The International Association for the Study of Lung Cancer (IASLC) showed that, based on a newly established large database, the 5-year survival rates for patients with LNM ranged from 26 to 53% [10]. The systematic dissection of LNs in lung cancer patients has been widely accepted, but the extent of LN dissection has remained a matter of debate due to the precise assessment of metastatic LNs [11,12]. LNM is an important factor that affects tumor and LN staging. Therefore, the noninvasive preoperative evaluation of the LN status is crucial for determining the lung cancer stage, surgical plan, and prognosis [13].
Currently, CT are the most routinely used noninvasive methods for the clinical diagnosis of LNs. The international standard for the diagnosis of metastatic LNs by CT in lung cancer is a short-axis LN diameter larger than 10 mm. However, due to the single diagnostic criterion, the accuracy of the diagnosis is limited to some extent. Also, PET/CT is a noninvasive method for staging cancer that has been increasingly employed by multidisciplinary lung cancer teams. Many studies have reviewed the diagnostic performance of PET/CT for LN staging in patients with NSCLC [14~16]. A systematic review showed that the summary sensitivity and specificity estimates for a maximum standard uptake volume (SUVmax) ≥2.5, which is the PET/CT positivity criterion, were 81.3% and 79.4%, respectively [17]. However, the low prevalence and high cost of PET/CT equipment have resulted in it being less commonly used than CT alone in preoperative examinations. If the accuracy of CT in the diagnosis of LNs could be improved, it would provide more important clinical guidance for identifying the radiotherapy targets and surgical range.
Recently, the development of radiomics has enabled medical images to be converted into high-throughput quantitative data, providing information that can be explored and used to guide clinical decision-making. In contrast to subjective descriptions of the volume and shape of lesions, radiomic features can more comprehensively describe the state of lesions, overcoming the disadvantages of traditional diagnostic methods [18~20]. Therefore, radiomics is expected to improve the accuracy of diagnosis based on CT images. Moreover, studies have demonstrated the feasibility of using radiomic features to predict LNM in rectal, breast and esophageal cancers, providing theoretical support for this study [21~23].
In the present study, we constructed radiomic models based on pathological diagnostic results to facilitate the preoperative identification of metastatic LNs in NSCLC patients. The results showed that the diagnostic models based on different phases all exhibited favorable discrimination (AUC values greater than 0.8, a maximum sensitivity of 97.9%, and a maximum specificity of 86.0%), and model 1 (plain CT) yielded the highest AUC, specificity, accuracy and PPV. The underlying reason for the better performance on non-contrast images may be that the biological heterogeneity within the LNs that can be described by radiomic features may be confounded by the intravenous injected contrast material, which may then lead to worse discrimination between malignant and benign LNs due to the existing intratumoral contrast material. On the other side, the result of this stduty showed that more texture features (10 texture features) were selected from non-contrast CT than contrast-enhanced CT (9 or 5 texture features), and the texture features from plain CT scan were found to be more significant in discriminating mediastinal metastatic lymph nodes. Moerover, previous researches have confirmed this interesting finding. He et al.[24] evaluated the effects of contrast-enhancement on the diagnostic performance of radiomics signatures in solitary pulmonary nodules (SPNs), which indicated that contrast-enhancement can affect the diagnostic performance of radiomics signatures in SPNs and that non-contrast CT is more informative. Similarly, Sui et al.[25] confirmed that the radiomics features of the non-contrast CT have a better predictive performance than those of contrast CT in anterior mediastinal lesion risk grading. In the research of classifying mediastinal LNM of NSCLC from 18F-FDG PET/CT images,Yao et al. [26] summarized the diagnostic results from 22 research centers and found that the overall sensitivity and specificity were 0.66 and 0.82, respectively. In addition, another study showed that the sensitivity and specificity of CT for the diagnosis of mediastinal LNM were 0.79 and 0.72, respectively [27]. Compared to those published studies, the methods proposed in our study have the advantages of being quantitative and reproducible, with a higher sensitivity and specificity than the previously reported methods.
Moreover, we not only extracted radiomic features from plain, arterial, and venous phase CT images but also calculated delta radiomic feature values between different phase CT images in different phases. The arterial phase mainly reflects the tissue perfusion of the tumor, and the venous phase mainly reflects the clearing of the tissue blood flow, which is also an important imaging feature of tumor metastasis [28]. Dynamic CT texture analysis can assess temporal changes in tumor heterogeneity after the administration of contrast material and could provide another dimension of physiologic tumor assessment [29]. The sensitivity and NPV of the model were significantly better when combined with arterial phase CT in our study, which may have been because temporal changes in texture features can potentially provide diagnostic and prognostic information and canincrease the utility of contrast-enhanced CT. In clinical practice, for NSCLC patients treated with neoadjuvant therapy and routine radical surgery, false-positive LNs will not result in insufficient treatment or lead to treatment delay. However, the higher NPV of this approach means that negative LNs will be more accurately identified, which may change the clinical treatment plan [30]. These findings suggest that the accuracy of models can be improved when combined with dual-phase radiomic features in future clinical applications.
This method of integrating a large number of features in CT images that cannot be recognized or distinguished by the human eye has high accuracy and sensitivity for diagnosing mediastinal LNM in NSCLC patients and is expected to improve the efficacy of treatments for NSCLC. However, there are still some limitations of this study. First, the data used in this study were obtained from the same center. Second, the diagnostic capacity of combined clinical and quantitative imaging features could not be evaluated. Third, a minority of patients in our study underwent both CT and PET/CT before surgery because of the high cost of PET/CT and the high radiation exposure. We will add related cases in future studies and compare the study results with the PET/CT results, and the predictive accuracy might be further improved by automatically combining features determined by radiomic and deep learning. In summary, all of the CT radiomic models based on different phases all showed high accuracy and precision for the diagnosis of LNM in NSCLC patients. The combination of plain and venous phase CT scans with arterial phase CT radiomic features can further improve the sensitivity and NPV.