Multimodality CT imaging contributes to improving the diagnostic accuracy of solitary pulmonary nodules: a multi-institutional and prospective study

This study was conducted at our three teaching hospitals and approved by the institutional review committee of Suining Central Hospital (Suining, China; approval no. LLSNCH20200004). All patients were included after providing informed consent.

Patients with SPNs were admitted and treated at one of the three hospitals from January 2019 to June 2021 and were consecutively enrolled into our study. The inclusion criteria were: (1) SPNs were detected with NECT in the lung window; (2) each SPN size is 1.5 cm–3.0 cm; (3) benign or malignant SPNs were confirmed by CT-guided percutaneous biopsy or pathology after surgery; (4) CTPI and DECT were performed within one week prior to CT-guided percutaneous biopsy or surgery; (5) SPNs were not treated by any antitumor therapies (e.g., radiotherapy, chemotherapy, or targeted drug therapy); (6) age > 18 years old; and (7) patients had acceptable liver and kidney functions and no history of allergies (for example, allergic to iodine or seafood). The exclusion criteria were: (1) SPNs appearing as pure ground-glass opacity (pGGO) or as subsolid nodules that could not be measured in the mediastinal window; (2) majority components of the SPNs have been calcified or liquefactive necrosis, leading to region of interests (ROIs) cannot be drawn; (3) body mass index (BMI) is greater than 30.0 kg/m²; (4) poor image quality (e.g., significant anomalies by heart beats or others artifacts); and (5) incomplete clinical data. Three hundred and twenty-three patients were included and 38 of them were excluded because of the exclusion criteria (Figure 1).

Figure 1

All patients were scanned with the same DECT system (Revolution CT, GE Healthcare, Milwaukee, WI, USA), and all the scanning parameters were set at the same level. The multimodality CT imaging protocol included three steps, i.e., NECT, CTPI, and DECT.

For NECT, the patients were asked to hold their breath at the end of inspiration, and the scanning ranges were taken from the thoracic inlet to the bilateral costophrenic angle. The tube voltage was 100 kVp; the tube current was modulated by a SmartmA technology, which provides reasonable tube current to decrease the radiation dose; and the preset noise index was adjusted for patient circumference.

For CTPI, the patients were asked to take relaxed and slow respiration. After the position of an SPN was determined by the NECT, a GSI Chest Perfusion protocol was used with a scanning range of 3.0 cm above and below the central level of a SPN. A non-ionic iodine contrast agent of 400 mg/ml (Iomeron^®, Shanghai Braccosine Pharmaceutical Co., Ltd.) was injected into an antecubital vein with a binocular power injector (Stellant D-CE, MEDRAD, Bayer Healthcare Co., Ltd.), at a flow rate of 5.0 ml/s (40.0 ml). After that, 20.0 ml of normal saline at a speed of 5.0 ml/s was used to flush the tube. The CTPI was performed after injection of the contrast medium with a delay time of 4.0 s. The tube voltage and tube current were 70 kVp and 250 mA respectively. There were 18 scanning phases with an interval of 2.5 s, and the total acquisition time was about 43.0 s.

For DECT, the patients were asked to hold their breath at the end of inspiration, and the scanning ranges were the same as the NECT. Another 50.0 ml contrast agent was injected into the antecubital vein with the binocular power injector, at a flow rate of 2.5 ml/s. After that, 20.0 ml of normal saline at a speed of 2.5 ml/s was used to flush the tube. A ROI was placed in the thoracic aorta at the level of bronchial bifurcation, and the threshold was set to 120 HU. After reaching the threshold, the arterial phase was scanned with a delay time of 7.0 s, and the venous phase delay time was 30.0 s. The tube voltage, tube current, and noise index were the same as the NECT.

The same scanning parameters for NECT, CTPI, and DECT were: detector coverage = 80 mm; pitch = 0.992: 1; coverage speed = 158.75 mm/s; rotation time = 0.5 s; preset adaptive statistical iterative reconstruction-V (ASIR-V) = 50%; postset ASIR-V = 60%; slice thickness = 5.0 mm; slice interval = 5.0 mm; reconstructed slice thickness = 1.25 mm; reconstructed slice interval = 1.25 mm.

After the multimodality CT imaging protocol, the volume CT dose index (CTDI_vol, mGy) and dose-length product (DLP, mGy.cm) were recorded. The estimated effective dose (ED, mSv) was calculated with the formula: ED = DLP × k, where the k equals 0.014 mSv/mGy.cm.

For NECT, the lung window images were further processed with the post-processing workstation AW 4.7 (GE Healthcare, Milwaukee, WI, USA) to obtain multiplanar reconstruction (MPR) coronal and sagittal images (1.25 mm).

For CTPI, the images were transferred to the post-processing workstation AW 4.7 (GE Healthcare, Milwaukee, WI, USA), CTPI parameters of the blood volume (BV, ml/100 g), blood flow (BF, ml/100 g/min), mean transit time (MTT, s), and permeability surface (PS, ml/100 g/min) with their perfusion artificial color maps were automatically generated by the CT perfusion 4D software. The ROI should be placed in the soft tissue area of a SPN at its maximum plane, and the thoracic aorta at the same level plane was selected as the reference vessel for calculating the arterial input function and a time density curve (TDC). When placing a ROI, the calcification, bleeding, liquefaction necrosis area, blood vessels, and anomalies by heart beats or other artifacts should be avoided as much as possible.

For DECT, the arterial and venous phase mediastinal window images were respectively processed with the GSI volume viewer software (AW 4.7, GE Healthcare, Milwaukee, WI, USA), to obtain the arterial and venous phase iodine-based images. The following principles should be followed in outlining the ROI: (1) ROI should be placed in the solid area of a SPN with uniform enhancement, and the outlined area should be as large as possible; (2) the calcification, bleeding, liquefaction necrosis area, blood vessels, and anomalies by heart beats or others artifacts should be avoided as much as possible; (3) the position and size of ROI measurement in each phase should be consistent.

The para meter measurements for CTPI, DECT, and the CT enhancement amplitude value were independently measured by two senior radiologists with more than 15 years of experience in chest CT. All indices were measured two times, and the average values were taken as the final results. Disagreements were resolved through consensus. Parameter measurements for CTPI included BV, BF, MTT, and PS. Parameter measurements for DECT included iodine overlay (OL), iodine concentration (IC), and normalized iodine concentration (NIC) at the arterial and venous phases (i.e., aOL, vOL, aIC, vIC, aNIC, and vNIC). The NIC was calculated with the formulae: NIC = IC_SPN / IC_{Thoracic aorta}. The CT values of a SPN at the venous CECT and NECT phases were also measured, and the CT enhancement amplitude value was calculated with the formula: CT enhancement amplitude value = venous CECT value - NECT value.

First, another two senior radiologists with more than 15 years of experience in chest CT randomly and independently interpreted the conventional NECT images of each SPN. Any disagreements were solved through consensus. The SPNs were then classified into five categories, i.e., I = most likely benign, II = possibly benign, III = uncertain benign or malignant, IV = possibly malignant, and V = most likely malignant. The diagnostic methods were as follows: NECT combined with CECT (Method A), and the diagnostic criteria for CT enhancement were¹: a CT enhancement amplitude value of 20–60 HU was classified as malignant, and a CT enhancement amplitude value of less than 20 HU or more than 60 HU was classified as benign; NECT combined with DECT (Method B); NECT combined with CTPI (Method C). If methods A, B, or C were consistent when classifying a benign or malignant SPN, the SPN would be judged as benign or malignant. However, if methods A, B, or C were not consistent when classifying a benign or malignant SPN, the following principles would be followed: (1) if a benign or malignant SPN was definitely diagnosed as category I or V by NECT, the diagnostic results of CECT, DECT and CTPI would not be considered, and the SPN would be directly characterized as benign or malignant; (2) if the nature of a benign or malignant SPN could not be confirmed by NECT (i.e. a category of II, III or IV SPN), the final diagnosis would be made according to the diagnostic thresholds of CECT, DECT and CTPI, respectively. There were four combined diagnostic methods: method A + B, method A + C, method B + C, and method A + B + C (Figure 2). In combined diagnostic methods, the results of benign or malignant SPNs may be inconsistent according to the diagnostic thresholds of CECT, CTPI or DECT. Our method is that if a SPN is diagnosed as malignant according to one of the thresholds (CECT, CTPI, or DECT), it would be preliminarily categorized as malignant; then, the result would be compared with the pathological result to calculate the sensitivity, specificity and accuracy of the combined diagnostic method.

Figure 2

Measurement data were expressed as the mean ± standard deviation or median and interquartile range (P₂₅, P₇₅). The comparisons were performed with the two independent samples t-test or Mann Whitney rank sum test, as appropriate. The count data were expressed as the percentage (%) or composition, and the comparisons were performed with the Fisher’s exact test. Receiver operating characteristic (ROC) curve analysis was performed on statistically significant CTPI and DECT parameters to obtain the area under the curves (AUC) and diagnostic thresholds, and the Delong-test was performed to compare the differences.¹¹

In using the pathological diagnosis as the gold standard, the sensitivities, specificities, accuracies, positive predictive values (PPVs) and negative predictive values (NPVs) of the three diagnostic methods separately used (Method A, B, and C) and in combination (Method A + B, Method A + C, Method B + C, and Method A + B + C) for the diagnosis of benign and malignant SPNs were calculated, and the McNemar’s test was used to compare the sensitivity, specificity, and accuracy of the various diagnostic methods.

All statistical analyses were performed by using the software of GraphPad Prism 8.0.0 and MedCalc 19.5.3. Significant differences were set at a p value < 0.05.

Ultimately, 285 SPNs were included in this study. Of these, 178 (62.46%, 178/285) SPNs were confirmed by CT-guided percutaneous biopsy, and 107 (37.54%, 107/285) SPNs were confirmed by pathology after surgery. The patients’ characteristics are shown in Table 1. There were no significant differences between the malignant and benign SPNs relative to gender, age, smoking status, history of tumors, or tumor biomarkers (all p values > 0.05).

Of the 285 SPNs, 118 were benign SPNs (41.4%, 118/285) and 167 were malignant SPNs (58.6%, 167/285). Of the 118 benign SPNs, 46 were tuberculosis (29.0%, 46/118), 32 were acute and chronic inflammation (27.1%, 32/118), and 14 were inflammatory pseudotumors (11.9%, 14/118). Of the 167 malignant SPNs, 116 (69.5%, 116/167) were primary pulmonary carcinomas (including adenocarcinomas in 52, squamous cell carcinomas in 35, and small cell lung cancer in 17, etc.), 23 (13.8% , 23/167) were solitary metastasis (this included five from the liver, five from the breast, etc). Pathological results of the 285 SPNs included in this study are shown in Table 2.

After the multimodality CT imaging protocol, the CTDI_vol, DLP, and ED in the SPNs were 66.88 ± 4.36 mGy, 768.29 ± 91.65 mGy.cm, and 10.76 ± 1.28 mSv, respectively.

In the malignant SPNs (n = 167), 49 (29.34%, 49/167), 115 (68.86%, 115/167), 108 (78.81%, 108/167), 38 (22.75%, 38/167), 89 (53.29%, 89/167), and 84 (50.30%, 84/167) cases were seen with smooth margins, lobulated sign, spiculated sign, vacuole sign, pleural indentation, and vessel convergence, respectively. In the benign SPNs (n = 118), the same CT findings were seen in 84 (71.19%, 84/118), 29 (24.58%, 29/167), 25 (21.19%, 25/118), 13 (11.02%, 13/118), 25 (21.19%, 25/118), and 33 (27.97%, 33/118) cases (Table 3), respectively. There were significant differences between the malignant and benign SPNs in above mentioned findings on NECT (all p values were < 0.05). However, no significant differences were noted between the malignant and benign SPNs in cavity sign, air bronchogram, calcification, fat (defined as CT attenuation of -40 HU to -120 HU), adjacent bronchial changes, location, or size on NECT (all p values were > 0.05).

For the classification of SPNs (Table 4), among the 118 benign SPNs, there were 39, 9, 48, 22, and 0 SPNs that were classified as the Category I, II, III, IV, and V nodules, respectively. Among the 167 malignant SPNs, there were 0, 5, 54, 19, and 89 SPNs that were classified as the Category I, II, III, IV, and V nodules, respectively.

For CECT, among the 167 malignant SPNs, 118 cases (70.66%, 118/167) were noted with a CT enhancement amplitude value of 20–60 HU; and among the 118 benign SPNs, 75 cases (63.56%, 75/118) were noted with a CT enhancement amplitude value of less than 20 HU or more than 60 HU (Figure 3).

Figure 3

The parameters of BF, BV, MTT, and PS of CTPI in malignant SPNs were higher than those of benign SPNs (Figure 4, Table 5). However, there was only significant difference in the PS parameter (p < 0.05). A ROC curve analysis was performed on the PS parameter for differentiating benign and malignant SPNs (Figure 5). It showed an AUC of 0.739 (95% confidence interval of 0.684–0.789, p < 0.0001) with the sensitivity, specificity, and diagnostic threshold of 85.03 %, 66.10%, and 9.88 ml/100g/min, respectively.

Figure 4

Figure 5

For DECT in evaluating SPNs (Figure 6), parameters of aOL, vOL, aIC, vIC, aNIC, and vNIC in malignant SPNs were significantly higher than those of benign SPNs (all p values were < 0.05, Table 6). ROC curve analyses were performed on these parameters for differentiating benign and malignant SPNs (Figure 7). The AUCs, sensitivities, and specificities were 0.636 to 0.790, 59.88% to 75.45%, and 61.02% to 80.51%, respectively. The diagnostic thresholds were 13.89 HU, 12.79 HU, 0.65 mg/ml, 0.85 mg/ml, 0.12, and 0.35, respectively (Table 7). Compared with the AUC of aOL, vOL, aIC, vIC, and aNIC by the Delong test, vNIC had the largest AUC (all p values < 0.05, Table 8).

Figure 6

Figure 7

For methods A, B, C, A+B, A+C, B+C, and A+B+C, the sensitivities, specificities, accuracies, PPVs, and NPVs were 83.23% to 97.60%, 63.56% to 88.14%, 75.09% to 93.68%, 76.37% to 92.09%, and 72.82% to 96.30%, respectively (Table 9).

There were no significant differences between the methods A, B, and C in sensitivities, specificities, and accuracies, and so did the methods A+B, A+C, B+C, and A+B+C (all p values > 0.05). However, the methods A+B, A+C, B+C, and A+B+C all had higher sensitivities, specificities, and accuracies than the methods A, B, and C for distinguishing benign from malignant SPNs (all p values < 0.05).

As low-dose CT scanning is more and more widely used in the early screening of lung cancers, to some extent, NECT scanning has become one of the most commonly used examination modalities for SPNs.¹² In addition to confirming the location, NECT is also used to perform a morphologic evaluation of SPNs. Generally speaking, the smaller the SPN, the more likely it is to be a benign lesion. In terms of the margins and contours of an SPN, it can be classified as smooth, lobulated, irregular, or spiculated.^1,5,6 Studies have reported that about 80% of benign SPNs are not larger than 2.0 cm in diameter.^1,5,6,13 However, some studies showed that about 15% of malignant SPNs are not larger than 1.0 cm in diameter, and 42% of malignant SPNs are not larger than 2.0 cm in diameter.^1,14 Most benign SPNs have smooth and well-defined margins. However, about 21% of malignant SPNs have well-defined margins.¹⁵ A lobulated, irregular, or spiculated contour is usually associated with a malignant SPN. However, lobulation also occurs in up to 25% of benign SPNs.¹⁶ Therefore, there are considerable overlaps in the size, margins, and contours of benign and malignant SPNs. Internal characteristics (e.g. homogeneous attenuation, cavitation, intranodular fat, presence and pattern of intranodular calcification, etc.) and abnormalities surrounding nodules (e.g. pleural indentation, vessel convergence, adjacent bronchial changes, etc.) are also helpful for distinguishing benign from malignant SPNs.¹ For example, intranodular fat and a popcorn like calcification are typical findings seen in hamartomas. However, intranodular fat or calcification alone cannot be used to differentiate benign from malignant SPNs confidently.

In our study, there were significant differences between the malignant and benign SPNs only in smooth margins (29.34% vs 71.19%), lobulated sign (68.86% vs 24.58%), spiculated sign (78.81% vs 21.19%), vacuole sign (22.75% vs 11.02%), pleural indentation (53.29% vs 21.19%), and vessel convergence (50.30% vs 27.97%) on NECT (all p values < 0.05). Our results are consistent with some previous studies.^17-19 Some previous studies also reported that these findings on NECT are risk factors for malignant SPNs. However, we did not identify cavity sign, air bronchogram, calcification, fat, adjacent bronchial changes, location, or size as risk factors for malignant SPNs (all p values > 0.05). This may be explained by the number of patients with SPNs included in our and previous studies which varied largely. Our results and previous studies also showed considerable overlaps in the findings of benign and malignant SPNs on NECT. Therefore, when we performed the classification of the 285 SPNs, there were 157 cases that were classified as the Category II, III, and IV nodules.

For CECT, it is helpful to improve the accuracy of differential diagnosis between benign and malignant SPNs. The degree of enhancement is directly associated with the possibility of malignancy and the vascularity of the SPNs. Swensen et al.²⁰ reported that nodular enhancement of more than 20 HU on CECT highly implies a malignant SPN. Whereas, nodular enhancement of less than 15 HU on CECT typically indicates a benign SPN, with a sensitivity, specificity, accuracy, PPV, and NPV of 98%, 58%, 77%, 68%, and 96%, respectively. In our study, the diagnostic criteria for CT enhancement were: a CT enhancement amplitude value of 20– 60 HU was classified as malignant, and a CT enhancement amplitude value of less than 20 HU or more than 60 HU was classified as benign. We also want to highlight that CT postprocessing technology of thin-section and multiplanar reconstruction (MPR) are indeed beneficial when differentiating benign and malignant SPNs. This is because more details would be identified with thin-section CT and MPR. On the other hand, comprehensive morphological features evaluation is far better than a single morphological features assessment because more negative (or positive) morphological features will increase the likelihood of a malignant (or benign) SPN.

CTPI allows the derivation of several physiologic parameters, including BV, BF, MTT, and PS. BV is defined as the integral under a corrected attenuation curve for enhancement values from a contrast material bolus. It is a relative measure of the blood volume within small vessels in a region of specific tissue. BV is related to the number and diameter of open vessels, etc. MTT is a measure of the time it takes for blood to pass through small vessels. BF refers to the blood flow passing through a section of the blood vessel in unit time which is related to the patency of drainage vein, lymphatic reflux, blood volume level, etc. BF is calculated with the equation: BF = BV/MTT, where BV is blood volume and MTT is mean transit time. PS refers to the diffusion coefficient of the unidirectional transmission velocity of the contrast agent through capillary endothelial cells.

CTPI can be used to evaluate the perfusion and vascularity of lesions in the lung and many other organs.^21-28 Wang et al.²¹ reported that parameters of CTPI (including BF, BV and PS but not MTT) of SPNs were significantly correlated with SPNs’ microvessel density (MVD) and luminal vascular parameters such as luminal vascular number (LVN), luminal vascular area (LVA), and luminal vascular perimeter (LVP). Huang et al.²² analyzed the CT perfusion parameters (BF, BV, MTT, and PS) and the microvessel parameters (MVD, LVN, LVA, and LVP) in the non-small cell lung cancer (NSCLC) patients with and without lymph node metastasis.

In our study, only the PS parameter of CTPI was found to be significant difference between the malignant and benign SPNs (p < 0.05), and a ROC curve analysis performed on the PS parameter showed that the AUC, sensitivity, specificity, and diagnostic threshold were 0.739, 85.03%, 66.10%, and 9.88 ml/100g/min, respectively. Our results are not completely consistent with previous reports.^7,8,29 For example, in the systematic review and meta-analysis performed by Huang et al.⁸, lung cancer was found having higher BV, BF, MTT, and PS values than benign lesions; and AUC values of BV and PS were 0.92 (0.90, 0.94) and 0.83 (0.80, 0.86), respectively. The explanations may be as follow. PS mainly reflects the structural integrity of the capillary wall and the osmotic pressure between plasma and tissue fluid. There are a large number of microvessels and rich blood supply in malignant tumors. However, the capillary wall is immature, which is mainly reflected in the incomplete structure of vascular wall, including the basement membrane and epithelial cells. As a result, the vascular permeability will increase, promoting a large number of contrast agents to enter the tissue space. By contrast, the capillary wall of benign nodules is mature with a complete basement membrane and continuous endothelial cells which decreases the contrast agents to enter into the tissue space. On the other hand, although there are abundant microvessels in inflammatory lesions in benign nodules, in addition to the mature development of vascular wall, the edema and congestion of surrounding tissues will lead to the increase of tissue hydrostatic pressure which will also slow down the penetration rate of contrast medium. Therefore, the overall result is that the PS value in malignant SPNs was higher than that of benign SPNs.

There are mainly three techniques commercially available for DECT scanning. These techniques include: (a) fast voltage switching DECT, (b) layer detector DECT, and (c) dual-source DECT.³⁰ In this study, we used a fast voltage switching DECT system. It has the ability to obtain virtual nonen-hanced images, arterial phase and venous phase contrast enhanced images, and iodine maps in one examination.³⁰

An iodine map is the decomposition of material components realized by dual DECT according to the different attenuation characteristics of substances under high and low energy. It is the imaging of iodine material density extracted from enhanced iodine components. Iodine maps can effectively inhibit the background CT value and match with artificial color maps which can directly reflect the difference in iodine concentration in the lesion. The iodine concentration in the lesions can be measured quantitatively, and the lesions with slight enhancement can be displayed more sensitively. The CT value of the lesions measured on the iodine maps is the net enhancement value of the lesions. By measuring the enhancement value of lesions in iodine maps, we can evaluate the hemodynamic changes of tumors and the curative effects after tumor treatments.

In our results, parameters of DECT, which included aOL, vOL, aIC, vIC, aNIC, and vNIC, in malignant SPNs, were significantly higher than those of benign SPNs (all p values < 0.05). Compared with the AUC of aOL, vOL, aIC, vIC, and aNIC, the vNIC had the largest AUC (all p values < 0.05). Our results are consistent with many previous publications.^9,10,31-34 For example, Ha et al.³¹ differentiated pulmonary metastasis from benign lung nodules in thyroid cancer patients by using DECT parameters finding that the DECT parameters (IC, NIC, NIC in using pulmonary artery, λ_HU, and Z-effective values) of the metastatic nodules were significantly higher than those of the benign nodules (all p values < 0.05).

Zhou et al.³⁵ analyzed the relationship between clinical data, tumor markers, chest high-resolution CT (HRCT), and pathology in patients with SPNs finding that a joint evaluation model had better diagnostic efficiency for the diagnosis of SPN with diameter ≤ 2.0 cm. In another study, Zhu et al.⁹ reported that IC from DECT was significantly correlated with low-dose volume perfusion CT (VPCT) parameters (BF, BV, MTT, flow extraction product (FED), pulmonary nodule enhancement peak (PPnod), and VPCT parameters (BV, FED, and PPnod) had better diagnostic performance for SPN than DECT parameters (IC). Both of the above studies showed that multimodal evaluation is helpful in improving the diagnostic efficiency of SPNs.

In this study, the diagnostic performances of the three diagnostic methods were separately used or in combination for differentiating benign and malignant SPNs were calculated, and the results showed that multimodality CT imaging had higher performances than single modality CT imaging in differentiating between benign and malignant SPNs in sensitivity, specificity, and accuracy (all p values < 0.05). To the best of the authors’ knowledge, this is the first study in the literature in which NECT, CECT, CTPI and DECT were comprehensibly used to differentiate benign from malignant SPNs.

Radiation is an important consideration when using a multimodality CT imaging protocol. Previous studies have reported that the radiation dose of CTPI is the key factor hindering its application.^9,36 In our study, the CTDI_vol, DLP, and ED in the SPNs were 66.88 ± 4.36 mGy, 768.29 ± 91.65 mGy.cm, and 10.76 ± 1.18 mSv, respectively. We have tried to make the radiation dose in our study at an acceptable level by using low tube voltage, SmartmA, and iterative reconstruction.^37-39 Tube voltage and tube current are positively correlated with the radiation dose.^40-42 In addition, it is recommended that postset ASIR-V% (60% in this study) should be higher than or equal to preset ASIR-V% (50% in this study).^43-45 All these radiation reduction techniques have been shown to decrease the radiation dose while maintaining the image quality.^40-45

Some limitations of this study are: (a) only a fast voltage switching DECT was used in our study, and we did not compare our results with parameters from a layer detector DECT or dual-source DECT; (b) we did not compare our results with parameters from a magnetic resonance imaging (MRI) or positron emission tomography computed tomography (PET/CT); (c) some parameters of CTPI and DECT were not included into our study for evaluating; (d) radiation dose of our multimodality CT imaging protocol is still at a relatively high level; and (e) we did not include some advanced medical image analysis methods, such as artificial intelligence (AI), deep learning, or radiomics in our study. As a result, further investigations are needed to strengthen our findings in the future.