With the development of imaging and the popularization of GGN follow-up, the risk prediction model of GGN based on high-resolution CT is increasing year by year(14), some models have been applied in clinic, but the prediction of GGN is not enough. Among them, McWilliams A's model and Garau N's model achieve high prediction accuracy, AUC 0.94,0.89(15, 16). However, their model was based on the Canadian and Nordic populations and included variables that were uncommon in the Chinese population and did not apply to the Chinese population projections. Sun y and Liu A's model was based on the Chinese population(17, 18), but their model AUC was 0.77 and 0.836, which was less accurate than our prediction model. Importantly, all of the above models are predictive of benign and malignant GGN, and we know that GGN is often initially confirmed by 3–6 months of observation, therefore, the prediction model for benign and malignant nodules may be less meaningful than follow-up, whereas our model is intended to determine the corresponding surgical approach but rather to predict whether GGN reaches the invasive phase; in addition, according to the IASLC lung cancer staging project(19), we know that GGN in different stages has different CTR ratios and different morphological characteristics. CTR ratios of 0804 and 0802.1211 are used as a predictor of prognosis (11–13), there is good consensus on the treatment of nodules smaller than 2 CMGGN, but 2–3 CMGGN may be in the invasive stage due to the large diameter of the nodules, and there is currently a lack of relevant clinical studies, therefore, the prediction model of 2-3cm GGN invasion stage has higher accuracy and can guide the choice of surgical method more effectively.
Our study found that GGN diameter and CTR were important predictors, and nodule diameter was the first predictor to emerge in a predictive model for the differential diagnosis of benign and malignant pulmonary nodules, using the Mayo Clinic Model (AUC = 0.833) constructed by Swensen et al(20), gould et al built the VA model using data from the Department of Veterans Affairs (AUC = 0.79)(21), also, our own Li et al, using the PUPH model constructed by the People's Hospital of Peking University, all used nodule diameter as a key molecule in predicting benign and malignant nodules(22). CTR values are derived from unique imaging features of the lung; The clinical implications of this concept are supported by multiple studies (23, 24). The eighth edition of tumor staging argues that, histologically, the ground-glass component of pulmonary nodules is associated with a lymphocyte growth pattern, whereas the solid component is associated with an invasive adenocarcinoma pattern(19). Some studies have found that the malignant degree of tumors increases significantly with the increase of the diameter of nodules and CTR. From the observation of patients' survival time, the Five-year survival rate rate of patients with CTR less than 0.75 is 97.4%, whereas 86.1% of patients with CTR greater than 0.75(25). Thus, these two indicators play a crucial role in the prediction of whether tumors reach the invasive stage, and previous studies agree with our results.
pleural depression sign is an imaging feature in which a subpleural nodule or tumor contacts the visceral pleura, causing the visceral pleura to be pulled toward the lesion (26). The pleural indentation sign has a variety of manifestations (27) and is associated with the invasiveness of adenocarcinoma of the lung(28). Pleural indentation has been studied as evidence of non-small-cell lung carcinoma infiltration into the pleura(29). Through our study, we can identify this sign as an imaging risk factor for the development of GGN in the invasive phase, and further understand the relationship between pleural indentation sign and the pathological components of adjacent pleural regions, can promote the development of GGN personalized treatment.
Vacuole sign is related to the infiltration and development of GGN. Vacuole is a residual cavity formed by lung necrosis and liquefied material discharged through Bronchus. Vacuolar features are important in distinguishing lung cancer from benign lesions(30). Based on the CT findings of the Solitary pulmonary nodule, Shi et al found that vacuole was a risk factor for malignancy, while calcification and satellite lesions were protective factors. And vacuole sign is the first imaging sign in the development of tumor(31). A study found 5-year survival rates of 68.42% and 59.46% in patients with homogeneous and vacuolar nodules, respectively (32), demonstrating that vacuolar sign has a negative impact on patient outcomes.
The burr sign was pathologically associated with increased lobular interstitial thickness, fibrotic or carcinomatous lymphatics due to small peripheral vessel occlusion(33), and in this study, the spiculate sign was found to occur during the invasive development stage of GGN, this is consistent with previous studies showing that burr-bearing nodules are more likely to be malignant than those with a smooth, well-defined margin(34) ,and another study suggesting that, the positive predictive value of Burr for malignancy was as high as 90%(35). In addition, classic predictive models such as the Mayo Clinic Model(20) and Brock Model (15) also identified the spiculate sign as a risk factor for malignant pulmonary nodules.
The lobulated sign is closely associated with the growth pattern of malignant tumors, with unbalanced growth of solid components within it, resulting in radiographic changes similar to cauliflower-like ones (36). In partially solid nodules, a lobulated border is more invasive (37, 38), and this phenomenon is present in multiple cancers (39). In this study, the OR of the lobulated sign was 3.006(95% CI 1.098–8.227), which I believe has a high value in predicting the development of lung cancer invasion.
The bronchial sign refers to the presence of an air-bearing track within the nodule(40), which is common in malignant nodules(41), and is mostly due to the fact that the trachea extends in reverse within the tumor when the tumor retracts due to fibrosis. This symptom is seen in all lung cancer cell types but is more common in adenocarcinoma (42). According to Qiang JW et al, there are five types of bronchial sign, including continuous open type, enveloping type, tree-like narrowing type, compressed narrowing type, and compressed flat type, the first three types are associated with malignant progression of tumours(43). Multidisciplinary studies in imaging and molecular biology have shown that bronchial signs are associated with mutations in EGFR activity(44). Therefore, our study confirms that it is significant to classify it as an imaging risk factor for the development of infiltration in GGN.
Our prediction model still has some limitations. First, our data came from a single centre and were investigated only in the Chinese population, thus limiting the generalizability of the model; and our study used a two-dimensional CTR value rather than a three-dimensional proportion of solid components; In addition, the characterization of the patient's imaging features is subjective and may have an impact on the outcome. I think that adding a three-dimensional dimension of the proportion of solid components, conducting a multi-center study may improve the model's predictive performance.