Logo of the Physikalisch-Technische Bundesanstalt

Computed Tomography Image Quality – Development of a Measurement Procedure

23.12.2021

A proposal that was put forward by the German Federal Office for Radiation Protection (BfS) for a measurement procedure to determine a test specification for the image quality of clinical CT systems was further developed and tested by the UAG‑CT Dosis BQ Task Group of the AG‑X (formerly AK‑RöV). The BfS and PTB closely cooperated to establish a measurement and analysis procedure that aggregates a group of data for low‑contrast detectability (LCD) into one test statistic.

The Radiation Protection Act (StrahlenschutzgesetzStrlSchG) requires the necessary image quality in the field of medical imaging to be realized with the lowest possible radiation exposure. The expert guideline (SachverständigenrichtlinieSVRL), the quality assurance guideline (Qualitätssicherungsrichtlinie, QSRL), and the standards do not currently include any measurement requirements for low‑contrast detectability in the field of X‑ray computed tomography (CT). The new filtered back projections for CTs are the reason for this shortcoming. These image reconstruction techniques are nonlinear, making established measures of image quality inadequate.

The proposal of the German Federal Office for Radiation Protection for a measurement procedure to determine a test statistic [1,2] was discussed, further developed, and tested in the UAG‑CT Dosis BQ (a subcommittee of the AG‑X, formerly AK‑RöV).

For an initial test on three CT devices from different manufacturers, the CCT189 MITA body phantom (The Phantom Lab, Salem, NY, USA) was chosen together with an attenuator that simulates a realistic body size (see Figure 1). A CT protocol for the examination of the upper abdomen (for example liver metastases) was used. A total of 200 scans of the test object were performed at three different dose values that are expressed by the CT dose index (CTDI, in mGy). In one image plane, four contrast rods with different diameters and contrasts are visible (or not). Both the standard reconstruction procedure (FBP, filtered back projection) and at least one iterative procedure per manufacturer were used. To determine the low‑contrast detectability d’, the established channelized Hotelling observer (CHO) was applied that represents the theoretically ideal linear model observer [3]. The BfS and PTB closely cooperated to establish an evaluation scheme that combines the data into one test statistic [4].

test object and CT scan

Figure 1: On the left: The test object, CCT189 MITA, together with an attenuator with an external diameter of 32 cm on the imaging table of PTB's own CT scanner; on the right: CT scan of the arrangement with a high dose.

When the protocol was applied, data for the low‑contrast detectability d’ for four contrast rods, for three dose values, and for at least two reconstruction procedures were acquired for each CT. To compile the data in one test statistic per modality (device + reconstruction procedure), the low‑contrast detectability values were first integrated over the dose using a trapezoidal approximation. Then a weighted mean was taken over the four contrast rods, using the reciprocal value of the product of the diameter and the contrast as a weight. The uncertainty estimation was made according to the Guide to the Expression of Uncertainty in Measurement (GUM) [5]. The uncertainty of the weighted mean was determined by means of a Monte Carlo calculation according to Supplement 1 to the GUM [6].

A mean value for the CTDI was also determined. Forming a dose efficiency index as a ratio of image quality and dose was not considered as reasonable, since the uncertainty of the dose is exceedingly high: The uncertainty of the dose efficiency would be determined by the (exceedingly high) uncertainty of the dose, as shown by the available data.

The results obtained from the weighted mean d’m for the three CT systems that were assessed so far are shown in Figure 2. In this context, the x‑axis represents the device and the reconstruction method. Two data sets were generated according to each modality in order to assess the reproducibility of the results. The conformity of the two data sets is very good. In general, d’m shows only minor differences between the FBP and the iterative procedures, with the exception of the most advanced iterative procedure of CT 3. The error bars represent a coverage interval of 90 %. This selection results from the acceptance test recommended by the German Commission on Radiological Protection (SSK) with unilateral limitation [7,8]: The detectability d’ should not fall below a minimum value limit.

diagramme (detectability)

Figure 2: Weighted mean of the detectability d’m for three different CT devices. Two nominally identical data sets with FBP reconstruction (FBP‑I and FBP‑S) and two and four nominally identical data sets with iterative reconstruction methods (IR*‑I and IR*‑S, respectively) were examined for each device. Error bars indicate the 90 % coverage interval, assuming an 8 % uncertainty in the CTDI.

At BfS, the evaluation procedure is also implemented in the Python language. The results are first to be compared with the results obtained at BfS by means of Matlab software. The BfS software is, to the greatest possible extent, to perform all evaluation steps automatically. In the next step, data from a large number of CT devices from as many manufacturers as possible are to be collected in order to test the procedure’s “suitability for the masses” and to be able to determine a meaningful (lower) threshold value for d’m for the type test.

Literature

[1] de las Heras Gala, H. & Schegerer, A.: Konzept zur Charakterisierung klinischer CT‑Systeme unter Einbeziehung von Bildqualität und Dosis (2019)

[2] Racine, D.; Viry, A.; Edyvean, S. & Verdun, F. R.: Characterization of clinical CT systems using a dose efficiency index (DEI) - project 3613S20007. Bundesamt für Strahlenschutz (BfS) (2017)

 [3] Wunderlich, A., Noo, F., Gallas, B.D. & Heilbrun, M.E.: Exact Confidence Intervals for Channelized Hotelling Observer Performance in Image Quality Studies, IEEE Transactions on Medical Imaging, 34 (2) (2015)

[4] Anton, M. & de las Heras Gala, H.: d’m als Messgröße für die Typprüfung, handed out in the UAG-CT Dosis BQ (July 2021)

[5] JCGM100, BIPM, Working Group 1 of the Joint Committee for Guides in Metrology (JCGM/WG 1): Evaluation of measurement data - Guide to the expression of uncertainty in measurement. GUM 1995 with minor corrections (2008)

[6] JCGM101, BIPM, Working Group 1 of the Joint Committee for Guides in Metrology (JCGM/WG 1): Evaluation of measurement data – Supplement 1 to the “Guide to the expression of uncertainty in measurement”. Propagation of distributions using a Monte Carlo Method (2008)

[7] JCGM106, BIPM Joint Committee for Guides in Metrology (JCGM 106): Evaluation of measurement data – The role of measurement uncertainty in conformity assessment (2012)

[8] Strahlenschutzkommission: Methodik zur Berücksichtigung von Messunsicherheiten bei messtechnischen Prüfungen im Geltungsbereich der Röntgenverordnung und der Strahlenschutzverordnung. Geschäftsstelle der Strahlenschutzkommission (2016)

Contact

Opens local program for sending emailM. Anton, Department 6.2. Working Group 6.24