The differentiation between solitary pulmonary inammatory lesion and solitary cancer using gemstone spectral imaging (GSI): the correlation with lesion size

Background: There were still nonnegligible overlaps between lung inammatory and malignant lesions in morphologic appearance and enhancement pattern using conventional computed tomography (CT) especially when the lesion size was smaller than 3cm, which made misdiagnosis often happen in clinical practice. The study was to verify whether spectral CT parameters can differentiate solitary lung cancer from solitary inammatory lesions or not, and to nd the correlation between GSI parameters and lesion size. Methods: 102 patients with solitary inammatory or lung cancer lesions underwent enhanced CT scans with GSI mode. While total of 78 ROIs (region of interest) were nally collected and divided into inammatory group (group I: ≤ 3cm, IA; >3cm, IB) and cancer group (group II: ≤ 3cm, IIA; >3cm, IIB) according to their maximum diameter (D max ). CT values measured on 40keV and 70 keV monochromatic images (HU 40kev , HU 70kev ), effective atomic number (Z eff ) , iodine concentration (IC), normalized IC (NIC) and slopes of spectral curve (λ 30 , λ 40 ) were calculated. Intergroup and intragroup comparisons were made using Mann-Whitney U test, and receiver operating characteristic curve (ROC) analysis were done. Correlation analysis was used to nd the relationship between D max and GSI parameters. Results: NIC and λ 30 under venous phase could distinguish general solitary inammatory from cancer group. No signicant correlation was found between GSI parameters and D max in inammatory group, while negative correlations were found between them in cancer group. The GSI parameters (except HU 70kev ) of group IIA were signicantly higher than those of group IIB. When D max was larger than 3cm, there were signicant differences between solitary inammatory group and cancer group in HU 40kev , IC, NIC, λ 30 and λ 40 , under both AP and VP, while the AUC for λ 30 under VP was largest, and the corresponding sensitivity and specicity were 96.32% and 85.71% respectively. However, only the HU 40kev and HU 70kev under arterial phase of IIA were signicantly higher than those parameters, group IA and IB, IIA and IIB were compared, and the correlation between D max and GSI parameters were done. Also, we compared group IA and IIA, IB and IIB to verify the performance of GSI parameters in the distinguishing lung cancer and lung inammatory when the lesion size was lower and higher than 3 cm respectively. ROC analysis about the performance of these parameters were done. P value < 0.05 was considered to indicate a signicant difference. All statistical analyses were performed using SPSS (Version 25.0; Chicago, IL, USA) and Prism (Version 8.3; Graphpad Software; San Diego, CA, USA) software.

Evaluation of the solitary pulmonary lesions remains a challenge in clinical. Among all imaging examination tools, conventional contrast-enhanced CT scan is still the most important one for the patient with solitary pulmonary lesion [4]. It can offer the morphological signs, inner densities and enhancement patterns to characterize pulmonary lesions. However, there is considerable overlap between in ammatory and malignant masses in morphologic appearance and enhancement pattern, especially when the size of lesion is less than 3 cm, making it di cult to produce an accurate diagnosis in clinical practice.
With the rapid development of dual-energy spectral CT imaging, the researchers turned to explore its potential in the diagnosis of solitary pulmonary lesions. As one of dual-energy spectral CT imaging techniques [5], GSI was based on rapid switching between high and low voltage (80kVp and 140kVp) during a single rotation of the tube. It can acquire multiple parameters, such as monochromatic spectral images at energy levels ranging from 40 to 140 keV, material decomposition images, effective atomic number and spectral HU curve [6,7]. Specially, iodine concentration measured on iodine-enhanced images, which is considered to be comparable to the real value of enhancement, could re ect the blood supply of pulmonary lesions, and is essential for differential diagnosis [8,9]. The spectral CT imaging has been applied to multiple studies, such as non-small cell lung cancer (NSCLC) subtype differentiation [9,10], gastric cancer evaluation [11,12] and hepatocellular cancer differentiation [13][14][15], evaluation of lymph nodes preoperatively [16] and differentiation between benign and malignant solitary pulmonary nodules (SPNs) [17], etc.. Several researchers reported that spectral CT imaging with quantitative parameters may be useful in the differentiation of solitary lung cancer from solitary in ammatory lesions [18][19], however, they haven't taken the lesion size into account. Therefore, the aim of our study was to verify whether spectral CT parameters under GSI mode can differentiate solitary lung cancer from solitary in ammatory lesions, and clarify the correlation between GSI parameters and the lesion size.

Study population and groups
This prospective study was reviewed and approved by the Medical Ethics Committee of the Second A liated Hospital of Xiamen Medical College (approval No. 2016011) and written informed consent was obtained from all subjects prior to the examination. A total of 102 patients with suspected solitary pulmonary lesions were enrolled in our study during past eight months.
Patients were selected according to the following inclusion criteria: (a) presence of a solitary pulmonary lesion, (b) no contraindications to the administration of iodinated contrast material, and (c) were able to nish the procedures cooperatively. They were underwent dual-phase CT scanning under GSI mode. And seventeen patients were excluded due to : (a) the lesion was combined with cavitation and signi cant calci cation (n = 8), (b) absence or inadequate pathologic diagnosis (n = 9), (c) unsatisfactory image quality because of artifacts caused by contrast material (n = 3), and (d) ground glass nodules and /or nonsolid nodules (n = 4). Thus, 78 patients (49 men, 28 women, average age, 59.1 ± 7.3 years) were nally included in our study.
Seventy-eight lesions in all patients were divided into two groups based on the pathologic results. Lung cancer group (n = 47) was composed of adenocancer (n = 28) and squamous cell cancer (n = 19). Lung in ammatory group (n = 31) was composed of granuloma (n = 9) and pneumonia (n = 22). Taking 3 centimeter as the cutoff value for the maximum diameter of lesions, the above each group can be divided into two subgroups respectively (Table 1).

Imaging techniques
Computed tomography examinations were performed using two-phase enhanced CT scanning on a spectral CT scanner (GE Discovery 750HD CT ; GE Healthcare, Milwaukee, WI, USA) under GSI mode.
Patients were injected with the given volume (1.1 ml/kg of body weight) of nonionic contrast media named Ioversol (320 mg I/ml, GE healthcare) through venous access. The injection ow rate was 3-3.5 ml/s, followed by 30 ml saline with same rate. When the net CT value of descending aorta increased by 100HU, the arterial phase (AP) scanning was triggered (about 22 ~ 28 s after the start of contrast medium injection). The venous phase (VP) scanning was 40 s delay after the AP. The GSI scanning parameters were: scan range including the entire chest, helical tube rotation time 0.8 seconds, helical pitch 1.375:1, tube current of 260 mA, slice thickness of 1.25 mm and spacing of 1.25 mm. The CT dose index volume (CTDI vol ) for each phase under GSI mode was 7.37 mGy. A dataset of monochromatic images at energies ranging from 40 to 140 keV was generated, which can be analyzed with GSI Volume Viewer software package at AW4.6 workstation (GE Healthcare).

Images analysis
Two radiologists (YW.W and QH.C with 7 and 22 years of experience in chest CT diagnosis respectively), who were blinded to pathologic diagnosis, interpreted and analyzed the images independent of each

General comparison
The solitary lung in ammatory group and lung cancer group were compared in Table 2. There were no signi cant differences in HU 40kev , HU 70kev , Z eff at both AP and VP between general solitary in ammatory other.
(a). Firstly, raw data containing both AP and VP of each patient were loaded using GSI Volume Viewer software simultaneously to ensure the size, shape and location of regions of interest (ROIs) afterwards were the same on both AP and VP to the greatest extent.
(b). After reviewing the lesion manifestations, three ROIs were placed at the upper, middle (maximum section) and lower slices of the lesion respectively. The ROIs were outlined as large as possible on the condition of avoiding necrosis, vessels and bronchi. And the average values of the three represented the measurement.
(c). Outline a round ROI on descending aorta at the carina level.
(d). Measured the maximum diameter (D max ) of each lesion to represent the lesion size.
For each ROI, CT values measured on 40 keV and 70 keV monochromatic images (HU 40kev , HU 70kev ), effective atomic number (Z eff ) and iodine concentration (IC) measured on iodine/water based material decomposition image under two phases were obtained. Normalized iodine concentration (NIC) and slopes of the spectral HU curve (λ 30 , λ 40 ) under two phases were calculated. The NIC was de ned as the IC ratio between lesion and descending aorta (NIC = IC leison / IC aortic ). The λ 30 and λ 40 were calculated as |HU 40keV -HU 70keV |/30, and |HU 40keV -HU 80keV |/40 respectively. Above seven parameters of each phase were used in our study.

Statistical analysis
The parameters were expressed as mean ± standard deviation. All the data were test for the normality before two sample t-test comparisons. When the data did not meet the normality, Mann-Whitney U test was used to statistically compare these parameters between groups. Group I and II were compared to nd out whether the GSI parameters were useful in their differentiation diagnosis. To test the in uence of lesion size to the GSI parameters, group IA and IB, IIA and IIB were compared, and the correlation between D max and GSI parameters were done. Also, we compared group IA and IIA, IB and IIB to verify the performance of GSI parameters in the distinguishing lung cancer and lung in ammatory when the lesion size was lower and higher than 3 cm respectively. ROC analysis about the performance of these parameters were done. P value < 0.05 was considered to indicate a signi cant difference. All statistical analyses were performed using SPSS (Version 25.0; Chicago, IL, USA) and Prism (Version 8.3; Graphpad Software; San Diego, CA, USA) software. group and lung cancer group. While the IC, NIC λ 30 , λ 40 under VP and λ 30 under AP of lung cancer group were signi cantly lower than those of in ammatory group. The ROC curves for predicting solitary in ammatory from cancer based on these indices were compared (Fig. 1), and the AUC of λ 30 under VP were highest (AUC = 0.74). The intra-group comparison of GSI parameters and correlation analysis between D max and GSI parameter The in ammatory group (group I) was divided into two subgroups (group IA: D max ≤3 cm) and (group IB: D max >3 cm). No signi cant differences were found between the GSI parameters of group IA and IB under both AP and VP ( Table 3). The Pearson correlation between D max and GSI parameters of solitary in ammatory group were also done. No signi cant correlation between D max and any GSI parameters were found (p (two tailed) > 0.05) under both AP and VP, and Fig. 2 showed the scatter chart of GSI parameters plotted with D max under VP.  (Table 4). The Pearson correlation between D max and GSI parameters of solitary cancer group were also done. Signi cantly negative correlation between D max and any GSI parameters (except HU 70kev ) were found (p (two tailed) < 0.05) under both AP and VP, and Fig. 3 showed the scatter chart of GSI parameters changed with D max under VP.

The inter-group comparisons between subgroup IA and IB, IIA and IIB under two phases
Signi cant differences can be found in HU 40kev and HU70 kev under AP between lung in ammatory and cancer when the D max of a solitary lesion was no larger than 3 cm (Table 5, Fig. 4, Fig. 5). The AUC, sensitivity and speci city were 0.72, 95.0% and 58.8% for HU 40kev under AP, and 0.74, 75.0%, 82.4% for HU70 kev under AP. When the D max of a solitary lesion was larger than 3 cm, the HU 40kev , IC, NIC, λ 30 and λ 40 under both AP and VP of the in ammatory group were signi cantly higher than those of the cancer group (Table 5, Fig. 4, Fig. 6). The ROC analysis (Fig. 7) showed that the AUC, sensitivity and speci city for λ 30 under VP performed best (AUC 0.90, sensitivity 96.32% and speci city 85.71%).

Discussion
The solitary in ammatory and cancer were two common kinds of lung lesions in the clinical practice. However, due to the similarity of their manifestations in conventional CT, the differentiation between them was di cult for radiologists. In our study, we investigated the potential of quantitative parameters derived from monochromatic images and material decomposition images (water-iodine) provided by gemstone spectral CT imaging in this eld. As we all known, iodine was the main ingredient of contrast medium, which could directly re ect the blood ow and distribution in the intravascular and extracellular spaces of the tissue [18]. The spectral CT can provide both the monochromatic image series from 40 to 140 keV and the material decomposition images. Multiple parameters could derive from spectral CT, such as iodine concentration, spectral curve, effective atomic number and CT value under different keV maps. The iodine concentration or normalized iodine concentration re ects the blood supply of the masses. Spectral curve re ected the dynamic change of measured CT numbers of ROIs against increasing keV values. Each tissue had its characteristic HU curve. The slope is used to evaluate the attenuation characteristics in different tissues.
According to our analysis above, IC, NIC, λ 30 , λ 40 under venous phase and λ 30 under arterial phase of general solitary in ammatory group were signi cantly higher than solitary lung cancer group, although their AUCs (Table.2) under the ROC curves were not high. The ndings were partly consistent with some previous studies [19,20] about the differences of NIC and λ 30 between in ammatory and cancer lesions.
However, there were no signi cant differences between these two groups when using the parameters under the arterial phase expect λ 30 . Some reasons could be used to explained this nding. Firstly, the injection ow rate used in our study was 3-3.5 ml/s, while some previous studies [17,19,20] set this value to 4-5 ml/s or even higher 6 ml/s. Secondly, the timings for AP and VP were different. We applied the triggering technology in our study to minimize the impact of individual differences instead of the experienced values, and the VP was set 40 s delay after the AP. The scan timing may cause the nding differences among previous studies.
When we took the D max into account, the study revealed some interesting results. In clinical, we usually classi ed lung lesions less than 3 cm in maximum diameter as nodules, and larger than 3 cm as masses. The in ammatory and cancer groups in our study were also classi ed into two subgroups respectively. No signi cant differences were found between the GSI parameters of group IA (D max ≤3 cm) and IB (D max >3 cm) of the in ammatory under both AP and VP. Also, there was no signi cant correlation between D max and any GSI parameters. Therefore, this may not only indicate that the value of GSI parameters will not change with the lesion size, but also the structure tends to be homogeneous in the solitary lung in ammatory lesion. Proliferation of in ammatory granulation tissue or residual of acute in ammation will form granulomatous in ammation and organizing pneumonia, and the rich and dilatate capillaries of masses are stimulated by in ammation gradually [22,23]. Our nding was in line with the growth characteristics of lung in ammatory.
For lung cancer group, signi cant differences were found between GSI parameters of group IIA and IIB under both AP and VP except parameter HU 70kev (Table 4). Signi cantly negative correlation between D max and GSI parameters (except HU 70kev ) were also found. That was to say, the larger the cancer lesions were, the lower the GSI parameters were. This nding indicated that the structure of solitary lung cancer lesion tends to be inhomogeneous. This nding was in line with the lung malignant tumor growth characteristics. As we all know, malignant tumors tended to grow towards the blood vessels to nd nutrition [24]. In malignant tumors, the establishment of angiogenesis and irregular vascular networks usually appeared, which could not support the rapid growth of tumor. Subsequently, the delivery of oxygen reduced, hypoxic in lung cancer occurred. With the continuous hypoxic, the necrosis might appear, which resulted in the tumor inhomogeneous.
The distinction between solitary lung in ammatory and cancer when the D max were not lager than 3 cm confused radiologists much using the conventional CT in practice. In our study, HU 40kev and HU 70kev under AP of solitary cancer were signi cantly higher than those of solitary in ammatory when the D max was no larger than 3 cm, and the sensitivity and speci city for HU 40kev and HU 70kev under AP were 95.0%, 58.8% and 75.0%, 82.45% respectively. However, signi cant differences were not found in iodine concentration and the slope of spectral curve between above two groups, which indicated that the blood supplies of solitary lung in ammatory and cancer were close when the D max were not lager than 3 cm.
Nevertheless, our nding showed that GSI parameter might be a new way to distinguish them despite our nding might be affected by several possible factors, such as the subject number, scanning protocol, injection ow rate etc.. When the D max were lager than 3 cm, the HU 40kev , IC, NIC, λ 30 and λ 40 under both AP and VP, Z eff under AP and HU 70kev under VP of the in ammatory group were signi cantly higher than those of the cancer group. And the λ 30 under VP performed the best (AUC 0.90, sensitivity 96.30% and speci city 85.71%).
This study also had some limitations. First, the number of subjects in subgroups was small. Further collection needs to be performed in the future. Second, we divided the groups using 3 cm as a D max cutoff, more cutoffs should be used to reveal more ndings about the differentiation, especially the characteristics of lesion when their D max was between 2 cm and 3 cm. We will do more research with a larger number of lesions in the future.

Conclusions
In conclusion, most of quantitative parameters of GSI can contribute to the differentiation between solitary lung cancer from solitary in ammatory lesions when D max is larger than 3 cm, while only few parameters are meaningful when D max is smaller than 3 cm .