The Ethics Committee of our hospital approved this retrospective study and waived the need for obtaining individual patient consent. A total of 132 consecutive patients with suspected left upper lobe lung cancer who had undergone pulmonary angiography using multidetector row CT (MDCT) and left upper lobectomy between August 2012 and March 2019 were retrospectively reviewed. After excluding 24 patients who had inadequate investigation of tumor-involved hilar structures and technical problems, the final cohort comprised 108 patients (59 men and 49 women; mean age, 69.0 years; age range, 14–85 years) (Fig. 1). Some data from 99 of these patients were previously utilized in another study [12].

Both 64-slice MDCT (Aquilion 64, Toshiba Medical Systems, Tokyo, Japan) and 256-slice MDCT (Brilliance iCT, Philips Healthcare, Cleveland, OH, USA) scanners were used. The technical parameters for the 64- vs. 256-slice MDCT, respectively, were as follows: a detector row configuration of 0.5 mm vs. 0.625 mm, a pitch of 53 vs. 106 (detector pitch of 0.83 vs. 0.83), a reconstruction increment of 0.4 mm vs. 0.5 mm, and a section thickness of 0.5 mm vs. 0.67 mm. An x-ray tube voltage of 120 kV and an automatic exposure control for tube current were used in all examinations. The examinations were performed with the patient in the supine position during a single breath hold at end-inspiration.

A dual-head power injector (Dual Shot GX, Nemoto Kyorindo, Tokyo, Japan) was used for all patients for the bolus administration of the contrast material iohexol (Omnipaque 350, GE Healthcare Pharma, Tokyo, Japan) or iopamidol (Iopamiron 300, Bayer Yakuhin, Osaka, Japan) via a cubital vein. In patients weighing ≥ 55 kg, 100 mL of iohexol (350 mg iodine/mL) was injected at a rate of 3.3 mL/s, and scanning was performed 18 s afterwards. In patients with a body weight of 44–55 kg, 85 mL of iohexol (350 mg iodine/mL) was injected at a rate of 3.3 mL/s, and scanning was performed 15 s later. In patients weighing < 44 kg, 85 mL of iopamidol (300 mg iodine/mL) was injected at a rate of 3.3 mL/s, and scanning was performed 15 s later.

Another protocol for PA and pulmonary vein (PV) separation images was determined from the time-density curve using a test bolus dose. The injection rate was 4 mL/s, with a 20-mL test bolus injected prior to the main injection. The test injection determined the adequate timing for the PA/PV scan. The PA/PV scan was performed with 50 mL of iohexol (350 mg iodine/mL). The saline chaser was 40 mL, and the injection rate was 4 mL/s. These two protocols were comparable to investigating the PA branching pattern in detail. The volume data obtained from the arterial phase were transferred to a workstation (Zio STATION, Ziosoft, Tokyo, Japan), where the data were converted to a 3D-CTA format using the volume-rendering technique.

Thin-section transverse images were reviewed at a width of 1600 HU and level -200 HU window settings with paging on a viewer (EV insite, PSP Corporation, Tokyo, Japan). The 3D-CTA images were interpreted by rotating the same viewer. The window, level, and opacity of the volumes were subjectively selected to optimize PA visualization. In the present study, the number and origin of PA branches in the left upper lobe were meticulously identified using 3D-CTA and thin-section images on the same viewer. These images were reviewed with an interval of several days between interpretations.

Two board-certified thoracic radiologists, with 12 and 22 years of experience, respectively, independently reviewed each CT image. In cases of discrepancy over branching, the images were re-evaluated with both 3D-CTA and thin-section images until a consensus was reached to avoid interobserver variability. The intraoperative findings of the PA branches of the left upper lobe were compared with the preoperatively obtained 3D-CTA and thin-section images in each patient’s case.

The nomenclature used to describe the segmental PA is that of Yamashita [14]. The branches to the left upper lobe arise from the anterior, posterosuperior, and interlobar portions of the vessel: A1 + 2, A3, A4, and A5. The lingular artery (A4 and A5) originates from the interlobar portion (pars interlobaris [PI]) of the left PA (LPA) and may arise from the anterior portion of the mediastinal part of the left arterial trunk (pars mediastinalis [PM]). Moreover, the lingular arteries of PI may sometimes originate from the lower portion, from A8 or the common trunk of A8 and A9 [7]. In the present study, the lingular arteries are identified separately as PI and PI originated from the lower portion (from A8 or the common trunk of A8 and A9, denoted as PI’) [12, 15].

Statistical analyses were performed in SPSS version 26 (IBM Corp., Armonk, NY). Inter-observer agreement regarding the PA branching pattern in the left upper lobe was analyzed by calculating Cohen’s kappa coefficient before reaching a consensus with the thoracic radiologist who reviewed the images. The inter-observer agreement was classified as follows: excellent (κ = 0.81–1.00), substantial (κ = 0.61–0.80), moderate (κ = 0.41–0.60), fair (κ = 0.21–0.40), and poor (κ = 0–0.20).