| 1. Establishment of the optimal threshold segmentation of 18F—FDG—PET/CT image for accurate target volume delineation of lung cancer——a phantom studyPurpose: To establish the optimal threshold segmentation of 18F—FDG—PET/CT image for accurate target volume delineation of lung cancer with the PET/CT machine of our hospital by a phantom study, for that different size spheres filled of 18F-FDG solution suspended with different radioactivity in background with different radioactivity 18F-FDG solution were used to simulate the corresponding size and target to background(T/B)ratios of non-small cell lung cancer observed in our clinical practice.Materials and Methods: A NEMA IEC—Body Phantom SetTM was used to simulate the human thorax, 6 standard glass spheres with volumes of 0.52 ml,1.15 ml,2.6 ml,5.5 ml,11.4 ml,26.5ml(inner diameter of 10 ram, 13 mm, 17 mm, 22mm, 28mm, 37mm)were fixed on the top lid of the phantom through glass stems, and a standard plastic sphere with volume of 127ml (inner diameter of 62mm)prepared by ourselves were used to simulate the corresponding size of the NSCLC clinically observed. In order to acquire the high, medium and low T/B observed clinically ( 17.7:1,10.1:1,3.2:1), 18F-FDG solution of 0.563～3.572uCi/ml and 0.122～0.203uCi/ml were filled into all the spheres and the background region of the phantom respectively. All PET/CT images acquired with Siemens biograph 16 (HR) PET/CT scanner were transferred to Pinnacle 7.0g radiotherapy treatment planning system(TPS). Maximum mutual information automatic image fusion and manual image fusion were combined to achieve the image fusion of PET and CT scans. The fusion error was made sure to less than 1mm. The inner contour of all the spheres were delineated on CT images with window level 40Hu and window width 200Hu was regarded to represent the true volume of the spheres and used as a golden standard. The threshold of PET was defined as Ithreshold=Iback+X%Imax-back(slice), Ithreshold was the threshold radioactivity of single voxel, Iback was mean background radioactivity, X% Imax-back(slice) was a certain percent of the contrast of maximum radioactivity of each slice and the mean background radioactivity. Adjust the threshold of PET images to make the contour of all the spheres close to the inner contour delineated on CT images, the threshold was accepted as the optimal threshold as the two contours were closest. The volume and the spatial location of the contours delineated with the optimal threshold we have choosed were compared to that of the contours delineated with the threshold used world widely, which was defined as 42% of the maximum radioactivity in the whole target.Results: The distribution of the maximum radioactivity of each slice in a standard sphere was correlated with slice volume. In the slices which located in the middle half of the spheres with volume larger than 5.5ml, the contrast of maximum radioactivity and the mean background radioactivity for each slice were close to that of the slice with maximum volume, the mean discrepancy was—3.2±2.7%, and the optimal % Imax-back(slice) to acquire a volume closest to the true volume for each slice was close to that of the slice with maximum volume, the mean discrepancy was 2.0%±0.3%, the corresponding value of the slices located at the peripheral half of the spheres was—20.0±14.8% and 10.1%±1.7% (P＜0.05). For the spheres with volume larger than 5.5ml with different T/B, the optimal %Imax-back(slice) for the slices with maximum volume approach to 15%～26%, the discrepancy between the volume delineated and the true volume was≤5% as %Imax-back(slice) changed between 20～25%.Our phantom study results indicated that the optimal %Imax-back(slice) was a function of slice volume, the bigger the slice volume was the lower %Imax-back(slice) was required to acquire a slice volume closest to the true volume. For the spheres with same volume, the optimal %Imax-back(slice) for the slices with similar volume were higher in the low T/B group than that of medium or high T/B group (P＜0.03), and the optimal %Imax-back(slice) was close between the medium T/B and high T/B group (P＞0.05).For the the spheres with volume larger than 5.5ml with different T/B, as far as the coverage of true volume was concerned, the volumes delineated with the threshold segmentation of Iback+20% Imax-back(max) and Iback+20%Imax-back(slice) (for brevity sake, 20%max and 20%slice were named for the threshold segmentation respectively) seemed to be superior to those delineated with the threshold of 42%Imax(total) (42% was named for brevity). The threshold segmentation of 20%max and 20%slice resulted in delineated volumes larger than the true sphere volumes, whereas, issue was on the contrary for the volumes delineated with the method of 42% (P=0.000). There were no significant difference between the volumes delineated with the method of 20%max and 20%slice(P＞0.05). The mean volume of background which was involved in the target volume by mistake was 5.3±4.7 cm3 and 4.4±4.0 cm3 for the methods of 20%max and 20%slice respectively. There were no edge slices were missed for the method of 20%slice, whereas, the missing of edge slice occured in 42% spheres for the method of 20%max. For spheres with different size, the discrepancy between the volumes delineated by the threshold segmentation of 20%slice and the true volumes were 11.2%±3.0%,12.8%±3.8%和13.9%±5.2% for high, medium and low T/B respectively (P＞0.05)。Conclusions: For the PET/CT system of our hospital, the threshold segmentation of Iback+20%Imax-back(slice) seemed to be superior to the threshold segmentation of 42%Imax(total). It showed the potential of delineating target volume without the volume information provided by CT image, the influence of background radioactivity was taken into account by this method, the missing of edge slice might be avoid as well as. This method might result in a delineated volume larger than the true volume, but the volume of background which was involved in the target volume by mistake was very small due to the major enlarged volume located in the peripheral half slices of the spheres. Theoretically, it might be safe to be used for the gross target volume(GTV) delineation of NSCLC with atelectasis. As far as the coverage of the true target volume and the enlarged volume were concerned, there were no significant difference between the method of 20%max and 20%slice, but the missing of edge slice might be taken into account as the method of 20%max was used. The results of phantom study indicated that it might be helpful for decreasing the discrepancy between the delineated volume and the true target volume to improve the T/B ratio of targets with low T/B. Further investigation was required on this problem. 2. Impact of different threshold segmentation of 18F—FDG—PET/CT image for target volume delineation and radiotherapy treatment planning of NSCLCPurpose: To verify the accuracy of the optimal threshold segmentation of PET image which acquired from the former phantom study when it was used for the target volume delineation of NSCLC. To evaluate the possible influence of the optimal threshold we have established on the radiotherapy treatment planning by a dosimetry analysis.Materials and Methods: Patients with following conditions were selected: with NSCLC verified by cytological or histological proof, the tumor located in the upper lobe of the lung with clear margin on CT scans, without combination of atelectasis or post-obstructive pneumonitis, the ultimate respiratory movements at the craniocaudal and lateral direction were≤5mm, with no local region lymph node metastasis and remote metastasis. The patients were fasted over 8h before PET/CT scanning. PET/CT images were acquired at 1h and 2h after the intravenous injection of 6.7～7.4MBq/Kg 18F-FDG respectively. Both PET/CT images for each patient were transferred to Pinnacle 7.0g Radiotherapy treatment planning system(TPS). Maximum mutual information automatic image fusion and manual image fusion were combined to achieve the image fusion of PET and CT scan. The fusion error was made sure to less than 1.5mm. After the fusion of PET and CT image was done well, the gross target volume (GTV) of each was delineated on CT image with the 'lung window setting' (window level -700Hu, window width 700Hu), which was named GTVCT. Then, GTV was delineated on 1h and 2h PET image with three different threshold segmentation(42 % Imax(total), Iback+20%Imax-back(max) and Iback+20% Imax-back(slice), for brevity sake, 42%, 20%max and 20%slice was named to represent the three methods respectively), the corresponding GTV was named as GTV42%,GTV20%max and GTV20%slice. Not only the volume discrepancy between GTV42%, GTV20%max, GTV20%slice and GTVCT, but also the spacial coincidence with GTVCT and the coverage over GTVCT for each GTV was compared. The above compare was also done between the 1h and 2h PET image of each patient. A three dimensional margin of 1cm were added for GTVCT, GTV42%, GTV20%max and GTV20%slice of 2h PET/CT image of each patient to form corresponding PTVCT, PTV42%, PTV20%max and PTV20%slice. Four radiotherapy treatment plan were designed based on PTVCT, PTV42%, PTV20%max and PTV20%slice respectively for each patient by a physicist. All PTVs were delivered dose of 6600 cGy in 33 fractions in 6.6 weeks. Both the volume within PTVCT accepted dose less than 95% of prescribed dose and the lung V20 (the percent of the volume accept dose≥20Gy within total normal lung volume) were compared between the four plans based on different PTVs. Both TCP(tumor control probability) and lung NTCP(normal tissue complication probability, here the occur rate of radiation induced pneumonitis was regard as a endpoint) were compared within the treatment plans based on different PTVs.Results: 2006. 11～2007. 3, 8 patients were involved in our study, the median tumor volume was 40.1cm3 (5.01～95.34cm3), the maximum transverse diameter was 2.1～5.7cm. GTV42% was not large enough to cover the whole GTVCT was observed in all the patient. Even in slices which located in the middle 1/2 of the maximum diameter at the scanning axial direction of the tumor, it covered only 51.4% of GTVCT. As far as the coverage for GTV was concerned, GTV20%max and GTV20%slice were superior to GTV42% (P＜0.01). For the peripheral 1/2 slices of the tumor, GTV20%slice covered more GTVCT than GTV20%max did (P＜0.05). But for the middle 1/2 slices, there was no significant difference of the coverage on GTVCT between GTV20%slice and GTV20%max (P＜0.01). In all patients, T/B increased with the delay of PET scan time. It increased from 13.9±5.4 for 1h PET image to 17.6±6.2 for 2h image (P＜0.01). There were no significant difference of the GTVs delineated by the same method between 1h and 2h PET/CT image. All PTV42% were smaller than PTVCT, the mean discrepancy was 37.8%. PTV20%slice seemed to be closest to PTVCT, the mean enlarge volume was 2.0% of PTVCT(P＜0.01). When the treatment plan based on different PTV were taken into account, there was 8% volume within PTVCT accepted dose less than 95% of prescribed dose in 42% group, thus might induce 1% decrease of TCP. Whereas, there were only 1.4% and 0.01% volume within PTVCT accepted dose less than 95% of prescribed dose for 20%max group and 20%slice group respectively. That didn't influence TCP significantly. Both 20%max group and 20%slice group were superior to 42% group(P＜0.01). The lung V20 and lung NTCP showed no significant difference within 20%max group, 20%slice group and PTVCT group. For a special case with tumor with large volume of necrosis, neither GTV delineated by the three method could cover the GTVCT entirely, GTV20%slice seemed to be closest to the GTVCT.Conclusions: Because of possibly more complete tumor coverage, for the PET/CT system of our hospital, the threshold segmentation of Iback+20%Imax-back(slice) might be more suitable for target volume delineation of the majority of lung caner than the threshold segmentation of 42%Imax(total) and Iback+20%Imax-back(max). It might be recommend to use for the target volume delineation of NSCLC with atelectasis in corresponding clinical research. The accuracy of the threshold segmentation of Iback+20%Imax-back(slice) required further investigation as it was used for the delineation of tumors with great inhomogeneous of glucose metabolism.
|
PDF Full Text Request