Quality control in dual head γ- cameras: Comparison between methods and softwares used for image analysis

Transcription

1 Quality control in dual head γ- cameras: Comparison between methods and softwares used for image analysis Abdalrhman Nayl Edam 1, *, Maria Rosa Fornasier 2, Mario de Denaro 2 Abdelmoneim Sulieman 3, Mohammed Alkhorayef 4,5, David Bradley 5,6 1 *Radiation Safety Institute, Sudan Atomic Energy Commission Khartoum, Sudan 2 Medical Physics Department, Azienda Sanitaria Universitaria Integrata Trieste (ASUITS), Italy 3 Radiology and medical imaging department, College of Applied Medical Sciences Prince Sattam bin Abdulaziz University, P.O.Box 422, Alkharj 11942, Saudi Arabia 4 Department of Radiological Sciences, College of Applied Medical Sciences King Saud University, PO Box Riyadh 11433, Saudi Arabia 5 Department of Physics, University of Surrey Guildford, Surrey GU2 7XH, UK 6 Sunway University, Institute for Healthcare Development Jalan Universiti, PJ, Malaysia * abdwsh10@hotmail.com Abstract Patient radiation dose and image quality are the main issues in nuclear medicine (NM) procedures. Currently, many protocols are used for image acquisition and analysis of quality control QC tests. National Electrical Manufacturers Association (NEMA) methods and protocols are widely accepted method used for providing accurate description, measurement and reporting of γ-camera performance parameters. However, no standard software is available for image analysis. The aim of this study was to compare between the vendor QC software analysis and three software from different developers downloaded free from internet; NMQC, NM Toolkit and imagej-nm Toolkit software. The three software are used for image analysis of some QC tests for γ-cameras based on NEMA protocols including non-uniformity evaluation. Ten non-uniformity QC images were taken from dual head γ-camera (Siemens Symbia) installed in Trieste general hospital (Italy), and analyzed. Excel analysis was used as baseline calculation of the non-uniformity test according NEMA procedures. The results of the non-uniformity analysis showed good agreement between the three independent software and excel calculation (the average differences were 0.3%, 2.9%, 1.3% and 1.6% for UFOV integral, UFOV differential, CFOV integral and CFOV differential respectively), while significant 27

2 difference was detected on the analysis of the company QC software with compare to the excel analysis (the average differences were 14.6%, 20.7%, 25.7% and 31.9% for UFOV integral, UFOV differential, CFOV integral and CFOV differential respectively). NMQC software was the best in comparison with the excel calculations. The variation in the results is due to different pixel sizes used for analysis in the three software and the γ- camera QC software. Therefore, it is important to perform the tests by the vendor QC software as well as by independent analysis to understand the differences between the values. Moreover, the medical physicist should know the pixel sizes used in each; the independent software and company QC software. The standard analysis software is needed to give the possibility to compare between the γ-camera systems in order to improve the image quality in NM imaging. Keywords: Quality control; gamma camera; Nuclear Medicine; Image quality. 1. INTRODUCTION The concept of quality control is primarily based on standard measurements and analysis. For gamma cameras quality controls, the NEMA publications are commonly used and provide accurate description about the measurement, analysis and the reporting of gamma camera performance parameters [NEMA 2012], but standard analysis tools are still missing in the recommendations. The analysis of Some quality control (QC) tests are not easy to perform manually, and require a long time, e.g. non uniformity, spatial resolution and linearity, and center of rotation (CoR). Moreover the accuracy and the reproducibility of the analysis are important issues. New models of SPECT/CT dual head gamma cameras are frequently equipped with QC software tools developed by the company. The use of this facility in managing the periodical QC shows the evident advantage to perform the tests directly by the equipment console, with optimization in saving working time and results record. Nevertheless, in some cases it is not so clear the processing steps and algorithm followed by the company to analysis the QC images and one could wonder if the evaluation method follows or not the analysis recommendations of the NEMA recommendations. Due to these considerations, it can be a 28

3 good practice to perform the tests by the company QC tools as well as by independent measurements in such way to find an agreement in the results. Non uniformity test is the most important test in gamma cameras performance check, usually performed daily or weekly depending on the manufacturer recommendations. The new SPECT/CT systems equipped with 57 Co source to perform the test automatically by simulating NEMA approach, but with a few differences; in the source ( 57 Co instead of m99 Tc), and the detector irradiation distance (30 cm instead of five times Field of View FOV). These differences can lead to more processing in the images after acquisition to calculate the non uniformity by the system QC software. Moreover the software of the company is not standard comparing with the other company QC software, so in this situation the uniformity performance of the gamma camera doesn t evaluated by standard manner, and the acceptance criteria should be vary between gamma camera to other. Some authors developed independent software to analyze gamma camera non uniformity images [Rova et al., 2008; NM Tool kit 2016; Carlier et al, 2005; Kelmpa 2011; Demirkaya and Al Mazrou; 2007]. The present study was aimed to perform the analysis of integral and differential non uniformity in Useful Field of View UFOV and Central field of View CFOV for ten flood images by using three free software downloaded free from internet independently from the QC software of the SPECT/CT system based on NEMA protocol, and to compare the results. 2. MATERIALS AND METHODS The present study was conducted in Sanitaria Universitaria Integrata Trieste (ASUITS) hospital, department of Nuclear Medicine, Trieste-Italy. Dual head Siemens Symbia Intevo Excel SPECT/CT system was used in this study. The system is fully automatic collimator changing design and equipped with QC software. The basic specifications of the two detectors are show in Table (1). 29

4 Table (1): Symbia SPECT/CT basic specifications. Model Detector Dimensions Crystal Intevo 16/6/2 Excel FOV Diagonal FOV Size Diagonal 53.3 x 38.7 cm 63.5 cm 59.1 x 44.5 cm 69.2 cm Thickness 9.5 mm (3/8 inch) Photomultiplier Tubes Total number 59 Array Hexagonal Intrinsic Flood Field 75 kcps Integral UFOV & CFOV 3.7 & 2.9 (uncorrected) Differential UFOV & CFOV 2.7 & 2.5 Intrinsic Spatial Resolution FWHM in UFOV & CFOV 3.9 & 3.8 The system also has built-in point source 57 Co for automatic QC of intrinsic non uniformity, calibration and verification measurements. Another 157 Gd line source for automatic alignment and extrinsic measurements. Figure (1) shows the sources configuration and setup for the detectors irradiation in automatic non uniformity test acquisition. 30

5 Point source Line source in shielding position Line source without shielding CT Gamma camera gantry Head 1 Head 2 Patient couch Fig (1): Configuration of built-in sources in Symbia SPECT/CT system. The NEMA non uniformity test images were produced in the system (10 images, every week two images for detector 1 and detector 2). Later the images were extracted from the consol of the system. Firstly the system was performed the automatic calibration and verification automatically and then the uniformity for the two detectors by simulating NEMA procedures, this done by using built-in 57 Co source by the following automatic setup, Figure (2). The results of non uniformity directly reprted after the acqustion. Fig (2): Setup for calibration, verification and uniformity of the two detectors. 31

6 Also According to NEMA setup the non uniformity test was performed with 99m Tc source in small vial (5 cc, 30 MBq = 0.8 mci), and polystyrene stand for supporting the vial in front of the detectors [NEMA 2012]. The matrix size for this acquisition was 128*128 (4.795 mm pixel size), this was done in order to produce the non uniformity images independently from the automatic QC software. The obtained images were processed by manual calculation for integral and differential non uniformities for each UFOV and CFOV by an Excel sheet using Equation (1). (1) Also three simple software were used for analyze the images, namely; Nuclear Medicine Quality Control (NMQC), Nuclear Medicine Toolkit (NM Toolkit) and imagej nuclear medicine toolkit. The criteria for choosing the software were to be; free, easy from the practical point of view and fast in processing. NMQC is a free, MATLAB based software and reads both DICOM and interfile images format. NMQC follows NEMA NU in image processing and analysis including basic quality control tests for gamma camera i.e. intrinsic uniformity, intrinsic resolution, center of rotation and tomographic uniformity [Rova et al., 2008]. NM Toolkit also is a free software package supporting NEMA based acceptance testing procedures and routine SPECT QC testing (intrinsic resolution, uniformity, Jaszczak phantom measurements and Specphan phantom measurements). The application supports DICOM images format [IRIS 2016]. In addition, the images were analyzed using a plug-in developed in imagej called Nuclear Medicine Toolkit. The plug-in supports quality controls for gamma cameras and PET cameras based on NEMA and IPEM analysis for intrinsic uniformity and CoR [Carlier et al., 2005]. 32

7 3. RESULTS AND DISCUSSION The integral and differential non uniformities were calculated firstly for the ten images by using excel sheet using the default pixel size selected in the images of Symbia (4.795 mm), the analysis had been done for both UFOV and CFOV. The macro was made for the reminder images, and then the results were obtained easily. The results of excel were taken as the base line values for the non uniformities. Also the non uniformity was analyzed by NMQC software, but in this case the pixel size was modified to be 4.8 mm, while for the NM toolkit and imagej toolkit plug-in the default pixel size was used, because in the two later software no possibility to change the pixel size. Figures (3) and (4) show the results of integral non uniformity in the UFOV and CFOV respectively, for the four analysis tools as well the company QC software. Fig (3): The results of integral non uniformity in UFOV by the four analysis tools and the Symbia QC software. 33

8 Fig (4): The results of integral non uniformity in CFOV by the four analysis tools and the Symbia QC software. As clearly showed in Figures (3) and (4), there is some correlation between the three software analysis and the excel analysis, while some gab detected with the Symbia QC software. In fact the NMQC software was the best one among them with respect to excel analysis. The little differences between the three software and the excel calculation mainly due to determination of the UFOV and CFOV, because it is automatically determined in each of the three software, also it is well known, on the periphery pixels of the detectors the account is not accurate for the events and the extraction of the un-useful pixels is not easy process in the analysis. The differences between the excel and the Symbia QC software analysis resulting from the different pixel size used i.e. in the excel calculation the pixel size used was mm, and for Symbia QC software was 7.79 mm, as well as the different geometrical setup for the detector irradiation (figures 3.6 and 3.7), other difference in the sources i.e m99 Tc for excl and 57 Co for Symbia QC software). Figures (5) and (6) show the results of differential non uniformity for UFOV and CFOV by the five software. In the Figures (5) and (6), it is clearly the same behaviors of the five methods in the integral uniformity. Since the non uniformity for each five pixels affecting by the periphery pixels including in the analysis, again the determination of the UFOV and CFOV was a little different between the five analysis tools. 34

9 Fig (5): The results of differential non uniformity in UFOV by the four analysis tools and the Symbia QC software. Fig (6): The results of differential non uniformity in CFOV by the four analysis tools and the Symbia QC software. 35

10 Table (2) represent the averaged percentage differences between the best tool i.e. NMQC software, and the excel analysis for the four performance (integral and differential non uniformity for UFOV and CFOV). Also the differences between excel and Symbia software are reported in the Table (2), in addition, the averaged differences between the exel and the three independant software. Table (2): The averaged percentage differences between excel, NMQC software and Symbia QC software. Parameter UFOV, Integ % UFOV, Differ % CFOV, Integ % CFOV, Differ % Excel Vs three software Excel Vs. Symbia Excel Vs. NMQC The important point we noted in our study is that; the recommended pixel size by NEMA is 6.4±30% mm; actually this range is wide (i.e. fom 4.48 to 8.32 mm), and can lead to variation in the non uniformity results due to smoothing effect of the pixel counts, that is happened when the dimension of the pixel is changed from small to larger size within the range of 30%. For this reason it is not easy to establish acceptance criteria for non uniformity test, because that is should be based on the analysis tool you used and the pixel size, e.g. Zanzonico recommended the NEMA value should not exceed 5% for floods acquired using between 10 and 15-million counts [Zanzonico 2008]. In fact this value is accepted as a criteria for non uniformity according to the equation (1), in other hand the maximum pixel and minimum pixel depend on the analysis tool, also for determination of the pixels including in the analysis, and agin this is depent in the determination of the UFOV and CFOV. For this concedrations the medical physiscst should understand very well the proccesing of the vendor 36

11 QC software and the pixel size used in the analysis by simble descution with the company engineer. Recently in the new models of SPECT/CT, the QC test for non uniformity is done automaticaly by the system and the resuls reported directly by the system software and recorded for the future measurements. Obvioslty this can be useful during clinical work flow and the bussy work load, but the results are unique for that system and the medical physicist have no possibility to compair the results with the other system to have an idea about the level of the system performance, i.e the x value for the non uniformity in one system can not be the same as the x value in the other system. In the other imaging modalities like in diagnostic x-ray, there is some sort of standarization of measurement tools i.e the protocol, and the measurement tools usaly are calibrated in primary or secondary labrotaies, so they can used for comparison between the different systems, but in gamma camera QC masurments there is some sort of lack in the standarization of the analysis partially in non uniformity, spatial resolution and linearity tests. For these concidrations it is important to find customary software to compair gamma cameras system by independant way. 4. CONCLUSIONS Non uniformity analysis for ten flood images was performed by three different software and by the vendor QC software based in NEMA standards. Excel was used to analyze the images to establish the base line values for the non uniformity. Variation was detected in the results between the five tools (i.e. company QC software, excel, NMQC, NM toolkit and imagej toolkit plug-in). A little different was found between the excel and the three software due to differnt determination of the UFOV and CFOV. While segnificant differnet was funded between the excel and the vendor QC software due to differnt pixel size used in the analysis. As a recommendation of our experiance, The medical physicist should understand the non uniformity analysis processing by the gamma camera automatic QC software. Pixel size is 37

12 very important factor in the uniformity analysis and it should be noted by the medical physicist when the analysis is performing. The exesting of independant stadard tool is important in the quality cotrol measurements in gamma camera systems. The benefit of standard tools could be for comparison of the gamma camera systems in one department or ever with the other hospitals for research purposes and to improve the image quality of the system. REFERENCES Carlier T; Ferrer L; Jean B; Berruchon R; Cuissard A; Martineau P. Loonis; O. Couturier O(2005). Quality controls for gamma-cameras and PET-cameras: development of a free open-source ImageJ program. Proceedings of the SPIE, 5745, Demirkaya O; Al Mazrou R. (2007). Performance Test Data Analysis of Scintillation Cameras:, IEEE transactions on nuclear science. 54 (5) IRIS QA.NM Tool kit, Nuclear Medicine gamma camera software package, IRIS QA LLC. Available at: Klempa K. (2011). Uniformity Testing: Assessment of a Centralized Web-Based Uniformity Analysis System: Journal of Nuclear Medicine Technology. 39: National Electrical Manufacturers Association (NEMA), Performance Measurements of Scintillation Cameras: Rosslyn, NU-1 NEMA Rova A; Celler A; Hamarneh G (2008) Development of NEMA-based software for Gamma Camera Quality Control: Journal of Digital Imaging. 21(2): Zanzonico P (2008). Routine Quality Control of Clinical Nuclear Medicine Instrumentation: A Brief Review:, J Nucl Med 2008; 49: , DOI: /jnumed