Respiratory-gated 18F-FDG PET imaging in lung cancer: effects on sensitivity and specificity

Abstract

Background

Respiratory motion is known to deteriorate positron emission tomography (PET) images and may lead to potential diagnostic errors when a standardized uptake value (SUV) cut-off threshold is used to discriminate between benign and malignant lesions.

Purpose

To evaluate and compare ungated and respiratory-gated 18F-fluorodeoxyglucose PET/computed tomography (CT) methods for the characterization of pulmonary nodules.

Material and Methods

The list-mode acquisition during respiratory-gated PET was combined with a short breath-hold CT scan to form the CT-based images. We studied 48 lesions in 43 patients. PET images were analyzed in terms of the maximum SUV (SUV_max) and the lesion location.

Results

Using receiver-operating characteristic (ROC) curves, the optimal SUV cut-off thresholds for the ungated and CT-based methods were calculated to be 2.0 and 2.2, respectively. The corresponding sensitivity values were 83% and 92%, respectively, with a specificity of 67% for both methods. The two methods gave equivalent performance levels for the upper and middle lobes (sensitivity 93%, specificity 62%). They differed for the lower lobes, where the CT-based method outperformed the ungated method (sensitivity values of 90% and 70%, respectively, and a specificity of 73% with both methods) – especially for lesions smaller than 15 mm.

Conclusion

The CT-based method increased sensitivity and did not diminish specificity, compared with the ungated method. It was more efficient than the ungated method for imaging the lower lobes and smallest lesions, which are most affected by respiratory motion.

Keywords

PET fluorine-18 fluorodeoxyglucose respiratory motion lung cancer

One of the most frequent causes of death in industrialized countries is primary or secondary malignant lung disease. This is due in part to a poor prognosis, since screening is not powerful (1). 18F-fluorodeoxyglucose (18F-FDG) positron emission tomography (PET) is an imaging modality in which lung lesions are considered in qualitative terms (e.g. visual aspect, relative to structures such as the mediastinum) and/or semi-quantitative terms (e.g. the standardized uptake value [SUV]) (2). A number of research groups have assessed the use of a semi-quantitative index (typically the maximum SUV [SUV_max]) to define a cut-off threshold which can discriminate between benign and malignant pulmonary nodules. Most recent studies on this topic have shown sensitivities ranging from 58% to 89% and specificities of between 77% and 93% (3–8).

Standard PET acquisitions last typically several minutes, during which anatomical structures are subjected to physiological motion. Furthermore, CT is usually performed in a free-breathing mode and can therefore be altered by respiratory motion artifacts (especially near the diaphragm) (9, 10); hence, the respiratory state at the time of the CT acquisition is unknown. All these phenomena spread out (11) and lower the intensity distribution (12) in the reconstructed PET images and lead to poorly co-registered PET vs. CT images (13). These effects are especially significant for small foci (4, 14). Thus, quantification is altered (i.e. the SUV may be erroneous) (15) and the uptake foci's location on morphological images is imprecise (16). Only a study by Garcia Vicente et al. has taken into account the respiratory motion issue when computing sensitivity/specificity for pulmonary lesions (17). However, the latter study involved a phase-based respiratory motion compensation technique. This kind of respiratory-gated PET technique (amplitude- or phase-based methods) divides the respiratory signal and the acquired PET data into several frames (18, 19). The main drawback of this strategy is that several PET volumes are attenuation-corrected with the same CT-derived attenuation map. Thus, the quantification problem is not solved and PET-to-CT lesion co-localization remains uncertain (18).

One solution to this issue is to synchronize both PET and CT scans (20, 21) but this requires identical breathing patterns during the two acquisitions (22) and exposes the patient to a much higher radiation dose.

Nehmeh et al. (23) suggested performing both PET and CT acquisitions during a deep-inspiration breath-hold. For the PET examination, 20-second episodes of apnea were repeated several times, so as to recover enough counting statistics. However, repetition of these apneic periods can be stressful and tiring – especially for dyspnoic patients.

We have suggested a ‘CT-based’ solution –a motion compensation technique in which a single PET volume is generated by selecting appropriate events from a respiratory-gated PET acquisition with respect to a single, short, breath-hold CT acquisition (BH-CT) (24). This method significantly improved PET-to-CT lesion co-localization and gave a significantly different SUV_max, relative to standard (‘ungated’) PET images, yet more accurate.

The aim of the present study was to assess use of the CT-based method to diagnose pulmonary nodules in a semi-quantitative analysis. With both ungated and CT-based images, we calculated the SUV threshold which gave the best sensitivity/specificity trade-off. The methods' respective sensitivity and specificity values were then compared. We also investigate the value of the CT-based method as a function of the lesion's location within the lung parenchyma.

Material and Methods

Patient data and preparation

Individuals with at least one pulmonary nodule (greatest dimension <35 mm) were referred to the Amiens University Hospital's Nuclear Medicine Department for malignancy assessment and were prospectively enrolled in the study. The characteristics of the 43 included patients are given in Table 1.

Table 1

Patient characteristics

Characteristic	Data
Gender (n)
Men	32
Women	11
Age (y) (mean; range)	65; 37–86
Body mass index (mean; range)	26; 18–34
Activity concentration (MBq/kg) (mean; range)	5.1; 4.5–5.7

The study was approved by the local independent ethics committee and all patients provided their written, informed consent.

Each patient was instructed to fast for at least 6 hours. Following verification of a normal glucose blood level, 226–519 MBq 18F-FDG was intravenously injected. After a 60-minute uptake phase in a quiet environment, patients underwent the PET/CT examination.

The final diagnosis of malignancy was based on a histological confirmation or imaging and clinical follow-up.

PET/CT acquisitions and image reconstructions

All acquisitions in this study were performed on a Hi-Rez Biograph^TM 6 PET/CT system (Siemens Medical Solutions, Erlangen, Germany) in 3-dimensional (3D) mode. Attenuation correction coefficients at 511 keV were calculated from CT acquisitions. Respiratory gating of PET acquisitions was performed using the AZ733V system (Anzai Medical Co. Ltd., Tokyo, Japan), which measures the abdominal wall pressure. After the examination, respiratory data were saved.

The study was performed as follows:

The ‘ungated’ method: whole-body, free-breathing CT (‘CT-std’, 110 kV, reference tube current: 85 mAs; pitch: 1) and standard, multistep PET (3 min per step) acquisitions.

The ‘CT-based’ method: a single-step, 10-min list-mode respiratory-gated PET acquisition (centered on the lesions previously visualized on the CT-std acquisition) and then by a BH-CT acquisition (110 kV, reference tube current: 50 mAs; pitch: 2) were added to the clinical protocol, with continuous recording of the respiratory signal (i.e. a respiratory-gated PET acquisition and a BH-CT). During the latter session, patients were asked to hold their breath at the normal end-expiration. This period of apnea defined a plateau on the respiratory curve and corresponded to a specific tissue position. In turn, the plateau enabled us to position an event selection range around it. The portions of the respiratory curve which fell within this event selection range defined the portions of the List Mode for binning into the CT-based sinogram. A more complete description has been published in (24). On average, the CT-based session was performed 16.6 min after the ungated acquisition of the same area.

All 3D data (i.e. ungated and CT-based sinograms) in this study were Fourier-rebinned into two dimensions and scatter-corrected. Ungated PET images were attenuation-corrected with a map derived from the CT-std volume, whereas CT-based PET attenuation correction was based on the BH-CT volume.

All PET volumes (i.e. for both ungated and CT-based acquisitions) were reconstructed using a two-dimensional attenuation-weighted ordered subsets expectation maximization algorithm with the following parameters: four iterations, eight ordered subsets in 168 × 168 × 81 matrices (4.06 mm × 4.06 mm × 2 mm). Lastly, a 3D isotropic Gaussian post-filter with a full width at half maximum of 5 mm was applied to each reconstructed volume.

Data analysis

For each lesion detected on CT volumes (i.e. CT-std and BH-CT), the clinician was asked to measure the lesion's greatest dimension on the BH-CT image and specify the lung lobe. Moreover, SUV_max were measured for both methods, inside volumes of interest (VOIs) drawn around the lesion on CT-std and BH-CT volumes and carried over onto ungated and CT-based PET volumes, respectively. VOIs were large enough to encompass the entire PET uptakes.

The sensitivity and specificity of the ungated and CT-based methods were calculated on the basis of histological reports or imaging and/or clinical follow-up. To compute the optimal SUV cut-off threshold of each method, receiver-operating characteristic (ROC) analyses were performed on the SUV_max of pulmonary uptakes for both ungated and CT-based PET images. We then calculated the vector norm between the ‘ideal point’ (i.e. 100% sensitivity and specificity) and each point on the curve. For each method, the SUV threshold giving the lowest vector norm was considered to be the optimal cut-off value.

First, all lesions (regardless of their location) were taken into account to compute the overall cut-off thresholds: SUV_threshold_ungated and SUV_threshold_CT-based.

Second, in order to evaluate the impact of the lesion's location, we also computed the sensitivity and specificity values (using to the above-defined thresholds) for lower lobe (LL) lesions on one hand and upper and middle lobe (UML) lesions on the other.

Lesions were also classified into two groups, according to their greatest dimension (less than or equal to 15 mm or greater than 15 mm). We chose this cut-off value because it corresponds to the limit of detection of our PET device (i.e. two to three times the spatial resolution) (25, 26). After assessing the two samples' comparability in terms of lesion size, we also compared the SUV_max measured on the ungated and CT-based volumes for LLs and for UMLs.

Results

Histological reports were obtained for 28 patients, including cases of non-small-cell lung cancer (n = 16, of which seven were adenocarcinomas), small-cell lung cancer (n = 1), metastases of various types (n = 4) and benign lesions (n = 7). For the 15 other patients, the final diagnosis was based on CT imaging and/or clinical follow-up (two malignant and 13 benign lesions). For these patients, the mean follow-up period was 142 days (range 84–240 days).

Overall, 48 lesions were studied. The lesion distribution within the lungs is shown in Fig. 1. Twenty-seven lesions were located in the UMLs (14 ≤15 mm; 13 >15 mm) and 21 were located in the LLs (12 ≤15 mm; 9 >15 mm).

Fig. 1

Lobe distribution of the lesions examined in the present study (whether malignant or benign)

An illustration of the CT-based method's influence on PET-to-CT lesion co-localization is presented in Fig. 2. On the ungated images (Fig. 2a), the uptake visible on PET was not in the same position as the pulmonary nodule seen in the CT-std scan. The CT-based method overcame the respiratory motion issue and enabled anatomical and functional images of the lesions to be matched accurately (Fig. 2b).

Fig. 2

Coronal 18F-FDG PET/CT images of a patient with a lesion on a CT scan, corresponding to a basaloid carcinoma: (a) free-breathing CT (left) and PET (right) images for the ungated method; (b) breath-hold CT (left) and PET (right) images for the CT-based method. The lesion identified on CT images was not visible in the same axial slice location on an ungated PET image, as shown by the solid lines. Application of the CT-based method resulted in accurate PET-to-CT lesion co-localization

For the sensitivity/specificity study, ROC curves for ungated and CT-based methods are presented in Fig. 3. The area under the curve for the ungated and the CT-based method was 0.828 and 0.823, respectively. The difference between the latter values was not statistically significant. Our analysis defined the optimal SUV_threshold_ungated

Fig. 3

Receiver-operating characteristic curves for the ungated method (left) and the CT-based method (right). The circles highlight the best sensitivity/specificity trade-off for each method

as 2.0 (vector norm = 37.27) and the optimal SUV_threshold_CT-based as 2.2 (vector norm = 35.60), as shown in Fig. 4. For all lesions (regardless of the location), these cut-off values yielded sensitivities of 83% and 92% for ungated and CT-based images, respectively, and a specificity of 67% for both methods. A McNemar test showed that the methods did not differ significantly in terms of sensitivity.

Fig. 4

Plots of the vector norms for the ungated method (left) and the CT-based method (right). The vector norm was defined as the distance between the upper left-hand corner (i.e. sensitivity and specificity = 100%) and each point on each receiver-operating characteristic curve. The optimal global thresholds for each method, (SUV_threshold_ungated and SUV_threshold_CT‐based) were defined as the smallest vector norm

By applying the above-defined thresholds (i.e. 2.0 for the ungated method and 2.2 for the CT-based method) to lesions localized in the UMLs, we obtained a sensitivity of 93% and a specificity of 62% for both methods (Table 2). For lesions in the LLs, the sensitivity and specificity were 70% and 73% for the ungated method and 90% and 73% for the CT-based method, respectively. Fig. 5 illustrates a case where, unlike ungated method, the CT-based method allowed to correctly classify the lesion. A melanoma metastasis is visible on the CT image (Fig. 5a, arrow). On the corresponding PET image, SUV_max was calculated to be 1.6; hence, the lesion was classified as ‘benign’ in a semi-quantitative analysis (i.e. as a false-negative). With the CT-based image (Fig. 5b), the same lesion gave an SUV_max of 2.7. This lesion was classified as malignant in an SUV-based interpretation (i.e. as a true-positive).

Fig. 5

18F-FDG PET/CT images of patient 41 with a suspect lesion on CT, corresponding to a melanoma metastasis (see arrows): (a) free-breathing CT (left) and PET (right) images for the ungated method; (b) breath-hold CT (left) and CT-based PET (right) images. The lesion was located in the lower lobe of the left lung (the lateral basal segment). On the ungated PET image, semi-quantitative analysis wrongly classified this lesion as benign. The CT-based method gave a higher value for the lesion's FDG uptake, so that SUV-based interpretation then described the lesion as a true-positive (malignant)

Table 2

Sensitivity/specificity results for upper and middle lobes and lower lobes calculated for each method, using the respective overall thresholds

	Upper and middle lobes		Lower lobes
	Ungated	CT-based	Ungated	CT-based
Threshold value	2.0	2.2	2.0	2.2
Sensitivity	93%	93%	70%	90%
Specificity	62%	62%	73%	73%

A Mann-Whitney test showed that there was no significant difference in lesion size between the UML and LL groups. The two methods yielded the same sensitivities/specificities for lesions >15 mm in the LLs and also in the UMLs (Table 3). For LL lesions smaller than or equal to 15 mm in size, the CT-based method yielded a higher sensitivity than the ungated method.

Table 3

Sensitivity/specificity results for upper and middle lobes and lower lobes calculated for each method, using the respective overall thresholds (SUV_threshold_ungated = 2.0 and SUV_threshold_CT-based = 2.2), as a function of the lesion size

		Upper and middle lobes		Lower lobes
		Ungated	CT-based	Ungated	CT-based
Size ≤15 mm	Sensitivity	75%	75%	40%	80%
	Specificity	80%	80%	86%	86%
Size >15 mm	Sensitivity	100%	100%	100%	100%
	Specificity	0%	0%	50%	50%

The normality test for SUV_max data in ungated and CT-based volumes for LLs and UMLs revealed that none of the series followed a normal distribution. Thus, a Wilcoxon test showed that there was significant difference between the SUV_max obtained with each method for LL lesions (P < 0.001) but not for UML lesions. Without taking account of the lesions' location, the mean (±SD) SUV_max for lesions smaller than or equal to 15 mm in size was significantly greater (P < 0.005 in a Wilcoxon test) for the CT-based method (2.3 ± 1.8) than for the ungated method (1.9 ± 1.3). For lesions greater than 15 mm in size, the mean (±SD) SUV_max were 5.2 ± 2.5 and 5.7 ± 2.5 for the ungated and CT-based methods, respectively. The difference between the latter two values was not statistically significant. The same conclusions were also drawn when we took the lesions' location into consideration.

Discussion

The best way to evaluate pulmonary lesions with 18F-FDG PET/CT is still subject to debate. Many research groups have suggested and evaluated an SUV threshold for classifying lung lesions as benign or malignant (3, 4, 6, 14). However, it was found that using the SUV_max as a critical decision criterion did not improve PET sensitivity/specificity.

The present study sought to compare the efficiency of semi-quantitative analysis on ungated PET images (which are currently used to diagnose pulmonary lesions) and images to which a particular type of respiratory motion compensation (our CT-based method (24)) had been applied. For each technique, we computed the sensitivity and specificity with several SUV thresholds (regardless of the lesions' locations, in an initial step) and extracted the optimal SUV cut-off from ROC curves.

The ROC curve analysis yielded a SUV_threshold_ungated of 2.0 – exactly the value obtained by Kim et al., who also worked on a lutetium oxyorthosilicate-based PET gantry (4). The CT-based method yielded an optimal SUV cut-off of 2.2. Application of the CT-based method gave a higher sensitivity than the ungated method (92% vs. 83%, respectively) but was just as specific (67% for both methods). Garcia Vicente et al. recently studied the impact of respiratory motion compensation on the assessment of malignancy in pulmonary nodules. They found that by using a SUV cut-off threshold of 2.5, the sensitivity increased from 0% to 52% and the specificity fell from 100% to 74% (compared with non-gated acquisitions) (17). These results can be explained by the fact that Garcia Vicente et al. only selected faint uptakes observed on non-gated images (i.e. uptakes with SUV_max < 2.5). This was not the case in our present study.

The decrease in specificity can be attributed to the behavior of FDG. Indeed, it is known that 18F-FDG does not accumulate only in tumor cells. Tissues affected by inflammation and infectious diseases are also 18F-FDG-avid and can generate false-positive lesions in both ungated and CT-based PET images. In the present study, eight lesions classified as malignant by both imaging methods were ultimately confirmed as benign by a histological report (n = 3) or clinical follow-up (n = 5). Three of these lesions were located in the UMLs and were greater than 15 mm in size.

Various histological and physical factors can induce false-negatives in semi-quantitative PET analyses. Indeed, some tumor types are known not to be intrinsically 18F-FDG-avid. This is the case for well- or moderately differentiated adenocarcinoma (6) and bronchioalveolar cell carcinoma (27). In the present study, two false-negative lesions (according to both the ungated and CT-based methods) were well-differentiated adenocarcinomas. In terms of physical effects, lesion size is a major limitation in PET imaging. Detecting small uptake foci (i.e. foci whose greatest dimension is less than two or three times the gantry's spatial resolution) is tricky, since it corresponds to the detection limit for a static target (26). This situation is accentuated if the small lesion is also subject to physiological motion. In the present study, lesions that were incorrectly described as benign by the ungated method were correctly classified as malignant by the CT-based method. These lesions measured between 12 and 14 mm in greatest dimension, which is close to the detection limit of our PET device. Removing the motion component by application of the CT-based method avoided the smearing effect (Fig. 2) and enabled us to detect smaller lesions than the ungated method did. In addition, the CT-based method ensures appropriate attenuation correction, yields a more reliable SUV and thus provides a more accurate semi-quantitative analysis (15).

The PET-to-CT lesion co-localization and attenuation correction improvements obtained by use of our CT-based method appeared to be greater for LL lesions than for UML lesions. Ungated and CT-based methods gave the same sensitivity/specificity results in the UMLs (Table 2). This is explained by the fact that the upper areas of the lungs are less subject to motion than the lower parts are (28). Moreover, Rodarte et al. (29) showed that even though the middle lobe (or lingula) extends toward the diaphragm, it is less subject to respiratory motion than the LLs. Indeed, the lung's oblique fissures favor inter-lobe sliding and rotations which mechanically compensate for the difference in motion magnitude between the UMLs and the LLs (30, 31). Consideration of the SUV_max values confirmed this assertion, since there was significant difference between SUV_max obtained with both methods for LL lesions but not for UML lesions. The CT-based method offers better prospects for accurate lesion detection in the LLs, since it removes the motion effects in PET and CT images. Indeed, the CT-based method is more sensitive than the ungated method for LL lesions smaller than 15 mm (Table 3) and is just as specific.

Both the ungated and CT-based methods displayed lower specificity values for lesions larger than 15 mm than for small lesions. This can be attributed not only to the non-specificity of 18F-FDG for tumor cells but also to the fact that large lesions are more easily detected with PET. This result is concordant with existing literature (32, 33).

Since conventional binning methods (based on a frequency- or amplitude-based analysis of the respiratory signal) provide numerous PET volumes for only one CT volume, they are still flawed by potentially inaccurate matching between CT and PET volumes and attenuation correction errors (18). To overcome this issue, Nehmeh et al. suggested a method based on a deep-inspiration CT acquisition and repeated deep-inspiration breath-hold PET acquisitions (each of which lasts 20 s) (23). This strategy may be very inconvenient for dyspnoic patients who have trouble holding their breath. Our CT-based method is easier for patients to tolerate, since the respiratory-gated PET acquisitions are performed in free-breathing mode. The only breath-hold sequence is the BH-CT, which lasts less than 10 s.

Our study had a number of limitations. Firstly, the small size of the study population may explain the lack of a statistically significant difference between the method's respective sensitivities. Secondly, a great number of lesions were classified as benign on the basis of clinical follow-up alone. These classifications must be interpreted with caution because of the short, ongoing follow-up period and the lack of histological evidence. Here, we presented the intermediate results from a wider study; our findings need to be confirmed in a larger population and with longer-term follow-up.

In conclusion, the results of the present study show that our CT-based gated PET method was more sensitive than the ungated method but just as specific for pulmonary nodules. We suggest that respiratory-gated PET acquisition should be performed when patients are addressed for the evaluation of LL lesions (especially when the lesion size is close to the PET gantry's detection limit).

Footnotes

Acknowledgements

The authors thank Dr Olivier Leleu from Abbeville Hospital for his contribution to this study.

They also thank the Nuclear Medicine Department's technical specialists for their advice and their valuable help with acquisition of the study data. They also wish to thank Dr David Fraser (Biotech Communication, Damery, France) for his helpful advice on the English language in this paper.

Conflict of interest: None.

References

Infante

, Pedersen

. Screening for lung cancer: are we there yet? Curr Opin Pulm Med 2010;16:301–6

Lowe

, Hoffman

, DeLong

, Semiquantitative and visual analysis of FDG-PET images in pulmonary abnormalities. J Nucl Med 1994;35:1771–6

Kagna

, Solomonov

, Keidar

, The value of FDG-PET/CT in assessing single pulmonary nodules in patients at high risk of lung cancer. Eur J Nucl Med Mol Imaging 2009;36:997–1004

Kim

, Allen-Auerbach

, Goldin

, Accuracy of PET/CT in Characterization of Solitary Pulmonary Lesions. J Nucl Med 2007;48:214–20

Bar-Shalom

, Kagna

, Israel

, Noninvasive diagnosis of solitary pulmonary lesions in cancer patients based on 2-fluoro-2-deoxy-D-glucose avidity on positron emission tomography/computed tomography. Cancer 2008;113:3213–21

Nomori

, Watanabe

, Ohtsuka

, Visual and semiquantitative analyses for F-18 fluorodeoxyglucose PET scanning in pulmonary nodules 1 cm to 3 cm in size. Ann Thorac Surg 2005;79:984–8; discussion 9

Sun

, Park

, Sheen

, Clinical usefulness of the fluorodeoxyglucose (FDG)-PET maximal standardized uptake value (SUV) in combination with CT features for the differentiation of adenocarcinoma with a bronchioloalveolar carcinoma from other subtypes of non-small cell lung cancers. Lung Cancer 2009;66:205–10

Schillaci

, Travascio

, Bolacchi

, Accuracy of early and delayed FDG PET-CT and of contrast-enhanced CT in the evaluation of lung nodules: a preliminary study on 30 patients. La Radiologia medica 2009;114:890–906

Beyer

, Antoch

, Blodgett

, Dual-modality PET/CT imaging: the effect of respiratory motion on combined image quality in clinical oncology. Eur J Nucl Med Mol Imaging 2003;30:588–96

10.

Goerres

, Burger

, Kamel

, Respiration-induced attenuation artifact at PET/CT: technical considerations. Radiology 2003;226:906–10

11.

Werner

, Parker

, Kolodny

, Respiratory gating enhances imaging of pulmonary nodules and measurement of tracer uptake in FDG PET/CT. Am J Roentgenol 2009;193:1640–5

12.

Lupi

, Zaroccolo

, Salgarello

, The effect of 18F-FDG-PET/CT respiratory gating on detected metabolic activity in lung lesions. Ann Nucl Med 2009;23:191–6

13.

Goerres

, Kamel

, Heidelberg

, PET-CT image co-registration in the thorax: influence of respiration. Eur J Nucl Med Mol Imaging 2002;29:351–60

14.

Hashimoto

, Tsujikawa

, Kondo

, Accuracy of PET for diagnosis of solid pulmonary lesions with 18F-FDG uptake below the standardized uptake value of 2.5. J Nucl Med 2006;47:426–31

15.

Daouk

, Fin

, Bailly

, Improved attenuation correction via appropriate selection of respiratory-correlated PET data. Comput Methods Programs Biomed 2008;92:90–8

16.

Osman

, Cohade

, Nakamoto

, Clinically significant inaccurate localization of lesions with PET/CT: frequency in 300 patients. J Nucl Med 2003;44:240–3

17.

Garcia Vicente

, Soriano Castrejon

, Talavera Rubio

, (18)F-FDG PET-CT respiratory gating in characterization of pulmonary lesions: approximation towards clinical indications. Ann Nucl Med 2010;24:207–14

18.

Daouk

, Fin

, Bailly

, Respiratory-Gated Positron Emission Tomography and Breath-Hold Computed Tomography Coupling to Reduce the Influence of Respiratory Motion: Methodology and Feasibility. Acta Radiol 2009;50:144–55

19.

Dawood

, Buther

, Lang

, Respiratory gating in positron emission tomography: a quantitative comparison of different gating schemes. Med Phys 2007;34:3067–76

20.

Nehmeh

, Erdi

, Pan

, Four-dimensional (4D) PET/CT imaging of the thorax. Med Phys 2004;31:3179–86

21.

Wolthaus

, van Herk

, Muller

, Fusion of respiration-correlated PET and CT scans: correlated lung tumour motion in anatomical and functional scans. Phys Med Biol 2005;50:1569–83

22.

Nehmeh

, Erdi

, Pan

, Quantitation of respiratory motion during 4D-PET/CT acquisition. Med Phys 2004;31:1333–8

23.

Nehmeh

, Erdi

, Meirelles

, Deep-inspiration breath-hold PET/CT of the thorax. J Nucl Med 2007;48:22–6

24.

Fin

, Daouk

, Morvan

, Initial clinical results for breath-hold CT-based processing of respiratory-gated PET acquisitions. Eur J Nucl Med Mol Imaging 2008;35:1971–80

25.

Brambilla

, Secco

, Dominietto

, Performance characteristics obtained for a new 3-dimensional lutetium oxyorthosilicate-based whole-body PET/CT scanner with the National Electrical Manufacturers Association NU 2-2001 standard. J Nucl Med 2005;46:2083–91

26.

Hoffman

, Huang

, Plummer

, Quantitation in positron emission computed tomography: 6. effect of nonuniform resolution. J Comput Assist Tomogr 1982;6:987–99

27.

Lee

, Kim

, Han

, Bronchioloalveolar carcinoma: clinical, histopathologic, and radiologic findings. Radiographics 1997;17:1345–57

28.

Wade

. Movements of the thoracic cage and diaphragm in respiration. J Physiol 1954;124:193–212

29.

Rodarte

, Hubmayr

, Stamenovic

, Regional lung strain in dogs during deflation from total lung capacity. J Appl Physiol 1985;58:164–72

30.

Hubmayr

, Rodarte

, Walters

, Regional ventilation during spontaneous breathing and mechanical ventilation in dogs. J Appl Physiol 1987;63:2467–75

31.

Ding

, Yin

, Cao

, Evaluation of lobar biomechanics during respiration using image registration. Med Image Comput Comput Assist Interv 2009;12:739–46

32.

Khalaf

, Abdel-Nabi

, Baker

, Relation between nodule size and 18F-FDG-PET SUV for malignant and benign pulmonary nodules. J Hematol Oncol 2008;1:13

33.

Kim

, Machac

, Krynyckyi

, Fluoro-deoxy-glucose positron emission tomography for evaluation of indeterminate lung nodules: assigning a probability of malignancy may be preferable to binary readings. Ann Nucl Med 2008;22:165–70