Bio-Medical Materials and Engineering 24 (2014) 85–94 DOI 10.3233/BME-130787 IOS Press
85
4D-CT reconstruction based on pulmonary average CT values ZHANG Shu-xua,* , ZHOU Ling-hongb, LIN Sheng-qua, YU Huia, ZHANG Guo-quana, WANG Rui-haoa and QI Binga a
Radiotherapy Center, Affiliated Tumor Hospital of Guangzhou Medical University, Guangzhou 510095, China b Department of Biomedical Engineering, Southern Medical University, Guangzhou 510095, China
Abstract. To date, commercial 4D-CT systems typically depend on an external respiratory monitoring device. Immobilizing patients in a thermoplastic mask while receiving radiotherapy may result in a failure of 4D-CT reconstruction. The aim of this study is to investigate the feasibility of 4D-CT reconstruction based on a method using pulmonary average CT values (ACV) without an external respiratory monitoring device. The ACV of the whole lung assumes cyclical variation during respiration. Phases of CT images were identified by calculating the ACV over time. Subsequently, five sets of 4D-CT images based on a Real-time Position Management (RPM) system were selected to verify the ACV method. The entire lung CT datasets of another sixteen free-breathing patients were acquired in Cine scan mode for multiple couch positions. The phase of every CT image was identified and re-sorted into different phase 4D-CT volumes by analyzing the time dependence of the corresponding ACVs. This paper demonstrates the ACV method using the 4D-CT data sets based on the RPM system. Convenient and reliable 4D-CT reconstruction can be accomplished without any external respiratory monitoring device using ACVs. Key words: 4D-CT; image reconstruction; average CT value; respiration phase
1. Introduction Recently, the use of imaging techniques to improve treatment planning and delivery for patients undergoing radiotherapy has generated a number of divergent four-dimensional computed tomography (4D-CT) resorting methods, e.g. [1-5]. Compared with conventional free-breathing CT scans, 4D-CT can not only reduce motion artifacts but also represent the motion states of internal organs. Existing 4D-CT systems are typically based on the Real-time Position Management (RPM) system (Varian Medical Systems, Inc., Palo Alto, CA, USA) to record respiratory cycle information [2-4]. During the image acquisition, the RPM system correlates the patient's breathing cycle with CT imaging. After scanning, all CT images are grouped and sorted according to their corresponding phases to obtain eight to 10 series of complete CT data sets. Each is equivalent to a static ordinary CT scan taken at a particular point in the respiratory cycle. All series together reflect the movements and changes of internal organs across the respiratory cycle. The system is then used to track the motion amplitude of a reflective marker placed on the patient’s abdomen during treatment, which is converted *
Address for correspondence: Professor ZHANG Shu-xu, Radiotherapy Center, Affiliated Tumor Hospital of Guangzhou Medical University. NO.78 Hengzhigang Road, Guangzhou, 510095, China. Tel: +86-13005110376; Fax: +86-2066673635; E-mail: [email protected] 0959-2989/14/$27.50 © 2014 – IOS Press and the authors. All rights reserved
86
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
into the corresponding respiratory breathing phase so that delivery can be gated to a fraction of the respiratory cycle. Other types of devices have also been used to acquire an external respiratory cycle signal in other 4D-CT reconstruction methods: a spirometer to measure the tidal volume through the mouth [1], and elastic belts to measure pressure around the abdomen [6]. However, respiratory cycle signals obtained from these external surrogates may not always accurately represent internal organ motion, especially when the breathing patterns become irregular. In addition, when a patient receiving radiation therapy is immobilized in a thermoplastic mask, the external respiratory monitoring device mentioned above cannot measure the respiratory cycle signal, which may result in the failure of 4D-CT reconstruction and delay or prevent treatment. Therefore, a novel 4D-CT reconstruction method based on pulmonary average CT values (ACVs) without any external respiratory monitoring device is proposed and demonstrated in this study. 2. Materials and methods 2.1. ACV method and verification According to the principle of CT imaging, the CT value of tissue X is computed by the following equation: HU(x) = 1000 × (µx-µw) / µw,
(1)
where HU(x) is the CT value of tissue X, µx is the attenuation coefficient of the tissue X, and µw is the attenuation coefficient of water. The attenuation coefficient of air is defined as 0. Thus, the CT values of water and air is 0 (±10) and −1000 HU, respectively. During expiration, the decrease of air volume in the lung leads to an increase of the average CT value, vice versa. Thus, it seems reasonable to presume that the ACV of the whole lung changes periodically. If the breathing cycle begins from the end-exhale to end-inhale and then return to end-exhale, the air content of lung therefore changes from the minimum to the maximum and return to the minimum. The plot of ACV of the lung against time will appear like the shape of the letter “V”, and can be used as the respiratory signal. The respiratory cycle and phase of the CT image can then be identified by the ACV-time plot. This method is named the “ACV principle”. To verify this method, the 4D-CT datasets of five patients were analyzed and reconstructed using the RPM system [7]. The lung ACVs of each 4D-CT volume at different phases were calculated (Table 1) and the ACV-phase graphs were plotted (Fig. 1). 2.2. Image acquisition The existing RPM 4D-CT reconstruction method is under the assumption of a repeated respiratory cycle and amplitude of the scanned object throughout treatment. Without this hypothesis, the acquired 4D-CT images cannot be reconstructed. The methods of CT image acquisition and reconstruction was previously reported [8], which are simply stated as follows: Before CT data acquisition, patients underwent breathing training to improve the reproducibility of respiration. The entire thorax and upper abdomen of patients with free breathing were acquired in Cine mode on a 16-slice GE LightSpeed CT scanner (GE Medical Systems, Waukesha, WI, USA). The imaging parameters were set as follows: 120 kVp, 0.5 sec gantry rotation,
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
87
0.45 sec Cine interval, 512 × 512 matrix with 2.5 mm slice thickness. The duration of a Cine scan at a couch position was slightly greater than or equal to the average respiratory cycle plus 1 sec, which was set as 6 sec in this study. A volume of 2 cm in length along the cranial-caudal direction was imaged with 12 reconstructions in each Cine scan and eight layers for each reconstruction with 2.5 mm in slice spacing. The CT acquisition was turned off when a couch position Cine scan was completed. Then, the table was moved to the next position and another Cine scan was launched. The acquisition was repeated until the entire thorax and upper abdomen volume was fully acquired. 2.3. Image grouping and segmentation Based on the ACV principle, a 4D-CT reconstruction software (named ACV 4D-CT) was developed using VC++ language based on the Windows platform. The acquired CT images were read into the system and grouped according to the couch position and reconstruction order in the Cine scan. Then, the lung tissues were segmented from the CT images using an improved adaptive threshold segmentation algorithm [9]. As a result, the lung tumor focus was retained, the regions of trachea and main bronchi were removed, and the lung edges were smoothed. When no abnormal lesion was discovered in the inspected patients, the segmentation result was considered to be ideal. When the peripheral lung cancer adhered to the chest wall or the central-type lung cancer adhered to the mediastinum, artificial auxiliary segmentation was incorporated. As a result of the image segmentation, only the transaxial section of lung tissue was present, while other pixels corresponding to the couch, clothing, bed sheet, immobilization device and muscles were removed. Finally, the results were reevaluated by experienced physicians of imaging diagnosis. 2.4. Computing of pulmonary ACV The ACV of lung in each CT image is computed by the following equation: ACV(k) = ΣHU(i, j) / N,
(2)
where k indicates the slice number at a reconstruction of Cine scan, HU (i, j) is the CT value of the pixel at row i and column j of a lung CT image, and N is the total number of the lung pixels. At each couch position, the duration of a Cine scan is 6 sec, and a total of 12 reconstructions were performed. The ACVs for eight slice images of each reconstruction were calculated, and the ACV variation with the reconstruction time sequence is shown in Figure 3. 2.5. Image sorting and 4D-CT reconstruction When the patients were breathing regularly, after Cine scanning the relationship between lung ACV versus the reconstruction time is very similar among different couch positions. The starting time of Cine scanning is random at different couch position, so the initial phase of the ACV-time graph is not the same at different couch positions. The time cost of the data acquisition is only 1-2 min. If the patient had received breathing training, the respiratory amplitude and frequency is assumed to be constant. The phase of each CT image can be identified according to the ACV-time graph. Sorting by phase, multiple CT series with different phases can be obtained and are named 4D-CT data sets, which cover the whole lung. Every single-phase 4D-CT data set is equivalent to a static ordinary CT scan taken at a particular point in the respiratory cycle. After the axial 4D-CT data sets are acquired, the coronal, sagittal section of the 4D-CT data sets can be reconstructed based on the axial 4D-CT.
88
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
3. Results 3.1. ACVs of 4D-CT change with breathing phase Table 1 lists the entire lung ACV of single-phase 4D-CT data set for the five subjects whose 4D-CT were reconstructed with an RPM system [7]. For each subject, the ACV of lung changes with phase from a lower value, which is corresponding to the end-inhale with maximum content of air, to a larger value (end-exhale) and return to a lower value (end-inhale again). To intuitively demonstrate the regularity of the ACVs of the whole lung changing with the respiratory phase, the entire lung ACV versus image phase is shown in Figure 1 for each patient based on the data mentioned in Table 1. It demonstrates that the whole lung ACVs of each patient present a cyclic variation during respiration. From end-inhale to exhale and back to inhale, a sound change of ACV over the respiratory phase was revealed. Table 1 The entire lung ACV of different phase 4D-CT data of five patients, whose 4D-CT reconstructions are based on the RPM system and Cine model scan. Entire Lung ACV of different phase 4D-CT volume (HU) Phase Patient No.1 Patient No.2 Patient No.3 -674 -817 -688 1 -667 -814 -676 2 -646 -800 -652 3 -628 -795 -645 4 -608 -792 -644 5 -598 -785 -627 6 -597 -776 -632 7 -620 -777 -646 8 -641 -794 -661 9 -668 -807 -676 10
Patient No.4 -765 -755 -752 -741 -726 -720 -723 -730 -751 -763
Patient No.5 -770 -757 -742 -735 -723 -714 -710 -728 -744 -763
Fig. 1. The graphs of the whole lung ACV-phase of five patients, whose 4D-CT reconstructions are based on the RPM system, demonstrate that the whole lung ACVs present periodic variation during respiration process. When breathing motion changes from end-inhale (phase 1) to end-exhale (phase 6/7/8), the air content of the lung decreases, which results in an increase in the ACVs. When breathing motion is from end-exhale to end-inhale (phase 10), the air content of the lung increases leading to a decrease in the ACVs.
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
89
3.2. Lung tissue segmentation With proper acquisition parameters and image segmentation methods, the lung tissue on a transaxial section CT image can be completely segmented as mentioned in Section 2.3 and shown in Fig. 2. Then the segmented lung tissue image can be used to compute the ACV of the lung.
Fig. 2. The lung tissues were segmented from the CT images acquired with proper scanning parameters set using improved adaptive threshold segmentation algorithm for (a) 20 mA, slice thickness of 5.0 mm and (b) 250 mA, slice thickness 2.5 mm.
3.3. ACV variation with reconstruction time After image segmentation, the ACV of the lung in each CT image is calculated by Eq. (2). The duration of Cine scanning at a couch position is 6 sec and eight slices of CT image can be reconstructed for each acquisition with 2.5 mm slice spacing. Also, 12 reconstructions can be performed for a Cine scan at each couch position. The total ACV of the lung on the CT images of 8 slices reconstructed at the same time can be calculated. Then, lung ACV against reconstruction time
90
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
(sequence number of CT reconstruction) for different couch positions are shown in Fig. 3, from which the breathing cycle and phase can be identified.
Fig. 3. After the Cine scanning for each couch position with a duration of 6 sec, a total of 12 CT reconstructions were performed, where eight slices can be acquired per reconstruction. The lung ACV of each reconstruction periodically changes with the sequence number of CT reconstruction at a couch position.
3.4. 4D-CT reconstruction Fig. 3 shows that all of the axial CT images with the same phase for all couch positions were aligned forming a single-phase 4D-CT volume, and all of which contain a complete 4D-CT series of a subject. Subsequently, the coronal and sagittal 4D-CT data sets can be obtained from the axial 4D-CT series. For example, a set of 4D-CT images with 10 different phases in a complete respiratory cycle is shown in Fig. 4 (coronal section) and Fig. 5 (sagittal section). It is clear that the locations of diaphragmatic dome and tumor, which are relative to the double transverse lines in the figures, change with the breathing cycle periodically.
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
91
Fig. 4. Images in the coronal view of 10 phases of 4D-CT reconstructed in the study. From left to right, top to bottom, the breathing state varies from end-inhale (phase 1) to end-exhale (phase 6) and back to end-inhale (phase 10). The two lines were added to show the relative amplitudes of the tumor and the top of the diaphragm moving with the breathing cycle.
Fig. 5. Images in the sagittal view of 10 phases of 4D-CT reconstructed in the study, which is based on the ACV principle. From left to right and top to bottom, the breathing state varies from end-inhale (phase 1) to end-exhale (phase 6) and back to end-inhale (phase 10). The two lines were added to show the relative amplitudes of the tumor and the top of the diaphragm moving with the breathing cycle.
92
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
4. Discussion The highlight of 4D-CT reconstruction based on the ACV principle are the simplicity, easiness and rapid imaging, which completely overcomes the dependence on the external respiration monitoring system, such as RPM, pressure sensors and spirometers. This can be achieved in all conventional CT. Compared with previous iteration model reconstruction, which typically requires several hours [10-11], 4D-CT reconstruction has the advantage of being much more efficient. Of course, like other methods, it requires the patients to breathe smoothly, naturally and regularly during the entire process of image acquisition. If the patient breathes irregularly during data acquisition, the motion artifacts are very obvious in the reconstructed 4D-CT series and may cause 4D-CT reconstruction to fail. For this reason, to enhance the quality of 4D-CT reconstruction, patient breathing training before data acquisition must be performed. Similar with other 4D-CT reconstruction methods [12], the 4D-CT reconstruction based on the ACV principle is also sensitive to the noises. To investigate the optimal scanning parameter set, the reconstruction results in different scanning parameters are compared. For instance, the milliamperes were set to 20, 50, 100 and 200 mA, and the slice thicknesses were set to 0.6, 1.25, 2.5 and 5 mm. In this study, it was found that the intrinsic noise was evident and the image was blurred when using a low milliampere scanning (Fig. 6). When liminal value segmentation was performed on this kind of CT image, it is difficult to segment the lung, muscle, skin, clothes and sheets from each other, which have been blurred. At the same time, noise may completely overwhelm the change of pulmonary CT value caused by the variation of air content in the process of breathing, which makes the ACV-time relation change irregularly leading to a failure of 4D-CT reconstruction. The results suggested that to ensure the normal reconstruction of 4D-CT and its quality, the signal-to-noise ratio should be increased as high as possible during CT image acquisition. The number of CT milliampere should not be lower than 100 mA and should reach about 200 mA. Meanwhile, the slice thickness is also suitable at 2.5 mm (Fig. 2). When the slice thickness is < 1.25 mm, the images are of poor quality and the noise influence is substantial. When the slice thickness is > 5 mm, although image quality is high and noise influence is slight, the longitudinal spatial resolution is low. For example, Fig. 6 illustrates the effect of noise on lung tissue segmentation of different scanning conditions.
(a) 200 mA, slice thickness 1.25 mm
(b) 50 mA, slice thickness 1.25 mm
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
(c) 100 mA, slice thickness 1.25 mm
(d) 200 mA, slice thickness 0.6 mm
(e) 200 mA, slice thickness 2.5 mm
(f) 20 mA, slice thickness 2.5 mm
93
Fig. 6. The effect of noise on the lung tissue segmentation of different scanning parameter sets. When the slice thickness is ≤ 1.25 mm (a, b, c, and d), it is difficult to distinguish the lung, muscle, skin, clothes and sheets, which have been blurred. At the same time, the noise may completely overwhelm the intensity variation of pulmonary CT caused by the variation of air content in the process of breathing. When the slice thickness is 2.5 mm and the milliampere is ≤ 200 mA (e, f), image quality is better than that with lower slice thickness, and the effect of noise is minimal.
5. Conclusion In conclusion, based on the ACV principle, 4D-CT reconstruction is convenient and reliable and it eliminates the need of external respiratory monitoring equipment. It is not constrained by the software and hardware of a CT scanner. The thermoplastic mask used to fix patients receiving radiotherapy has no effect on it. Therefore, it is an accurate, cost-efficient yet simple method, which can be readily implemented in any conventional multi-slice CT scanner.
94
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
6. Acknowledgements This work is partially supported by the National Natural Science Foundation of China (Grant No. 81170078), Guangdong Province Science and Technology Agency Grant (Grant No. 2011B031800111) and Guangzhou Municipal Science and Technology Agency Grant (Grant No. 2011J4300131). References [1]
Low DA, Nystrom M, Kalinin E, Parikh P, Dempsey JF, Bradley JD, et al. A method for the reconstruction of four-dimensional synchronized CT scans acquired during free breathing. Med Phys. 2003; 30(6): 1254-1263. [2] Keall P J, Starkschall G, Shukla H, Forster K M, Ortiz V, Stevens C W, et al. Acquiring 4D thoracic CT scans using a multislice helical method. Phys Med Biol. 2004; 49 2053–2067. [3] Pan T, Lee TY, Rietzel E, Chen GT. 4D-CT imaging of a volume influenced by respiratory motion on multi-slice CT. Med Phys. 2004; 31(2): 333-340. [4] Rietzel E, Pan T, Chen GT. Four-dimensional computed tomography: image formation and clinical protocol. Med Phys. 2005; 32(4): 874-889. [5] Berlinge K, Sauer O, Vences L. A simple method for labeling CT images with respiratory states. Med Phys. 2006; 33(9): 3144-3148. [6] Kleshneva T, Muzik J, Alber M. An algorithm for automatic determination of the respiratory phases in our-dimensional computed tomography. Phys Med Biol. 2006; 51(16):269–276 [7] http://www.dir-lab.com/Downloads.html, 2013-06-16. [8] Zhang SX, Zhou LH, Chen GJ, Lin SQ,Ye YS, Zhang HN. Four-dimensional computerized tomography (4D-CT) reconstruction based on the similarity of spatial adjacent images. Chin J Biomed Eng (Engl Ed). 2008; 17(3): 106-113. [9] Jia T, Meng L ,Zhao DZ, Wang X. Automatic Lung Parenchyma Segmentation on CT Image. J Northeastern Univ (Nat. Sci.). 2008; 29(7): 965-967,975. [10] McClelland JR, Blackall JM, Tarte S, Chandler AC, Hughes S, Ahmad S, et al. A continuous 4D motion model from multiple respiratory cycles for use in lung radiotherapy. Med Phys. 2006; 33(9): 3348-3358. [11] Zeng R, Fessler JA, Balter JM, Balter PA. Iterative sorting for 4D CT images based on internal anatomy motion. Med Phys. 2008; 35(3):917-26. [12] Yamamoto T, Langner U, Loo BW Jr, Shen J, Keall PJ. Retrospective analysis of artifacts in four-dimensional CT images of 50 abdominal and thoracic radiotherapy patients. Int J Radiat Oncol Biol Phys. 2008; 72 (4):1250–1258.
Copyright of Bio-Medical Materials & Engineering is the property of IOS Press and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
85
4D-CT reconstruction based on pulmonary average CT values ZHANG Shu-xua,* , ZHOU Ling-hongb, LIN Sheng-qua, YU Huia, ZHANG Guo-quana, WANG Rui-haoa and QI Binga a
Radiotherapy Center, Affiliated Tumor Hospital of Guangzhou Medical University, Guangzhou 510095, China b Department of Biomedical Engineering, Southern Medical University, Guangzhou 510095, China
Abstract. To date, commercial 4D-CT systems typically depend on an external respiratory monitoring device. Immobilizing patients in a thermoplastic mask while receiving radiotherapy may result in a failure of 4D-CT reconstruction. The aim of this study is to investigate the feasibility of 4D-CT reconstruction based on a method using pulmonary average CT values (ACV) without an external respiratory monitoring device. The ACV of the whole lung assumes cyclical variation during respiration. Phases of CT images were identified by calculating the ACV over time. Subsequently, five sets of 4D-CT images based on a Real-time Position Management (RPM) system were selected to verify the ACV method. The entire lung CT datasets of another sixteen free-breathing patients were acquired in Cine scan mode for multiple couch positions. The phase of every CT image was identified and re-sorted into different phase 4D-CT volumes by analyzing the time dependence of the corresponding ACVs. This paper demonstrates the ACV method using the 4D-CT data sets based on the RPM system. Convenient and reliable 4D-CT reconstruction can be accomplished without any external respiratory monitoring device using ACVs. Key words: 4D-CT; image reconstruction; average CT value; respiration phase
1. Introduction Recently, the use of imaging techniques to improve treatment planning and delivery for patients undergoing radiotherapy has generated a number of divergent four-dimensional computed tomography (4D-CT) resorting methods, e.g. [1-5]. Compared with conventional free-breathing CT scans, 4D-CT can not only reduce motion artifacts but also represent the motion states of internal organs. Existing 4D-CT systems are typically based on the Real-time Position Management (RPM) system (Varian Medical Systems, Inc., Palo Alto, CA, USA) to record respiratory cycle information [2-4]. During the image acquisition, the RPM system correlates the patient's breathing cycle with CT imaging. After scanning, all CT images are grouped and sorted according to their corresponding phases to obtain eight to 10 series of complete CT data sets. Each is equivalent to a static ordinary CT scan taken at a particular point in the respiratory cycle. All series together reflect the movements and changes of internal organs across the respiratory cycle. The system is then used to track the motion amplitude of a reflective marker placed on the patient’s abdomen during treatment, which is converted *
Address for correspondence: Professor ZHANG Shu-xu, Radiotherapy Center, Affiliated Tumor Hospital of Guangzhou Medical University. NO.78 Hengzhigang Road, Guangzhou, 510095, China. Tel: +86-13005110376; Fax: +86-2066673635; E-mail: [email protected] 0959-2989/14/$27.50 © 2014 – IOS Press and the authors. All rights reserved
86
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
into the corresponding respiratory breathing phase so that delivery can be gated to a fraction of the respiratory cycle. Other types of devices have also been used to acquire an external respiratory cycle signal in other 4D-CT reconstruction methods: a spirometer to measure the tidal volume through the mouth [1], and elastic belts to measure pressure around the abdomen [6]. However, respiratory cycle signals obtained from these external surrogates may not always accurately represent internal organ motion, especially when the breathing patterns become irregular. In addition, when a patient receiving radiation therapy is immobilized in a thermoplastic mask, the external respiratory monitoring device mentioned above cannot measure the respiratory cycle signal, which may result in the failure of 4D-CT reconstruction and delay or prevent treatment. Therefore, a novel 4D-CT reconstruction method based on pulmonary average CT values (ACVs) without any external respiratory monitoring device is proposed and demonstrated in this study. 2. Materials and methods 2.1. ACV method and verification According to the principle of CT imaging, the CT value of tissue X is computed by the following equation: HU(x) = 1000 × (µx-µw) / µw,
(1)
where HU(x) is the CT value of tissue X, µx is the attenuation coefficient of the tissue X, and µw is the attenuation coefficient of water. The attenuation coefficient of air is defined as 0. Thus, the CT values of water and air is 0 (±10) and −1000 HU, respectively. During expiration, the decrease of air volume in the lung leads to an increase of the average CT value, vice versa. Thus, it seems reasonable to presume that the ACV of the whole lung changes periodically. If the breathing cycle begins from the end-exhale to end-inhale and then return to end-exhale, the air content of lung therefore changes from the minimum to the maximum and return to the minimum. The plot of ACV of the lung against time will appear like the shape of the letter “V”, and can be used as the respiratory signal. The respiratory cycle and phase of the CT image can then be identified by the ACV-time plot. This method is named the “ACV principle”. To verify this method, the 4D-CT datasets of five patients were analyzed and reconstructed using the RPM system [7]. The lung ACVs of each 4D-CT volume at different phases were calculated (Table 1) and the ACV-phase graphs were plotted (Fig. 1). 2.2. Image acquisition The existing RPM 4D-CT reconstruction method is under the assumption of a repeated respiratory cycle and amplitude of the scanned object throughout treatment. Without this hypothesis, the acquired 4D-CT images cannot be reconstructed. The methods of CT image acquisition and reconstruction was previously reported [8], which are simply stated as follows: Before CT data acquisition, patients underwent breathing training to improve the reproducibility of respiration. The entire thorax and upper abdomen of patients with free breathing were acquired in Cine mode on a 16-slice GE LightSpeed CT scanner (GE Medical Systems, Waukesha, WI, USA). The imaging parameters were set as follows: 120 kVp, 0.5 sec gantry rotation,
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
87
0.45 sec Cine interval, 512 × 512 matrix with 2.5 mm slice thickness. The duration of a Cine scan at a couch position was slightly greater than or equal to the average respiratory cycle plus 1 sec, which was set as 6 sec in this study. A volume of 2 cm in length along the cranial-caudal direction was imaged with 12 reconstructions in each Cine scan and eight layers for each reconstruction with 2.5 mm in slice spacing. The CT acquisition was turned off when a couch position Cine scan was completed. Then, the table was moved to the next position and another Cine scan was launched. The acquisition was repeated until the entire thorax and upper abdomen volume was fully acquired. 2.3. Image grouping and segmentation Based on the ACV principle, a 4D-CT reconstruction software (named ACV 4D-CT) was developed using VC++ language based on the Windows platform. The acquired CT images were read into the system and grouped according to the couch position and reconstruction order in the Cine scan. Then, the lung tissues were segmented from the CT images using an improved adaptive threshold segmentation algorithm [9]. As a result, the lung tumor focus was retained, the regions of trachea and main bronchi were removed, and the lung edges were smoothed. When no abnormal lesion was discovered in the inspected patients, the segmentation result was considered to be ideal. When the peripheral lung cancer adhered to the chest wall or the central-type lung cancer adhered to the mediastinum, artificial auxiliary segmentation was incorporated. As a result of the image segmentation, only the transaxial section of lung tissue was present, while other pixels corresponding to the couch, clothing, bed sheet, immobilization device and muscles were removed. Finally, the results were reevaluated by experienced physicians of imaging diagnosis. 2.4. Computing of pulmonary ACV The ACV of lung in each CT image is computed by the following equation: ACV(k) = ΣHU(i, j) / N,
(2)
where k indicates the slice number at a reconstruction of Cine scan, HU (i, j) is the CT value of the pixel at row i and column j of a lung CT image, and N is the total number of the lung pixels. At each couch position, the duration of a Cine scan is 6 sec, and a total of 12 reconstructions were performed. The ACVs for eight slice images of each reconstruction were calculated, and the ACV variation with the reconstruction time sequence is shown in Figure 3. 2.5. Image sorting and 4D-CT reconstruction When the patients were breathing regularly, after Cine scanning the relationship between lung ACV versus the reconstruction time is very similar among different couch positions. The starting time of Cine scanning is random at different couch position, so the initial phase of the ACV-time graph is not the same at different couch positions. The time cost of the data acquisition is only 1-2 min. If the patient had received breathing training, the respiratory amplitude and frequency is assumed to be constant. The phase of each CT image can be identified according to the ACV-time graph. Sorting by phase, multiple CT series with different phases can be obtained and are named 4D-CT data sets, which cover the whole lung. Every single-phase 4D-CT data set is equivalent to a static ordinary CT scan taken at a particular point in the respiratory cycle. After the axial 4D-CT data sets are acquired, the coronal, sagittal section of the 4D-CT data sets can be reconstructed based on the axial 4D-CT.
88
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
3. Results 3.1. ACVs of 4D-CT change with breathing phase Table 1 lists the entire lung ACV of single-phase 4D-CT data set for the five subjects whose 4D-CT were reconstructed with an RPM system [7]. For each subject, the ACV of lung changes with phase from a lower value, which is corresponding to the end-inhale with maximum content of air, to a larger value (end-exhale) and return to a lower value (end-inhale again). To intuitively demonstrate the regularity of the ACVs of the whole lung changing with the respiratory phase, the entire lung ACV versus image phase is shown in Figure 1 for each patient based on the data mentioned in Table 1. It demonstrates that the whole lung ACVs of each patient present a cyclic variation during respiration. From end-inhale to exhale and back to inhale, a sound change of ACV over the respiratory phase was revealed. Table 1 The entire lung ACV of different phase 4D-CT data of five patients, whose 4D-CT reconstructions are based on the RPM system and Cine model scan. Entire Lung ACV of different phase 4D-CT volume (HU) Phase Patient No.1 Patient No.2 Patient No.3 -674 -817 -688 1 -667 -814 -676 2 -646 -800 -652 3 -628 -795 -645 4 -608 -792 -644 5 -598 -785 -627 6 -597 -776 -632 7 -620 -777 -646 8 -641 -794 -661 9 -668 -807 -676 10
Patient No.4 -765 -755 -752 -741 -726 -720 -723 -730 -751 -763
Patient No.5 -770 -757 -742 -735 -723 -714 -710 -728 -744 -763
Fig. 1. The graphs of the whole lung ACV-phase of five patients, whose 4D-CT reconstructions are based on the RPM system, demonstrate that the whole lung ACVs present periodic variation during respiration process. When breathing motion changes from end-inhale (phase 1) to end-exhale (phase 6/7/8), the air content of the lung decreases, which results in an increase in the ACVs. When breathing motion is from end-exhale to end-inhale (phase 10), the air content of the lung increases leading to a decrease in the ACVs.
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
89
3.2. Lung tissue segmentation With proper acquisition parameters and image segmentation methods, the lung tissue on a transaxial section CT image can be completely segmented as mentioned in Section 2.3 and shown in Fig. 2. Then the segmented lung tissue image can be used to compute the ACV of the lung.
Fig. 2. The lung tissues were segmented from the CT images acquired with proper scanning parameters set using improved adaptive threshold segmentation algorithm for (a) 20 mA, slice thickness of 5.0 mm and (b) 250 mA, slice thickness 2.5 mm.
3.3. ACV variation with reconstruction time After image segmentation, the ACV of the lung in each CT image is calculated by Eq. (2). The duration of Cine scanning at a couch position is 6 sec and eight slices of CT image can be reconstructed for each acquisition with 2.5 mm slice spacing. Also, 12 reconstructions can be performed for a Cine scan at each couch position. The total ACV of the lung on the CT images of 8 slices reconstructed at the same time can be calculated. Then, lung ACV against reconstruction time
90
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
(sequence number of CT reconstruction) for different couch positions are shown in Fig. 3, from which the breathing cycle and phase can be identified.
Fig. 3. After the Cine scanning for each couch position with a duration of 6 sec, a total of 12 CT reconstructions were performed, where eight slices can be acquired per reconstruction. The lung ACV of each reconstruction periodically changes with the sequence number of CT reconstruction at a couch position.
3.4. 4D-CT reconstruction Fig. 3 shows that all of the axial CT images with the same phase for all couch positions were aligned forming a single-phase 4D-CT volume, and all of which contain a complete 4D-CT series of a subject. Subsequently, the coronal and sagittal 4D-CT data sets can be obtained from the axial 4D-CT series. For example, a set of 4D-CT images with 10 different phases in a complete respiratory cycle is shown in Fig. 4 (coronal section) and Fig. 5 (sagittal section). It is clear that the locations of diaphragmatic dome and tumor, which are relative to the double transverse lines in the figures, change with the breathing cycle periodically.
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
91
Fig. 4. Images in the coronal view of 10 phases of 4D-CT reconstructed in the study. From left to right, top to bottom, the breathing state varies from end-inhale (phase 1) to end-exhale (phase 6) and back to end-inhale (phase 10). The two lines were added to show the relative amplitudes of the tumor and the top of the diaphragm moving with the breathing cycle.
Fig. 5. Images in the sagittal view of 10 phases of 4D-CT reconstructed in the study, which is based on the ACV principle. From left to right and top to bottom, the breathing state varies from end-inhale (phase 1) to end-exhale (phase 6) and back to end-inhale (phase 10). The two lines were added to show the relative amplitudes of the tumor and the top of the diaphragm moving with the breathing cycle.
92
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
4. Discussion The highlight of 4D-CT reconstruction based on the ACV principle are the simplicity, easiness and rapid imaging, which completely overcomes the dependence on the external respiration monitoring system, such as RPM, pressure sensors and spirometers. This can be achieved in all conventional CT. Compared with previous iteration model reconstruction, which typically requires several hours [10-11], 4D-CT reconstruction has the advantage of being much more efficient. Of course, like other methods, it requires the patients to breathe smoothly, naturally and regularly during the entire process of image acquisition. If the patient breathes irregularly during data acquisition, the motion artifacts are very obvious in the reconstructed 4D-CT series and may cause 4D-CT reconstruction to fail. For this reason, to enhance the quality of 4D-CT reconstruction, patient breathing training before data acquisition must be performed. Similar with other 4D-CT reconstruction methods [12], the 4D-CT reconstruction based on the ACV principle is also sensitive to the noises. To investigate the optimal scanning parameter set, the reconstruction results in different scanning parameters are compared. For instance, the milliamperes were set to 20, 50, 100 and 200 mA, and the slice thicknesses were set to 0.6, 1.25, 2.5 and 5 mm. In this study, it was found that the intrinsic noise was evident and the image was blurred when using a low milliampere scanning (Fig. 6). When liminal value segmentation was performed on this kind of CT image, it is difficult to segment the lung, muscle, skin, clothes and sheets from each other, which have been blurred. At the same time, noise may completely overwhelm the change of pulmonary CT value caused by the variation of air content in the process of breathing, which makes the ACV-time relation change irregularly leading to a failure of 4D-CT reconstruction. The results suggested that to ensure the normal reconstruction of 4D-CT and its quality, the signal-to-noise ratio should be increased as high as possible during CT image acquisition. The number of CT milliampere should not be lower than 100 mA and should reach about 200 mA. Meanwhile, the slice thickness is also suitable at 2.5 mm (Fig. 2). When the slice thickness is < 1.25 mm, the images are of poor quality and the noise influence is substantial. When the slice thickness is > 5 mm, although image quality is high and noise influence is slight, the longitudinal spatial resolution is low. For example, Fig. 6 illustrates the effect of noise on lung tissue segmentation of different scanning conditions.
(a) 200 mA, slice thickness 1.25 mm
(b) 50 mA, slice thickness 1.25 mm
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
(c) 100 mA, slice thickness 1.25 mm
(d) 200 mA, slice thickness 0.6 mm
(e) 200 mA, slice thickness 2.5 mm
(f) 20 mA, slice thickness 2.5 mm
93
Fig. 6. The effect of noise on the lung tissue segmentation of different scanning parameter sets. When the slice thickness is ≤ 1.25 mm (a, b, c, and d), it is difficult to distinguish the lung, muscle, skin, clothes and sheets, which have been blurred. At the same time, the noise may completely overwhelm the intensity variation of pulmonary CT caused by the variation of air content in the process of breathing. When the slice thickness is 2.5 mm and the milliampere is ≤ 200 mA (e, f), image quality is better than that with lower slice thickness, and the effect of noise is minimal.
5. Conclusion In conclusion, based on the ACV principle, 4D-CT reconstruction is convenient and reliable and it eliminates the need of external respiratory monitoring equipment. It is not constrained by the software and hardware of a CT scanner. The thermoplastic mask used to fix patients receiving radiotherapy has no effect on it. Therefore, it is an accurate, cost-efficient yet simple method, which can be readily implemented in any conventional multi-slice CT scanner.
94
S.-x. Zhang et al. / 4D-CT reconstruction based on pulmonary average CT values
6. Acknowledgements This work is partially supported by the National Natural Science Foundation of China (Grant No. 81170078), Guangdong Province Science and Technology Agency Grant (Grant No. 2011B031800111) and Guangzhou Municipal Science and Technology Agency Grant (Grant No. 2011J4300131). References [1]
Low DA, Nystrom M, Kalinin E, Parikh P, Dempsey JF, Bradley JD, et al. A method for the reconstruction of four-dimensional synchronized CT scans acquired during free breathing. Med Phys. 2003; 30(6): 1254-1263. [2] Keall P J, Starkschall G, Shukla H, Forster K M, Ortiz V, Stevens C W, et al. Acquiring 4D thoracic CT scans using a multislice helical method. Phys Med Biol. 2004; 49 2053–2067. [3] Pan T, Lee TY, Rietzel E, Chen GT. 4D-CT imaging of a volume influenced by respiratory motion on multi-slice CT. Med Phys. 2004; 31(2): 333-340. [4] Rietzel E, Pan T, Chen GT. Four-dimensional computed tomography: image formation and clinical protocol. Med Phys. 2005; 32(4): 874-889. [5] Berlinge K, Sauer O, Vences L. A simple method for labeling CT images with respiratory states. Med Phys. 2006; 33(9): 3144-3148. [6] Kleshneva T, Muzik J, Alber M. An algorithm for automatic determination of the respiratory phases in our-dimensional computed tomography. Phys Med Biol. 2006; 51(16):269–276 [7] http://www.dir-lab.com/Downloads.html, 2013-06-16. [8] Zhang SX, Zhou LH, Chen GJ, Lin SQ,Ye YS, Zhang HN. Four-dimensional computerized tomography (4D-CT) reconstruction based on the similarity of spatial adjacent images. Chin J Biomed Eng (Engl Ed). 2008; 17(3): 106-113. [9] Jia T, Meng L ,Zhao DZ, Wang X. Automatic Lung Parenchyma Segmentation on CT Image. J Northeastern Univ (Nat. Sci.). 2008; 29(7): 965-967,975. [10] McClelland JR, Blackall JM, Tarte S, Chandler AC, Hughes S, Ahmad S, et al. A continuous 4D motion model from multiple respiratory cycles for use in lung radiotherapy. Med Phys. 2006; 33(9): 3348-3358. [11] Zeng R, Fessler JA, Balter JM, Balter PA. Iterative sorting for 4D CT images based on internal anatomy motion. Med Phys. 2008; 35(3):917-26. [12] Yamamoto T, Langner U, Loo BW Jr, Shen J, Keall PJ. Retrospective analysis of artifacts in four-dimensional CT images of 50 abdominal and thoracic radiotherapy patients. Int J Radiat Oncol Biol Phys. 2008; 72 (4):1250–1258.
Copyright of Bio-Medical Materials & Engineering is the property of IOS Press and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.
