Early-stage lung adenocarcinoma exists on a spectrum from completely non-invasive precursors to frankly invasive cancers, and distinguishing between these stages has profound implications for treatment. Atypical adenomatous hyperplasia (AAH) and adenocarcinoma in situ (AIS) are non-invasive lesions with 100% five-year survival after surgical removal, while minimally invasive adenocarcinoma (MIA) and fully invasive adenocarcinoma require more aggressive surgery and carry meaningfully worse prognoses.
The challenge is that non-invasive lung lesions sometimes contain solid-appearing components on CT scans that look identical to invasive cancer. When CT shows a part-solid ground-glass nodule - a fuzzy haziness with a denser solid center - that solid center is usually cancer invading surrounding tissue. But in up to one-third of cases, that apparent solid component may actually be alveolar collapse: a benign process where the delicate air sacs (alveoli) that make up normal lung tissue become compacted with elastic and collagen fibers, creating density on CT without any cancer invasion at all.
Alveolar collapse has been recognized as a biological phenomenon in lung adenocarcinoma since 1980, when Shimosato and colleagues first described its distinct histological appearance. Noguchi et al. later formalized the classification: type A and B tumors (which include alveolar collapse without true invasion) show zero lymph node metastases and 100% five-year survival, while type C tumors with active fibroblast invasion have only 74.8% five-year survival. The diagnostic stakes of correctly distinguishing collapse from invasion are therefore extremely high.
The consolidation/tumor (C/T) ratio has been the dominant CT-based tool for assessing lung adenocarcinoma invasiveness for over a decade. It is calculated by dividing the diameter of the solid (consolidation) component by the total tumor diameter. A C/T ratio of 0.25 or lower is used by major Japanese clinical guidelines to predict non-invasive disease, and this threshold was validated in the landmark JCOG 0201 trial.
However, the C/T ratio has a critical limitation: it uses only one-dimensional diameter measurements. Real tumors are three-dimensional, irregular, and non-spherical. Measuring a single diameter - even the longest one - may not accurately represent the true proportion of solid to non-solid tissue within an irregularly shaped nodule. An elongated tumor viewed from one axis might show a very different C/T ratio than when viewed from another, leading to inconsistent measurements.
The JCOG 0201 trial itself acknowledged this problem: while the C/T ratio criterion achieved excellent specificity (98.7%) for predicting non-invasiveness, its sensitivity was only 16.2%. Among 254 lesions with C/T ratios above 0.25 - classified as invasive - 176 (69%) turned out to be pathologically non-invasive after surgery. This massive overestimation of invasiveness is largely attributable to alveolar collapse creating misleadingly high solid components on CT.
This study from Mie University, Japan, retrospectively analyzed 161 patients preoperatively diagnosed with clinical stage IA1 lung cancer (the earliest stage) who underwent surgical resection between January 2019 and December 2024. All patients had adenocarcinoma histology. After surgery, pathological examination classified 50 patients as non-invasive (AAH or AIS) and 111 as invasive (MIA or fully invasive adenocarcinoma).
The key innovation was three-dimensional volumetric CT analysis performed on dedicated radiology workstation software (SYNAPSE VINCENT, FUJIFILM). Rather than measuring a single diameter, this approach semi-automatically reconstructs the complete 3D shape of both the total tumor and the solid component, using CT density thresholds (values greater than -300 Hounsfield units indicate solid tissue) to separate solid from non-solid regions. This produces absolute volume measurements for both components.
From these 3D measurements, the researchers calculated the 3D-C/T ratio (solid component volume divided by total tumor volume), which is the three-dimensional analogue of the conventional linear C/T ratio. They compared the predictive performance of four measurements: conventional consolidation diameter (a simple length), conventional C/T ratio, absolute consolidation volume (a 3D measurement), and the 3D-C/T ratio. A histological review of 28 non-invasive cases was also performed to characterize the microscopic makeup of their solid-appearing CT components.
All four CT measurements successfully distinguished non-invasive from invasive lesions in univariate analysis, but they performed very differently in terms of diagnostic accuracy. The AUC values from receiver operating characteristic (ROC) analysis were: consolidation diameter 0.702, C/T ratio 0.634, consolidation volume 0.747, and 3D-C/T ratio 0.742. All three alternatives outperformed the conventional C/T ratio, which had the lowest AUC of the group.
Statistical testing confirmed these differences were not due to chance: DeLong's test showed significant differences in AUC between consolidation diameter and C/T ratio (p=0.0415), and between 3D-C/T ratio and C/T ratio (p=0.0208). The consolidation volume was the only independent predictor of invasiveness in multivariate logistic regression analysis adjusting for all other imaging variables (odds ratio 1.002 per cubic millimeter increase, p=0.045).
Practical cutoff values were identified for clinical use: a consolidation diameter of 4.5 mm (67.6% sensitivity, 70.0% specificity), a consolidation volume of 173.4 cubic millimeters (64.9% sensitivity, 78.0% specificity), and a 3D-C/T ratio of 0.121 (72.1% sensitivity, 66.0% specificity). These values perform better than the historical C/T ratio of 0.25 while offering the advantage of 3D measurement accuracy that accounts for irregular tumor geometry.
The histological review of 28 non-invasive cases with solid-appearing CT components revealed that alveolar collapse was present in 89.2% of cases - 25 of the 28 patients. This rate confirms that when non-invasive early-stage lung adenocarcinoma appears to have a solid component on CT, the overwhelming cause is alveolar collapse rather than any true cancer invasion.
Microscopically, alveolar collapse in these cases consisted primarily of elastic fibers, found in 96% of alveolar collapse cases (24 of 25), with collagen fibers present in 76% (19 of 25) and both fiber types together in 76%. This fiber-based architecture - a compaction of the normal lung scaffolding without cancer cell invasion of vessels, septa, or bronchi - creates CT density without malignancy. Four cases showed particularly dense perivascular fiber accumulation around blood vessels within the tumor.
These pathological findings explain why conventional CT measurement fails: alveolar collapse creates a solid-appearing component that is indistinguishable from invasive cancer by visual CT inspection and by the one-dimensional C/T ratio. However, the 3D volumetric measurements reflect the actual extent of this solid component more accurately, and combined with knowledge of the collapse phenomenon, help clinicians avoid unnecessary surgical over-treatment of truly non-invasive lesions.
Correctly identifying non-invasive lesions has direct surgical implications. Non-invasive lung adenocarcinoma (AAH and AIS) can be treated with limited resection - wedge resection or segmentectomy - that preserves more lung function. Invasive adenocarcinoma typically requires more extensive lobectomy. In patients with limited lung reserve, the difference between these approaches can determine whether a patient maintains acceptable lung function after surgery.
The finding that 31.1% of clinical stage IA1 patients (50 of 161) turned out to be non-invasive after surgery highlights the scale of potential over-treatment. If better imaging tools could identify even a portion of these patients preoperatively, it could shift the surgical approach from lobectomy to segmentectomy, reducing operative risk and preserving lung function without compromising cancer control in truly non-invasive disease.
The authors also raise the emerging concept that not all lung adenocarcinoma precursors follow a linear progression path to invasive cancer. Research suggests that KRAS-mutated atypical adenomatous hyperplasia rarely progresses to invasive cancer, while EGFR-mutated lesions are distributed across all stages of progression. This molecular heterogeneity means that some lesions with benign alveolar collapse might never progress, making the case for accurate non-invasive characterization even stronger as a guide to active surveillance versus surgical intervention.
Despite 3D volumetric measurements outperforming the conventional C/T ratio, none of the assessed measures achieved diagnostic accuracy adequate for clinical decision-making in isolation. The AUC values of 0.70 to 0.75 reflect the fundamental challenge that alveolar collapse and true invasive lesions occupy overlapping size ranges and volumetric proportions on CT, particularly in small or mildly solid lesions.
The authors propose machine learning as the most promising path forward. Subtle CT features that distinguish alveolar collapse from invasive lesions - including the homogeneity of density within the solid component, the sharpness of the border between solid and non-solid regions, and the spatial distribution pattern of the consolidation - exist in the imaging data but are too subtle for reliable visual assessment or simple measurement. Machine learning approaches trained to detect and quantify these three-dimensional texture differences could potentially achieve higher diagnostic accuracy than any single measurement metric.
The study has important limitations that require acknowledgment: it is single-center and retrospective with only 161 patients, observer variability in 3D segmentation was not formally tested, and the possibility of iatrogenic (procedure-induced) collapse from intraoperative lung deflation was acknowledged though steps were taken to minimize it by immediately fixing resected specimens with formalin. Multicenter prospective validation with standardized imaging protocols and inter-observer reliability testing will be essential before 3D volumetric analysis could be adopted in routine clinical practice.