The institutional ethical review board approved this retrospective study and waived the requirement for informed consent (2020–469). The department received a research grant from Canon Medical Systems (Otawara, Japan). This work was technically supported by Canon Medical Systems. The corresponding author had complete control of the data and the information submitted for publication.

Study sample

This retrospective study included eligible consecutive patients who underwent 4DCT at our institution to evaluate the resectability of BR-LAEC from January 2011 to April 2021. The inclusion criteria were borderline resection on initial conventional CT (i-CT) and consideration for surgery following multidisciplinary discussions among the Departments of Diagnostic Radiology, Endoscopy, Esophageal Surgery, Radiotherapy, and Gastrointestinal Oncology. The 4DCT was performed within a week of this decision. The exclusion criteria are shown in Fig. 1. All patients received NAC, CRT, radiotherapy only (RT), or surgery only (Ope) as initial radical treatment.

Fig. 1

Flowchart of patient selection. 4DCT four-dimensional dynamic ventilation CT, CRT combined chemoradiotherapy, RT radiotherapy

The clinical tumor category (clinical TNM category) was determined using CT and endoscopic ultrasonography based on the 7th/8th edition of the Union for International Cancer Control staging system, which has consistent criteria for T and N categories. All patients received NAC, CRT, RT, or Ope as the initial radical treatment. For NAC or CRT, oncologists tailored chemotherapy regimens based on individual patient factors. Adjustments were made in cases of severe hematological toxicity or significant tumor growth. The surgery involved transthoracic esophagectomy with two- or three-field lymphadenectomies. Patients unsuitable for chemotherapy underwent Ope or RT. The treatment decisions were maximally aligned with patient preferences.

CT image acquisitionInitial conventional CT technique

All patients underwent preoperative i-CT scans using area-detector CT scanners (Aquilion ONE Vision; Canon Medical Systems, Otawara, Japan) with the following scanning parameters: tube voltage, 120 kV; gantry rotation time, 0.5 s; detector collimation, 0.5 mm; and helical pitch (pitch factor), 65.0 (0.813). Automatic exposure control was applied to adjust the tube current and maintain a user-specified noise level in the imaging data; the target noise value (expressed as standard deviation) was set to 11. Additionally, patients were intravenously administered 60–150 mL (600 mg I/kg) of iodinated contrast medium (iohexol 300; Ioverin, Teva Takeda Pharma or iopamidol 370; Oypalomin, Konica Minolta) at a rate of 2.5 mL/s and were scanned at the portal phase (70 s after injection). Axial CT images were reconstructed using 5-mm slice thickness and 512 × 512 matrices, and the field of view was adjusted according to the body size of the patients (minimum, 300 mm; maximum, 420 mm).

Four-dimensional CT (4DCT) techniques

4DCT imaging was performed using 320 area-detector CT scanners (Aquilion One; Canon Medical Systems until November 1, 2019, and thereafter using Aquilion One Vision; Canon Medical Systems). The scanning parameters were as follows: tube voltage, 120 kV; tube current, 80 mA with a rotation time of 0.35 s (29 mAs). Images were captured in 11 phases, covering a 16 cm range from the cranial to the caudal aspect of the target area. The capture lasted 3 s, spanning from the patient’s maximum exhalation to maximum inhalation while intentionally performing chest breathing, and took place between 60 and 180 s after contrast injection. The radiation exposure was recorded as the CT dose index volume of 89.30 mGy and a dose length product of 1428.20 mGy·cm.

In our institution, we determined that the clinical benefits of 4DCT for BR-LAEC outweigh the disadvantages of radiation exposure, and we conducted 4DCT after obtaining informed consent from the patients. The 4DCT imaging protocol was based on a previous study (n = 15), wherein patient consent was obtained which was approved by the institutional review board (IRB no. 21–017).

Imaging evaluationImaging evaluation of initial CT

Patients with BR-LAEC were classified into three categories determined by the tumor board on i-CT: BR1; closer to resectable, BR2; resectability not assessable, and BR3; closer to unresectable. The most severe classification was selected if multiple critical organs showed borderline resectability. The main targets of the resectability assessment were aortic and tracheal/bronchial invasion. Aortic invasion was evaluated based on the Picus angle [15], which is the contact angle between the tumor and aorta, contact length, and boundary unclearness of the contact area between the tumor and the aortic wall. Tracheal and bronchial invasions were evaluated based on the compression and enclosure caused by the tumor and contact length. The detailed evaluation criteria are described below.

Aortic invasion

Supplementary Fig. 1 provides an overview of aortic invasion. All BR1-3 cases were characterized by an unclear contact boundary with wide contact. BR1 was diagnosed when the contact angle was < 90° with a contact length of ≥ 10 mm, or when the contact angle was 90°–110° with a contact length of ≤ 5 mm. BR2 was diagnosed when the contact angle was 90°–110° with a contact length of 5–10 mm, or when the contact angle exceeded 110° with a contact length of ≤ 5 mm. BR3 was diagnosed when the contact angle was 90°–110° with a contact length of 10–20 mm, or when the contact angle exceeded 110° with a contact length of 5–10 mm. Cases were classified as unresectable when the contact angle was 90°–110° with a contact length of ≥ 20 mm, or when the contact angle exceeded 110° with a contact length of > 10 mm.

Tracheal and bronchial invasion

Supplementary Fig. 2 provides an overview of tracheal and bronchial invasion. BR3 was diagnosed when the stratification of the contact boundary with the airway by the primary lesion was unclear, and the airway was significantly compressed, displaced, and narrowed. BR1 was diagnosed when there was a wide contact length with significant compression but without encirclement of the trachea/bronchus or there was a wide contact length with encirclement but no compression. BR2 was diagnosed when the tumor exhibited characteristics of either BR1 or BR3.

Imaging evaluation of 4DCT

In this study, we used software (Vitrea workstation version 7.14.3.197; Canon Medical Systems, Otawara, Japan) to analyze cross-sectional and three-dimensional (3D) image changes associated with respiratory motion. Two radiologists with 12 (S.O.) and 29 (T.Kobayashi) years of experience in oncological diagnostic radiology conducted the evaluation. The main targets were aortic and tracheal/bronchial infiltrations, which were classified as resectable and unresectable, respectively, after discussion. Aortic invasion was assessed based on whether the enclosure of the aorta changed with the respiratory motion. Tracheal/bronchial invasion was assessed based on whether enclosure and compression changed with respiratory motion. The detailed evaluation criteria are described below and in Supplementary Fig. 3. A bar chart was used to depict the correlation among i-CT, 4DCT, and treatment outcomes.

Aortic invasion

First, the appropriate sections were determined. The short-axis image that best displays the contact area between the aorta and tumor was identified, and the slice with a significant contact area and encircling appearance on the axial images was selected. Second, if the tumor encirclement of the aorta changed with respiration, both on the axial slice from the first step and on the sagittal image, non-infiltration was determined.

Tracheal/bronchial invasion

First, the appropriate sections were determined. The short-axis image that best displays the contact area between the trachea/bronchus and tumor was identified, and the slice with a significant contact area and encircling appearance on the axial images were selected. Second, there was a transition to the coronal image from the axial slice in the first step, and the image was adjusted to make the bronchus horizontal. Using the resulting oblique sagittal images, if the compression of the tumor on the bronchial section was alleviated by respiratory movement or if the tumor and bronchus moved independently owing to respiration, non-infiltration was determined.

Histopathological evaluation

Two experienced pathologists (each with > 10 years of experience in oncological pathology) routinely diagnosed the surgically resected specimens following the Union for International Cancer Control 7th edition guidelines. When the evaluation of surrounding organ involvement was difficult on pathology, pT was determined based on intraoperative findings. Patients with pT3 showed invasion of the esophageal adventitia but not the adjacent organs. Patients with pT4a had pericardial, pulmonary, or pleural invasion, whereas those with pT4b had an invasion of the aorta, trachea, bronchus, or major vessels.

The correlation between 4DCT findings and pathological outcomes for evaluating resectability was examined in a surgery-only subgroup. Surgery was performed within a median of 17 (range, 3–39) days after the decision was made. Figure 2 and Supplementary Fig. 4 show the relationship between the radiological resectability of i-CT/4DCT and tumor-node-metastasis (TNM) classification.

Fig. 2

Relationship between radiological resectability evaluation in this study on initial conventional CT and four-dimensional dynamic ventilation CT (4DCT). In cases where there is clear infiltration of vital organs, such as major vessel infiltration or tracheal infiltration, it is classified as “unresectable” (UR). If there is a clear absence of such infiltration, it is classified as “resectable” (R). However, if it is difficult to determine, it is classified as “borderline resectable” (BR). The “BR” category includes cases where vital organ infiltration is suspected. In this study, BR-LAECs were further classified as BR1 (closer to resectable), BR2 (resectability not assessable), or BR3 (closer to unresectable) on initial conventional CT. Subsequently, the resectability on 4DCT was reassessed and diagnosed

Statistical analyses

The Mann–Whitney U and chi-squared tests were used to compare demographic and clinical characteristics for continuous and categorical variables, respectively. For inter-rater agreement of the resectability assessment on 4DCT, 60 patients with suspected aortic (30 patients) and tracheal/bronchial invasion (30 patients) were selected. Two board-certified radiologists (with 12 [S.O.] and 18 [T.H.] years of experience in oncological diagnostic radiology) independently assessed each case. The ratings were as follows: (1), suspected unresectable, (2), likely unresectable but not certain; (3), likely resectable but not certain; and (4), suspected resectable. Inter-rater agreement was quantified using the weighted kappa statistic and a quadratic weighting scheme. The weighted kappa interpretation followed the criteria of Landis et al. [25]: 0.81–1.00, almost perfect; 0.61–0.80, substantial; 0.41–0.60, moderate; 0.21–0.40, fair; and 0.00–0.20, slight.

Statistical analyses were performed using EZR, which is a graphical interface for R [26]. Statistical significance was set at p ≤ 0.05.