Our institutional review board approved this retrospective study and waived the requirement for written informed consent (approval no.: 2022-0162).

The UHR-CT was installed at our hospital at the end of 2019 and began scanning for lung cancer in 2020. Therefore, we searched for postoperative histopathology reports of primary lung cancer cases operated on at our institution between January 2020 and April 2022 to identify pathologically proven pleural invasion and chest wall infiltration. Inclusion criteria were postoperative histopathological evidence of invasion of the visceral pleura or greater (pl1 as invasion beyond the elastic layer, pl2 as invasion to the surface of the visceral pleura, and pl3 as invasion of the chest wall, diaphragm, mediastinum or adjacent lung lobe). Exclusion criteria were the absence of chest wall-tumor contact on CT and the lack of UHR-CT data. Consequently, tumors involving the diaphragm, mediastinum, or interlobar pl3 were excluded. Selected lung cancers classified as pl1 and pl2 were categorized as the non-chest wall infiltration group, while those classified as pl3 were categorized as the chest wall infiltration group. We recorded patient characteristics and symptoms (age, sex, body weight, and chest pain) from medical records and tumor characteristics (lung lobe in the existence of a tumor, pathological tumor size, histological type, and TNM staging) from surgical and pathological records. TNM staging system (UICC, 8th edition) was used.

All CT scans were performed using the super-high-resolution mode of a UHR-CT scanner (Aquilion Precision; Canon Medical Systems Corp.) in the craniocaudal direction during inspiratory apnea, with dynamic contrast enhancement. The following scan parameters were used: beam collimation, 0.25 mm × 160 rows; x-ray tube voltage, 120 kV; tube current, auto exposure control (Standard Deviation 20); rotation speed, 0.5 s; helical pitch, 0.806.

At our hospital, two-phase (arterial phase and venous phase) dynamic CT scans are routinely performed for preoperative lung cancer, and we used arterial phase scan image data in the present study. The iodine contrast agent (96 mL) was administered intravenously at a rate of 4 mL/s, followed by 20 mL of saline at 4 mL/s. The iodine dosage was 240 mg/mL for patients weighing < 45 kg, 320 mg/mL for patients weighing 45–55 kg, and 370 mg/mL for patients weighing ≥ 55 kg. The bolus tracking method was used for the arterial phase, and the scan was initiated when the ROI of the ascending aorta exceeded 120 HU.

The UHR-CT images were reconstructed with a matrix of 1024 × 1024, field of view of 320 mm, slice thickness of 0.25 mm, slice interval of 0.25 mm, and deep-learning reconstruction kernel (AiCE Body_sharp Standard).

The reconstructed image data were transferred to the picture archiving and communication system (PACS) in our hospital and were displayed at window settings for lung (level −600 HU; width, 1600 HU) and soft tissue (level, 25 HU; width, 330 HU).

Two observers (a thoracic radiologist with 29 years of reading experience and a radiological resident with three years of reading experience) independently reviewed the UHR-CT images. They were blinded to the postoperative pathological diagnosis. Before the actual reading, findings were assessed using training UHR-CT images of 10 non-participating individuals (5 patients with chest wall infiltration and five patients without chest wall infiltration). They assessed the presence or absence of CWVI, rib destruction, pleural effusion, and ground-glass opacity (GGO) in the tumor. CWVI and rib destruction were rated as positive, equivocal, or negative. CWVI was evaluated as positive when vessels distributed from the intercostal artery to the tumor could be identified. Figures 1–3 show the CWVI-positive images. Pleural effusion and GGO were rated as positive or negative, respectively. When different findings were reported between the two observers, the decision was finalized by consensus.

Maximum intensity projection created from ultra-high-resolution CT images Ultra-high-resolution CT images of a 79-year-old female showing adenocarcinoma infiltrating the chest wall in the right upper lobe. a, b, and c are coronal sections arranged ventral to dorsal. d is a conventional HRCT image. The tumor diameter was 85 mm and showed chest wall vessel involvement in subpleural lung cancer (arrow). The postoperative pathology assessment confirmed parietal pleural invasion of the tumor

Maximum intensity projection created from ultra-high-resolution CT images of a 53-year-old female showing adenocarcinoma infiltrating the chest wall in the right upper lobe. a, b, and c are coronal sections arranged ventral to dorsal. d is a conventional HRCT image. The tumor diameter was 24 mm, and showed chest wall vessel involvement in subpleural lung cancer (arrow). The postoperative pathology assessment confirmed parietal pleural invasion of the tumor

Ultra-high-resolution CT images of a 73-year-old male showing squamous cell carcinoma infiltrating the chest wall in the left upper lobe. a, b, and c are axial sections arranged from bottom to top. d is a conventional HRCT image. The tumor diameter was 29 mm and showed chest wall vessel involvement in subpleural lung cancer (arrow). The postoperative pathology assessment confirmed parietal pleural invasion of the tumor

The maximum tumor diameter (Dmax) and the arch distance to the chest wall (Adist) were measured on an axial section using the PACS measurement tool, and the Adist-to-Dmax ratio (the ratio of maximum tumor diameter to arch distance with the chest wall (A/D ratio)) was calculated [13, 25]. Arch distance means the interface length between the primary tumor and the chest wall. Dmax and Adist were measured three times to reduce subjectivity, and the average values were used.

First, the inter-reader reliability between the two readers was analyzed using Cohen’s kappa coefficient (k). The degree of agreement was interpreted as follows: less than 0.20, poor agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, substantial agreement; and 0.81–1.00, almost perfect agreement.

Next, the selected tumors were classified into two groups: lung cancers not infiltrating the chest wall (pl1 and pl2) and lung cancers infiltrating the chest wall (pl3), and the two groups were compared. The age, pathological tumor size, Dmax, Adist, and A/D ratio were compared with t-tests, and the frequency of sex, histological type, lobe, pleural effusion, rib destruction, and CWVI were compared with the χ2-test. The exact comparisons were performed for patients in whom no rib destruction was observed.

Finally, univariate and multivariate logistic regression analyses were performed to determine the significance of chest wall infiltration in patients whose rib destruction was not observed. Significant variables in the univariate analysis were adopted for multivariate analysis, and a stepwise method was used for further variable selection.

Analyses were performed using SPSS version 28 (IBM Corp.), Excel 2019 (Microsoft Corp.), with an add-in statistical software BellCurve for Excel version 4.04 (Social Survey Research Information Corp.). Statistical significance was set at p < 0.05.