Pulmonary nodules are detected with increasing frequency due to widespread use of computed tomography (CT).1,2 The prevalence of incidental pulmonary nodules on standard CT studies is around 13%, while lung cancer screening will detect lung nodules in up to 53% of subjects, leading to a lung cancer prevalence of around 1.4% (0.5–2.7%).3 The optimal diagnostic approach for the management of indeterminate pulmonary nodules has been the subject of much discussion.4
The widely accepted guidelines published by the British Thoracic Society (BTS) and the Fleischner Society recommend the minimum nodule diameter thresholds and CT follow-up time intervals for surveillance of solitary nodules smaller than 8 mm.3,5 For nodules of ≥ 8 mm (300 mm3), the BTS guidelines recommend risk assessment using the Brock model. The above guidelines recommend either 3-month CT follow-up, work-up with positron emission tomography (PET) with 2-deoxy-2-[fluorine-18]fluoro-D-glucose (18F-FDG), tissue sampling, or resection for nodules of ≥ 8mm. CT characterisation using only morphological features is imprecise6,7, leading to an increased interest in computer-based radiomics assessment.8, 9, 10, 11, 12, 13, 14, 15 Serial CT imaging to monitor nodule size can be problematic as nodule growth varies with different cancers and causes patient anxiety.16, 17, 18 18F-FDG PET has high sensitivity but lower specificity of 82% for detecting malignant pulmonary nodules, particularly in those smaller than 10 mm.19 Imaging guided sampling of small nodules is also difficult, is associated with complications, and its diagnostic yield decreases further as nodule size decreases.3,20,21
Neovascularisation is a complex process known to be central to carcinogenesis.22 Advances in the imaging technology in the last two decades have enabled the study of perfusion characteristics within pulmonary nodules.23, 24, 25, 26, 27 As benign and malignant lesions have different vascularity, different perfusion parameters and dynamic 18F-FDG uptake properties can be expected.27, 28, 29, 30, 31, 32
The purpose of this pilot study was to evaluate the feasibility and accuracy of CT perfusion and dynamic 18F-FDG PET imaging in differentiating proven benign and malignant pulmonary nodules.
This single-centre prospective study was approved by the local Research Ethics Committee (13/SS/0153) and written informed consent was obtained from all participants.
Between December 2014 and December 2015, 20 consecutive patients who were referred to our respiratory outpatient clinic for an indeterminate incidental pulmonary nodule were recruited. The inclusion criteria were: a) incidentally detected soft tissue (solid) pulmonary nodules measuring ≥ 8 mm and < 30 mm on CT, b) either surgical excision, imaging guided biopsy or imaging follow up of the nodule planned. The exclusion criteria were: a) abnormal renal function, b) previous adverse reaction to iodinated contrast agent, c) known history of malignancy, d) pregnancy or breast feeding, e) patients who refused or were unable to provide informed consent.
The patients underwent a dynamic 18F-FDG PET/ CT and dynamic perfusion CT imaging within a 3 week time frame (mean, 6.4 days: range 1–18 days). Due to technical reasons, the dynamic PET data could not be used in 4 patients for the analysis, one of these patients had two synchronous nodules. CT perfusion analysis was performed in 17 of the nodules. One patient declined the CT perfusion scan and 3 patients had significant breathing artefact on the scans, rendering analysis non-feasible. All nodules were classified into either benign or malignant on the basis of a histopathological diagnosis (n = 16), or stability during 2 years follow up CT imaging (n = 5).
All patients were fasted for at least six hours before the imaging. Following a low dose CT scan for attenuation correction and localisation (120 kV, 50 mAs, 5/3 mm), patients were administered 400 MBq of 18F-FDG intravenously, and a dynamic 60 minute image acquisition was performed using a Siemens Biograph PET/CT scanner (Siemens Healthcare, Erlangen, Germany). Respiratory-gated PET data were reconstructed using a 15-frame protocol (7 frames×180 s, 7×300 s, 1×240 s), a matrix size of 256×256×53 with a voxel size of 2.65×2.65×3.00 mm3, and subsets expectation maximization (OSEM) method. A conventional PET/CT scan was performed on completion of the dynamic phase of the scan at 1 hour after injection of the tracer.
Dynamic perfusion CT scans were performed as previously described 25,28,33,34 on a 320-detector row CT scanner (Aquilion ONE; Toshiba Medical Systems, Tokyo, Japan) with 16 cm field of view coverage. Imaging was performed at 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 24, 30, 40, 50, 60, 90, 120 seconds, 3 minutes, 4 minutes, and 10 minutes following the intravenous injection of 70 ml of iodinated contrast (Iomeron 400 mg/ml, Bracco, Milan, Italy) followed by a 30 ml bolus of saline both at 9 ml/s through a 16 G cannula sited in the ante cubital fossa. Acquisition parameters were 100 kV, 100 mA, 0.5 seconds rotation time, 320 x 0.5 mm collimation, 512 x 512 matrix.
Reconstructed images were imported into PMOD 3.409 software (PMOD Technologies, Zurich, Switzerland) and the input function was determined by placing a spherical volume of interest (VOI) with diameter of 1 cm in the ascending aorta. VOIs were drawn around the pulmonary nodules semi-automatically with a threshold of 50% of the maximum voxel value within the nodule, and then the VOIs were copied to the dynamic imaging sequence to obtain the time activity curves (TACs) (Figure 1). The influx constant
The maximum standardised uptake value (SUVmax) was measured for each nodule on conventional FDG PET/CT images. For the semi-quantitative analysis, the mean standardised uptake values (SUVmean) were measured of the ascending aorta at the level of the arch, and within the right lobe of the liver. SUV ratios (SUR) were calculated between the nodule SUVmax, and the SUVmean of the mediastinal blood pool (SURBLOOD) and liver (SURLIVER). Criteria for malignancy were specified as SUVmax ≥2.5; SURBLOOD ≥1.56; SURLIVER ≥1.12. Qualitative assessment PET features were specified as following: 0 = no visible uptake; 1 = uptake less than mediastinal blood pool; 2 = uptake comparable to mediastinal blood pool; 3 = uptake greater than mediastinal blood pool; 4 = distant metastases. Qualitative specified criteria for malignancy was PET grade ≥ 3.36, 37 VOIs were placed over the nodules, the ascending aorta at the level of the arch, and within the right lobe of the liver for determination of the SUVmean and SUVmax values using OsiriX software (OsiriX, version 8.0.1 64 bit; OsiriX Imaging Software, Geneva, Switzerland).
Perfusion analysis was performed using Body Perfusion Application on a Vitrea Workstation (Vitrea fX 6.0; Vital Images, Minnetonka, MN, USA). Regions of interest (ROIs) were placed over the pulmonary nodules and contralateral lung parenchyma (diameter range, 7–29 mm) on all perfusion CT images. Arterial input was determined by placing 1 cm ROI over the main pulmonary artery. Time-density graphs were then reviewed and adjustments to start point and end point of the maximum slope were made if needed to define the optimal slope range. Arterial flow perfusion maps overlaying CT images were visually analysed and ROIs were placed over the nodules to obtain the equivalent blood volume parameter calculated by Patlak plot model (BV, expressed in ml per 100 ml) and Arterial Flow (AF, expressed in ml per 100g per minute) using single-input maximum slope model for calculation.
All results were expressed as mean ± standard deviation (SD) unless indicated.
The demographic data, average nodule size, SUVmax, metabolic parameter relating to the pulmonary nodules through dynamic 18F-FDG PET/ CT, and perfusion parameters through perfusion CT for the benign and malignant nodules are summarised in Table 1 and Figure 3. We analysed 21 soft tissue nodules in 20 patients (male/female = 11/9; mean age ± SD: 65.3 ± 7.4; age range: 50–76 years) with mean nodule diameter ± SD of 20.1 ± 7.5 mm (9–29 mm); mean nodule volume ± SD: 2849 ± 2338.7 mm3 (247–9348 mm3). 52% of the nodules were located in the upper lung lobes (right upper lobe 7/21, left upper lobe 4/21), 48% were in middle and lower lung lobes (right middle lobe 2/21, right lower lobe 6/21 and left lower lobe 2/21). Final diagnosis was determined after surgical resection in 10 patients, core CT guided biopsy or bronchoscopy in 6 patients, and over 2 years stability on follow up CT imaging in 5 patients.
The demographic data, average nodule size, standardized uptake value (SUVmax), metabolic parameter relating to the pulmonary nodules through dynamic 18F-FDG PET/CT, and perfusion parameters through perfusion CT for the benign and malignant nodules
Benign Nodules | Malignant Nodules | p value | |
---|---|---|---|
Total Number of nodules | 9 | 12 | |
Number of male patients (%) | 5/9 (55 %) | 6/12 (50 %) | |
Average patient age (years ± SD) | 63 ± 7.5 | 68 ± 6.7 | |
Average nodule size, range (mm) | 18, 9–29 | 22, 12–30 | |
Average SUVmax 18F-FDG PET/CT ± SD | 2.2 ± 1.7 | 7.0 ± 4.5 | 0.0148 |
Number of nodules analysed for dynamic 18F-FDG PET/CT | 7 | 9 | |
Average |
0.0057 ± 0.0071 | 0.0230 ± 0.0155 | 0.0311 |
Number of nodules analysed for perfusion CT parameters | 7 | 10 | |
Average BV ± SD (Patlak, ml/100ml) | 11.6857 ± 6.7347 | 28.3400 ± 15.9672 | 0.0250 |
Average AF ± SD (ml/100g/min) | 74.4571 ± 89.0321 | 89.2000 ± 49.8883 | 0.1613 |
AF = Arterial flow; BV = blood volume;
As shown in Table 1 and Figure 3, SUVmax derived from the conventional 18F-FDG PET/CT and
The benign outlier on 18F-FDG PET/CT (SUVmax = 6.3) and dynamic 18F-FDG PET/CT (
Table 2 and Figure 4A show diagnostic accuracy of conventional PET/CT derived parameters with pre-specified and derived cut-point values though ROC analysis.36, 37 SURBLOOD parameter had overall highest accuracy, however, pairwise comparison of AUCs showed no significant difference (p = 0.5308
Comparison of the diagnostic accuracy of different techniques and parameters with pre-specified and derived cut-point values for malignancy
Parameter | Cut-point value/grade | Sensitivity (95% CI) | Specificity (95% CI) | Accuracy | |
---|---|---|---|---|---|
SUVmax | Pre-specified |
≥ 2.5* |
75.0% (46.8 to 91.1%) |
66.7% (35.4 to 87.9%) |
71.4% |
SURBLOOD | Pre-Derived specified | ≥ 1.56 | 83.3% (55.2 to 97.0%) | 88.9% (56.5 to 99.4%) | 85.7% |
SURLIVER | Pre-specified |
≥ 1.12 |
83.3% (55.2 to 97.0%) |
66.7% (35.4 to 87.9%) |
76.2% |
SUV grade | Pre-specified & Derived | ≥ 3 | 66.7% (39.0 to 86.2%) | 77.8% (45.3 to 96.0%) | 71.4% |
Derived | ≥ 0.01 min-1 | 77.8% (45.2 to 96.0%) | 85.7% (48.7 to 99.3%) | 81.3% | |
BV | Derived | ≥ 21 ml/100ml | 70% (39.7 to 89.2%) | 100% (64.6 to 100%) | 82.4% |
AF | Derived | ≥ 65 ml/100g/min | 70% (39.7 to 89.2%) | 85.7% (48.7 to 99.3%) | 76.5% |
* = adding cut-points standardized uptake value (SUVmax) ≥ 1.75 and ≥ 3.6 for nodules < 12 mm and > 16 mm, respectvely,37 resulted in sensitivity, specificity and accuracy of 72.7%, 70.0% and 71.4%, respectively;
AF = Arterial flow; BV = blood volume; CI = confidence interval;
Our results demonstrate that the metabolic parameter
Our study showed that the diagnostic accuracy of the conventional 18F-FDG PET/CT was best when semi-quantitative assessment and measuring the uptake ratio of the lung nodule to the mediastinal blood pool with cut-point criteria for malignancy SURBLOOD ≥1.56 was used. This has been confirmed in a larger multicenter trial by Evangelista
The accuracies of the new metabolic parameter
The benign outlier on dynamic 18F-FDG PET/CT with a high
The malignant outliers on dynamic 18F-FDG PET/CT with low
Dynamic enhancement CT studies help identify false positive results in both inflammatory and infective conditions, and sometimes in benign vascular tumours.46, 47 The perfusion CT parameters for the inflammatory nodule in our study were low and indicative of a benign lesion despite high metabolic activity on 18F-FDG PET/CT. We have shown that the parameters of perfusion CT of both malignant nodules with low metabolic activity were higher than the BV and AF in benign nodules. Therefore, our findings indicate parameters of perfusion CT may aid in the identification of benign nodules with high glucose metabolic activity and in the identification of malignant nodules with low glucose metabolic activity. Ohno
The AF parameter of the perfusion CT obtained by the maximum slope method was not significantly different between benign and malignant nodules. Benign nodules had a lower AF parameter value than malignant nodules overall with one significant benign outlier with markedly high AF. Histopathologically, this represented an extremely rare ‘light cell’ or ‘sugar type’ PEComa. There are only about 50 cases of this neoplasm described in the literature.48, 49 PEComas are more commonly found as angiomyolipomas in the kidneys, or as lesions in the retroperitoneal space, gastrointestinal tract, or uterus. Only 7 cases of malignant pulmonary PEComa have been reported.50 A case report of a benign pulmonary PEComa showing early wash-in enhancement with an early washout pattern of a malignant lesion on perfusion CT has been reported by Kim
Our study has limitations. This pilot study was performed in a small sample of patients and appropriately powered studies will be required for further validation. The mean nodule size was 18 mm for benign and 22 mm for malignant nodules, which would not normally be referred for imaging follow-up. The BTS and Fleischner Society recommended lower thresholds for nodule follow up (5 mm and 6 mm, respectively). More novel reconstruction methods in PET/CT such as specific point spread function (PSF) are enabling better spatial resolution and enable its use in 6 mm pulmonary nodules.52
Perfusion CT is quite demanding on patients with a prolonged breath-hold, which limits the availability of reliable data in some patients. All 3 nodules in which analysis was non-feasible due to the significant breathing artefact were near the diaphragm (2 in the right lower lobe and 1 in the right middle lobe). Segmentation of the pulmonary nodules on image analysis is restricted when the images were affected by respiratory motion artefact, especially in small nodules which were also abutting the chest wall or mediastinal structures. Some authors recommend quiet breathing during the perfusion CT scans but this is only acceptable in larger lung masses.53 There is a need for further optimisation of nodule segmentation and advanced image registration techniques that allow accurate assessment of pulmonary nodules without need for long breath-hold.23,54 The effective radiation dose for dynamic 18F-FDG PET/CT was around 8 mSv and for perfusion CT around 20 mSv. The radiation dose for perfusion CT can be improved by reducing the field of view from 16 cm to the nodule only and reducing tube voltage in smaller size patients.55
Potential increase in the demand for these not widely available novel dynamic imaging studies would consequently put additional strain on the imaging departments with increased demand for scanner time, funding and training of the staff. Limited capacity for a wider use of the dynamic imaging in lung nodules could be overcome by developing systems of identification of nodules with highest diagnostic benefit from dynamic imaging. A multicentre prospective cohort observational study initiated in 2016 is set to assess the performance and the cost-effectiveness of the dynamic CT and PET/CT in the characterisation of solitary pulmonary nodules.56
The small sample size limits the assessments of accuracy in our study. However, on this small sample we showed increase diagnostic improvement in the accuracy of diagnosis in both dynamic studies when compared to the conventional 18F-FDG PET/ CT. Specificity in
Early identification of a lung nodule as benign or malignant by analysing its metabolic and perfusion parameters could reduce the need for CT to monitor lung nodule size, thereby reducing the number of CT scans required. It could also reduce the need for CT guided biopsy or other invasive procedures. Patients with malignant lung nodules could thus be identified more quickly and referred for radical treatment. With our study, we have demonstrated the potential of perfusion CT. The BV parameter assessed by perfusion CT was not only significantly lower in benign nodules, it also aided in correctly characterising the metabolically active inflammation, hypervascular benign PEComa and low density malignancy.
In conclusion, this study demonstrated the feasibility of dynamic 18F-FDG PET/CT and CT perfusion studies in differentiating benign and malignant pulmonary nodules. The dynamic 18F-FDG PET/CT and perfusion CT derived blood volume can assist to differentiate benign and malignant lung nodules and in indeterminate cases, a combined approach can be helpful.