Patients
The patients with GGNs (from January 2017 to December 2018) were pulled from our institution’s picture archiving and communication system (PACS, DJ Health Union Systems Corporation) under the approval of the institutional review board with a waiver allowing the data to be used retrospectively without informed consent. Inclusion criteria: (A) preoperative chest HRCT examinations performed within 2 weeks; (B) MIA or IAC confirmed pathologically after thoracic resection which tissue specimens were complete and reviewable; (C) P-pGGNs located next to the pleura on the preoperative HRCT, defined as pGGNs (window width, 1,500 HU; window level, -600 HU) contacted or connected to the pleural surface including the non-interlobar and interlobar pleura. Exclusion criteria: (A) underwent tumour therapy (radiotherapy, chemotherapy, etc.), puncture biopsy, or surgical resection were performed before the HRCT examinations; (B) impaired image sequences in PACS; (C) detectable soft-tissue attenuation inside the tumour on the mediastinal window images (window width, 400 HU; window level, 40 HU). (D) VPI cases observed during pathologic review. The demographic (age and sex) and smoking history characteristics of all P-pGGNs were recorded.
CT imaging parameters
Unenhanced chest CT examinations for all enrolled patients were obtained before P-pGGNs resection via scanning machines (Optima CT660, Discovery CT750 HD, Revolution CT or LightSpeed16 from General Electric, SOMATOM Perspective or Emotion 16 from Siemens, Brilliance 16P from Philips). All patients underwent chest examinations with their hands in a supine position on either side of the head from the lung apex to the lung base. All patients were instructed to hold their breath for the whole scan period in a deep-inhalation state. Though various chest CT imaging protocols were used in this study, all examinations were performed with contiguous 1.00–1.50 mm axial sections and 1.00–1.50 mm slice intervals and 0.625–1.50 mm section thickness after reconstruction. Imaging parameters as follows: a matrix of 512 × 512, a tube current of 170–200 mA, a tube voltage of 120 kVp, rotation times of 0.5–0.6 s, and a full field of view. Data were reconstructed with a lung kernel algorithm.
Review of pathological sections
All histological specimens were reviewed by a senior pathologist (Z.Q.P. with 30 years of experience) who blinded to the primary pathological diagnosis. The comparison between the reviewed and primary pathological diagnoses were recorded. Pathological diagnoses were classified in accordance with the lung tumor classification suggested by the IASLC/ATS/ERS in 2011 [11].
All histological specimens were formalin fixed and hematoxylin–eosin stained. The observation of pleura in all sections were based on hematoxylin–eosin staining. Ki-67 staining was performed with Ki-67 protein antibody (Santa Cruz Biotechnology, Santa Cruz, CA, USA) at a dilution of 1:100. Ki-67, a protein expressed in proliferating cells, has been used to identify high risk people group in LAC. Cells with brown-stained nuclei are considered positive cells, which are associated with cell proliferation. Increased Ki-67 expression is also significantly associated with lymph node metastasis and a poorer prognosis of LAC [12, 13]. The microscopic features of pathological components of the area adjacent pleura within lesion were categorized by using Olympus microscope as: alveoli/lepidic, acinar/papillary and solid/micropapillary type [10]. Distance was defined as the shortest distance between the largest invasive component area and the pleura; Depth was defined as the depth of the largest invasive component area (the vertical distance between the point closest to the pleura and the point farthest from the pleura within invasive component area). Distance and depth were calculated microscopically. The detailed schematic diagram of measurement is shown in Fig. 1. The largest area closest to the pleura was selected when the invasive components were scattered. The subpleural Ki-67 staining were counted. The range of Ki-67 counting was from visible pleural radiation to the center of the lesion and the average was recorded.
Image characteristics analysis
All HRCT images of P-pGGNs were observed by two thoracic radiologists (a junior radiologist and a senior thoracic radiologist with 20 years of experience) who blinded to the pathological results in the lung window (window width, 1500 HU; window level, -600 HU). The following imaging features were recorded and the discrepancies in the observation between two radiologists were resolved by consensus: (1) tumor location; (2) shape: round and oval, irregular; (3) tumor-lung interface: clear or not; (4) lobulation: defined as a portion edge of the nodule was wavy or fan-shaped; (5) vacuoles: defined as 1.00–3.00 mm cystic transparency of air attenuation within nodules; (6) airbronchogram: defined as the lucency along the regular bronchial wall within the nodules. The maximum axial layer selected from the HRCT image on the lung window: (7) the maximum diameter (MD); (8) the maximum vertical diameter (MVD) of the MD; (9) CTv: the CT attenuation value. When measuring all quantitative CT features, it should be noted that region of interest of P-pGGNs were delineated by a unregular curve and large vessels, bronchus and vacuoles should be avoided, when they existed in the measurement layer.
The five subgroups were allocated according to pleural deformation [2]: Type a, pleural attachment without pleural distortion; Type b, pleural tag with pleural folding; Type c, pleural tag without pleural folding; Type d, pleural retraction showing enfoldment of the pleura into the tumour; Type e, tentiform indrawing of the pleura toward the tumour (Fig. 2).
Statistical analysis
The normal distribution was confirmed by the Kolmogorov–Smirnov test. The T-tests were applied for normally distributed features expressing as the mean ± standard deviation, Mann–Whitney U tests were applied for nonnormally distributed features expressing as the median and quartile. The Kappa analysis was used to assess agreement for P-pGGNs primary and reviewed pathological diagnoses. Pearson’s chi-square tests or Fisher’s exact tests were applied for categorical variables. All statistical analyses for the present study were executed by SPSS (version 26.0, IBM, Armonk, NY, USA). A two-tailed P-value < 0.05 was considered to indicate statistical significance.