUREF or national QA programs) and potentially delay necessary maintenance or corrective actions. Therefore, ensuring an adequate number of images during CDMAM testing is not only a technical consideration but also a factor directly influencing diagnostic confidence and patient safety.

According to the EUREF recommendations from 2006, to obtain the correct value of the threshold image contrast, it was required to analyze at least 6 images in the "for processing" format. These recommendations changed in 2013, where it was recommended to perform the analysis of at least 16 images in the "for processing" format. The image analysis software provided by the image manufacturer requires the analysis of at least 8 images of the CDMAM phantom. It should be added here that between subsequent exposures the phantom should be moved so that the disks are not located over the same pixels, which ensures the extraction of information from the image about artifacts not related to the image detector. It can therefore be seen that the number of analyzed images has a key impact on the measurement result. It is important to use images in the analysis in the "for processing" format, which are characterized by a linear relationship between dose and pixel value. The use of images in the "for processing" format ensures the repeatability and reliability of the obtained analysis results. In the images in the "for processing" format, CAD (Computed Aided Detection/Diagnosis) software has not applied filters to improve the visibility of anatomical structures and bring their appearance and contrast closer to analog images. The use of such filters does not allow for maintaining the repeatability of the analysis of images obtained from different types of mammograms. Unfortunately, there is no research in the available scientific literature on the number of CDMAM phantom images analyzed for which the obtained threshold image contrast result will be the best.

Maintaining low threshold image contrast, especially for small diameter disks, ensures visibility of microcalcifications—key in early breast cancer detection. For CDMAM 4.0, the use of 24 images improved precision of measurements for 0.1 mm disks by approximately 1.2%, reducing uncertainty from ~ 3.6% to ~ 2.4%. Though this may appear minor, even small improvements in contrast detection can affect sensitivity in borderline cases.

In principle, with the increase in the number of analyzed images, the threshold image contrast value should be obtained, which is precisely determined, and its standard error is smaller. This situation is generally observed in Tables 1 and 2. In the case where the number of images analyzed had no effect on the obtained result, all values of the threshold image contrast should have the same value. The results obtained show that this is not the case.

To determine the optimal number of analyzed images, at which the result is the most reliable, two important parameters must be considered. The first is the obtained result of the threshold image contrast. According to the principles of statistics, it must take the values as small as possible with a simultaneous low standard error. The second is the number of analyzed images. This number must be low enough to perform the test using the CDMAM phantom in an acceptable time. Several or a dozen measurements can be performed at an acceptable time, while several hundred measurements are impossible to perform during routine operation of the device.

The low value of the threshold image contrast is particularly important in the context of controlling the exposure automation system. In accordance with the EUREF recommendations, the value of the threshold image contrast for a disk diameter of 0.1 mm is considered in the test for compensation for changes in the phantom thickness and the value of the high voltage. Therefore, an appropriately low and thus reliably determined value of the threshold image contrast for a diameter of 0.1 mm significantly affects the verification of the correct operation of the exposure automation, which in current mammographs is crucial for obtaining images of the highest diagnostic value.

In the CDMAM 3.4 phantom, it is not possible to clearly determine which group of results is characterized by the lowest or highest values. From the point of view of statistics, the highest values of the threshold image contrast should be obtained for the group in which only two images were analyzed—in this case, the highest values are obtained in the group in which 6 images were analyzed. It is not possible to indicate the group in which the threshold image contrast values were the lowest. This situation is very clearly visible in the statistical analysis performed using Tukey and NIR tests. Due to the large diversity of the threshold image contrast results in individual groups, many different groups of results are recorded that differ statistically from each other, which is clearly visible in the statistics performed. This means that it is not possible to clearly state how many images of the CDMAM 3.4 phantom should be analyzed to obtain the optimal result of the threshold image contrast. One can only venture to say that from the point of view of the threshold image contrast value for a 0.1 mm diameter disk, it is best to use 16 images for analysis—if we use this value to evaluate the exposure automation system. In the opinion of the authors of this paper, the EUREF recommendations regarding the number of images analyzed should be followed for the CDMAM 3.4 phantom. This suggestion does not stem from the present data, which showed inconsistent trends across groups and diameters, but rather reflects alignment with the EUREF 2013 guidelines. Given the observed variability in results, we recommend continuing with the existing standard of 16 images, particularly for 0.1 mm disks used in automatic exposure control tests, until more robust evidence allows for an updated recommendation.

The situation is completely different when analyzing groups of images obtained using the CDMAM 4.0 phantom. Statistical analysis showed that there is only a statistically significant difference between the group and the other groups and N4 vs. N24 and N8 vs. N24. Analyzing the values in the individual groups, it is clearly visible that the lowest values were shown in the group in which 24 images were analyzed and the highest in the group in which 2 images were analyzed. To obtain optimal results, it is recommended to analyze at least 24 images of the CDMAM 4.0 phantom based on the measurements performed. With such analysis, we obtain results characterized by the smallest values with the smallest standard error (the results are precisely determined). EUREF recommends taking at least 16 images every 6 months. If it takes about 1 min to obtain one image in a DR mammogram, the total test time increases by about 10 min at most. An increase in the number of images analyzed by 8 will not significantly extend the time of performing this test, especially since we perform it every 6 months. An increase in the number of analyzed images increases the precision of the measurement by approx. 1.2% in the range of small disk diameters, which is particularly important in the context of the exposure automation control test.

It is also worth noting that in the case of both phantoms, the differences in absolute values between the individual groups become blurred with the increase in the diameter of the analyzed disks. This is consistent with theoretical predictions. Although this tendency is better observed for CDMAM 4.0 phantom than for CDMAM 3.4, in the CDMAM 4.0 phantom, the threshold image contrast practically does not change for disks above 0.57 mm.

In the above analysis, software from one manufacturer was used, allowing for the simultaneous analysis of two types of phantoms. However, it should be remembered that the method of reading images is the responsibility of cdcom.exe files/programs, which are a fundamental part of every software for reading CDMAM images. It is responsible for, among other things, detecting disks on the phantom, smoothing the detection matrix with a Gaussian filter, and fitting a polynomial to the thickness-diameter relationship. There are different versions of cdcom.exe, which have been improved over the years and for phantom versions. It is cdcom.exe that is largely responsible for the result of the analysis with the CDMAM phantom and thus the analyses presented in this work. From a practical standpoint, implementing a 24-image protocol for CDMAM 4.0 phantom analysis appears feasible in most clinical settings, particularly those using DR systems with automated acquisition. While doubling the number of images from 12 to 24 increases test time, the gain in measurement precision is especially notable for small diameters. Facilities with resource constraints may adopt a tiered approach—performing 24 images biannually during full QA cycles and fewer images during interim checks.

One of the limitations of this study was the inability to determine an optimal number of images for CDMAM 3.4. Although statistical testing showed differences between several image groups, the mean contrast values fluctuated nonlinearly, and no stable minimum could be identified across all diameters. This may result from higher inherent variability of the phantom structure, lower object count per diameter, or limitations in phantom design. Additionally, the relatively smaller number of total repetitions (10 per group) may have influenced the statistical robustness. As a result, the study supports adherence to EUREF recommendations (≥ 16 images) for CDMAM 3.4 but does not provide a new, evidence-based recommendation for clinical users.

Of course, the results of this work may turn out to be different in the case of a significant increase in the number of repetitions in individual groups. However, in the authors’ opinion, the trends described for individual versions of CDMAM phantoms will not change significantly. Another certain limitation of this work is the fact that the images were obtained from only one mammogram (i.e. only with one exposure parameter).

The generalizability of these findings is limited by the use of a single mammography system (Hologic DR) and a fixed exposure parameter set. Results may differ with other vendors or technologies (e.g., CR systems), due to variations in detector characteristics, image processing algorithms, and exposure control methods.

It would be necessary to check in later studies how the described trends are presented on mammograms from other manufacturers or on CR (Computed Radiography) mammograms. At the same time, it should be remembered that a properly functioning digital mammograph has no problems with obtaining an image that meets the threshold image contrast requirements specified in the EUREF Guidelines. This can be seen in the presented analysis. Regardless of the number of images analyzed, the results in a properly functioning mammograph will be within the tolerance range. Of course, if the device functions at the border of acceptance (this can happen especially in the case of CR systems), the number of analyzed images is very important. The differences between the minimum and maximum values between groups, regardless of the phantom version, are at the level of about 5% (Table 1 and Table 2). The threshold image contrast must meet the requirements for disks with a diameter of 0.1, 0.25, 0.5, and 1 mm. If even one diameter does not fit within the specified limit, the device must be repaired accordingly. The absolute value of the threshold image contrast for a disk diameter of 0.1 mm is only used in the automatic exposure test. Therefore, every care must be taken to ensure that the threshold image contrast is reliably determined for this disk diameter.