Serial CT changes in different components of lung cancer associated with cystic airspace in patients treated with neoadjuvant chemotherapy

The aim of this study was to observe changes in different components (solid, cystic airspace, or entire tumor) in lung cancer associated with cystic airspace following treatment with neoadjuvant chemotherapy (NC), using computerized tomography (CT). We analyzed serial (baseline, first-time follow-up, and last-time follow-up) clinical data and CT imaging in six patients treated with NC. The diameters, areas, and volumes of different tumor components (solid, cystic airspace, and entire tumor) were measured. Delta (Δ) was used to represent changes in these parameters between two examinations: Δ1(%) represents the change from baseline to first follow-up after NC, and Δ2(%) represents the change from baseline to last follow-up after NC. We used the intra-group correlation coefficient (ICC) to test for consistency between parameters as measured by two radiologists. The diameter of solid components in all lesions showed a trend of continuous reduction compared with baseline (Δ1 ranged from − 8.3 to − 46.0%, Δ2 from − 30.8 to − 69.2%). For cystic airspace and entire tumors, different lesions showed different trends over the course of treatment. For diameter, area, and volume, Δ1 of changes in the solid component ranged from − 8.3 to − 46.9%, − 19.4 to − 70.8%, and − 19.1 to − 94.7%, respectively; Δ2 ranged from − 30.8 to − 69.2%, − 50.8 to − 92.1%, and − 32.7 to − 99.8% in diameter, area, and volume, respectively. Results were inconsistent between different components of lung cancer associated with cystic airspace that was treated with NC, but the diameter, area, and volume of solid components were continuously reduced during treatment. Furthermore, area and volume measurements showed more-significant variation than diameter measurements.

CT imaging segmentation and data acquisition. We analyzed CT images from three examinations during NC treatment: baseline, first follow-up examination, and last follow-up examination. Maximum diameters, maximum cross-sectional areas, and volumes of different components (solid tumor, cystic airspace, and entire tumor) of the lesion were measured. The cystic airspace was defined as the gaseous-density shadow, the solid components as soft-density or ground-glass opacity (GGO), including mural nodules, cyst wall, and GGO lesions.
We imported the CT imaging data into ITK-SNAP software (version 3.8; https:// www. fsf. org/) in DICOM format for image segmentation. Two radiologists (Y.K.M. and K.X.) with more than 20 years' combined lung cancer research experience selected the region of interest (ROI) and sketched it layer by layer along the edges of the lesions. The radiologists did not know the details of clinical treatment and survival data. The average measurements they recorded were used in this study's analysis.
Each facet of the lesion area was labeled green, while the cystic airspace was labeled yellow (Fig. 1). We computed the volumes of lesions with different components. The volume of the solid component was equal to that of the entire lesion minus the volume of the cystic airspace. To measure lesion area, the maximum cross-section of the lesion was selected and segmented along the tumor edge. After segmentation, the single-layer ROI volume was automatically output, and we obtained results after calculation according to the formula (ROI area = ROI volume / layer thickness) 14 (Fig. 1). The solid-component area was equal to the whole-lesion area minus the cystic-airspace area. We measured lesion diameter directly using the maximum cross-section of the lesion in accordance with the revised RECIST guideline (version 1.1) 15 .
Delta (Δ) represents changes in different parameters from examination to examination: in which "Baseline" was defined as the last CT before NC, "First Time" as the initial CT examination during NC, and "Last Time" as the last CT examination after NC.

Statistical analysis.
Descriptive results were expressed as percentages (%). We used the intra-group correlation coefficient (ICC) to test the consistency of the quantitative parameters measured by the two radiologists 16 . An ICC value of 0-0.2 indicated poor consistency, 0.21-0.40 indicated average consistency, 0.41-0.60 indicated moderate consistency, 0.61-0.80 indicated relatively good consistency, and 0.81-1.00 indicated good consistency.

Results
Baseline clinical, laboratory, and CT findings. All  www.nature.com/scientificreports/ Three (50%) patients were classified as type I, one (16.7%) as type II, and two (33.3%) as type III 4 . We detected a total of six lesions, five (83.3%) of which were in the right lung. Lesion size ranged from 17.9 to 49.5 mm. Three (50%) patients had ACs, and three (50%) had SSCs. One (16.7%) patient had brain metastasis (M1 stage). Only one (16.7%) patient died during the study. Range of follow-up was 7.0-33.0 months ( Table 1). Table 2 shows the ICCs of measurement parameters in lung cancer associated with cystic airspace as measured by the two radiologists. The results showed good consistency between the two physicians: ICC values of diameter, area, and volume In terms of cystic airspace, treatment had inconsistent effects on different lesions. At the first CT examination during treatment, the cystic airspace of only three lesions had decreased (Δ1 ranged from − 4.6 to − 25.0%), whereas that of the other three lesions had increased (Δ1 ranged from 0.5 to 81.9%). At the last CT examination after treatment, the cystic airspaces of only two lesions were reduced (Δ2, − 7.5% and − 29.5%). www.nature.com/scientificreports/ The maximum whole-lesion diameters of four lesions were reduced at the first CT examination during treatment and at the last one afterward (Δ1 ranged from − 12.6 to − 38.1% and Δ2 from − 28.4 to − 59.6%; Supplementary Table 3, Fig. 3).

ICC of measurement parameters between the two radiologists. Supplementary
Areas of different components. Compared with baseline, solid components in all six cases were reduced (Δ1 and Δ2 were − 1.94% to − 70.8% and − 50.8% to − 92.1%, respectively; Table 2). Interestingly, the cystic-airspace area of only two lesions decreased (Δ1, − 2.4% to − 38.8%; Δ2, − 35.1% to − 71.1%) while that of four increased during Comparisons of different measurement parameters (diameter, area, volume) with serial CT. The magnitude of variation was larger in volume and area than in diameter (   www.nature.com/scientificreports/

Discussion
In this study, we measured different components of lung cancer associated with cystic airspace. Interestingly, the trends in our results were not consistent among solid components, cystic-airspace components, and entire tumors on serial CT scan. Furthermore, area and volume showed significant changes, while diameter did not. We observed additional contrary tendencies in most lesions for solid and cystic-airspace components. Lung cancer associated with cystic airspace is a rare manifestation of the disease. At present, CT is one of the main modalities for cancer staging and for evaluating treatment response and outcomes 17,18 . Most studies have evaluated treatment response by size of solid tumors 19,20 . Furthermore, the RECIST guideline (version 1.1) is recommended for such evaluation 15 . However, the relevant evaluation criteria for cystic lung cancer are rarely discussed, which can confuse oncologists' attempts to assess the criteria for tumor treatment response and therefore influence treatment strategies for individual patients in clinical practice. Our results demonstrated that different measurements of tumor components showed different results and that only the solid component was reduced by NC treatment. We speculate that the change in the solid component might be more correlated with clinical treatment response. In cystic-airspace lung cancer patients treated with NC, this component was more suitable for evaluation than were the cystic-airspace component and total tumor. However, we observed only six patients, and our study did not discuss the correlations of different component results with tumor long-term outcomes. Therefore, these correlations must be further evaluated by survival outcomes in a larger cohort over a longer follow-up time. The results of such studies could help guide the treatment for lung cancer associated with cystic airspace.
Furthermore, when comparing the degrees of variation in the three measurement parameters (tumor diameter, area, and volume), we found that area and volume changes in treatment were more significant than diameter changes. Some studies show that analysis of the largest cross-sectional area and of volume might more accurately depict the burden of complex objects compared with traditional linear size measurement, which has great clinical implications [21][22][23] . This could be because small changes in linear size are amplified by corresponding changes in area and volume. Therefore, area and volume might more comprehensively reflect tumor burden by objective three-dimensional measurement, and so measuring these two parameters in lung cancer associated with cystic airspace could be a useful supplement to RECIST criteria. However, the measurement error of area and volume is large, and the working procedure is complicated, which might limit its clinical application.
Some studies that have investigated the formation of cystic airspace have shown that tumor cell proliferation can lead to the formation of valves in the bronchial lumen 2,24,25 . The small bronchial stenosis prevents partial  www.nature.com/scientificreports/ air exchange, resulting in increased residual-air volume in alveolar space. In our study, some solid components decreased, while some cystic lesions increased, over the course of NC treatment. Solid and cystic-airspace component parameters had contrary tendencies in some lesions. We thought that perhaps the reduction in the solid components resulted in enlargement of the "unidirectional valves. " The parameters of the cystic-airspace component including the total tumor in the lesion were increased. Therefore, we speculate that changes in the cystic-airspace component did not accurately and objectively reflect treatment response. This study had some limitations. First, the number of cases was small; we analyzed the serial CT images of only six patients. Second, patients' NC regimens were inconsistent, which might have affected treatment efficacy. Furthermore, two patients were T1N1M0, and one was T2N1M0. We speculate that perhaps the oncologist used an aggressive treatment strategy in clinical practice. Third, such patients' actual conditions and attitudes toward treatment might influence the oncologist's treatment decisions. Finally, the pathological lung cancer types of the enrolled patients were not consistent, so we did not perform different subgroup analyses for different pathological types.
In conclusion, after neoadjuvant chemotherapy in lung cancer with cystic airspace, measurements of different components showed inconsistent results, with only the solid component reduced during treatment. Area and volume measurements changed significantly compared with diameter measurements over the course of treatment.