Deep learning-based computed tomography quantification integrated with circulating tumor cells for prognostic evaluation in stage I lung adenocarcinoma
Original Article

Deep learning-based computed tomography quantification integrated with circulating tumor cells for prognostic evaluation in stage I lung adenocarcinoma

Lujie Li1#, Meicheng Chen1#, Ji Zhu2#, Ling Ma1, Fangzeng Lin1, Xiang-Min Li3, Qiong Li4,5, Ying Zhu1

1Department of Radiology, The First Affiliated Hospital of Sun Yat-sen University, Guangzhou, China; 2Department of Thoracic Surgery, Changhai Hospital, Naval Medical University, Shanghai, China; 3Department of Radiology, Hui Ya Hospital of The First Affiliated Hospital, Sun Yat-sen University, Huizhou, China; 4Department of Radiology, Sun Yat-sen University Cancer Center, Guangzhou, China; 5State Key Laboratory of Oncology in South China, Collaborative Innovation Center for Cancer Medicine, Guangzhou, China

Contributions: (I) Conception and design: Y Zhu, XM Li; (II) Administrative support: L Ma, Q Li; (III) Provision of study materials or patients: J Zhu, Y Zhu; (IV) Collection and assembly of data: M Chen; (V) Data analysis and interpretation: L Li, F Lin; (VI) Manuscript writing: All authors; (VII) Final approval of manuscript: All authors.

#These authors contributed equally to this work.

Correspondence to: Qiong Li, MD. Department of Radiology, Sun Yat-sen University Cancer Center, Guangzhou, China; State Key Laboratory of Oncology in South China, Collaborative Innovation Center for Cancer Medicine, No. 651 Dongfeng Road East, Guangzhou 510060, China. Email: liqiong@sysucc.org.cn; Ying Zhu, MD. Department of Radiology, The First Affiliated Hospital of Sun Yat-sen University, No. 58 Zhongshan Er Road, Guangzhou 510080, China. Email: zhuy45@mail.sysu.edu.cn.

Background: Prognosis varies among patients with early-stage lung adenocarcinoma (LUAD) even within the same clinical stage. This study aimed to evaluate the prognostic value of measurements of computed tomography (CT), circulating tumor cells (CTC), and their combinations in clinical stage I LUAD.

Methods: This retrospective analysis included 183 patients at stage I peripheral LUAD who underwent preoperative thin-section chest CT and CTC detection. The maximal solid size to tumor ratio on axial image was measured by radiologists (R-MSSA%). The three-dimensional volume (3D-SV) and mass (3D-SM) of solid components were measured by deep learning, followed by calculation of the consolidation-to-tumor ratios of 3D-SV (3D-SV%) and 3D-SM (3D-SM%). The differences in recurrence rates or overall survival of CTC, CT measurements, and their combinations were compared. Multivariate Cox proportional hazards models were used to identify independent risk factors.

Results: Our study results indicated that 3D-SM, 3D-SV, 3D-SM%, and CTC can stratify the recurrence risk in clinical stage I patients. For clinical stage IA patients with high values of 3D-SM, 3D-SV, R-MSSA%, or 3D-SM%, the recurrence risk showed no significant difference compared with stage IB patients. Furthermore, for stage IA patients stratified as high-risk by the integration of CTC with 3D-SM or 3D-SV, the prognosis was worse than that of stage IB patients. A model combining 3D-SM and CTC had a higher C-index [0.75; 95% confidence interval (CI): 0.70, 0.81] compared to that of 3D-SM or CTC alone.

Conclusions: The integration of CTC and CT measurements can enhance the stratification ability for early-stage lung cancer patients and improve the prognosis prediction, especially for those at stage IA.

Keywords: Lung adenocarcinoma (LUAD); circulating tumor cells (CTC); computed tomography (CT); prognosis

Submitted Nov 06, 2025. Accepted for publication May 06, 2026. Published online May 22, 2026.

doi: 10.21037/qims-2025-aw-2342

Introduction

Lung cancer, a leading cause of cancer-related mortality, has relatively better prognosis in early stage. Anatomical lung resection surgery is commonly recognized as the typical treatment for clinical stage IA patients of lung adenocarcinomas (LUAD), and adjuvant chemotherapy may be necessary for some patients with stage IB after surgery (1-3). However, patients with the same clinical stage demonstrate variability in prognosis (4-6). A subset of stage IA patients still harbors a significant risk of late recurrence (7-9). Therefore, a comprehensive understanding of characteristics is critical for patients of stage IA LUAD with high recurrence risk, allowing precise risk stratification, and supporting individualized treatment decision as well as follow-up strategies.

The stage evaluation based on computed tomography (CT) images is widely applied for treatment decision. Currently, the T-categories are primarily determined by the longest diameter of solid part on axial image. Moreover, the consolidation-to-tumor ratio (CTR) has been adopted as a critical reference for surgical approach determination (10,11), while a favorable prognosis is often indicated by a CTR ≤0.5 (12,13). However, its accuracy is affected by the misunderstanding of the boundary between the solid and ground-glass components (14,15) and inter- or intra-observer variation (16,17). Compared with traditional manual measurement, measurement by automatic artificial intelligence (AI) is more convenient and reproducible, and provides more information about lesions (18-20).

Early-stage LUAD is associated with a more favorable prognosis, but approximately 20% of patients who undergo surgery are likely to relapse within 5 years (21,22). Circulating tumor cells (CTC), regarded as the foundation of metastasis, may play an important role in metastatic processes (23,24). Many studies have demonstrated the ability of CTC to predict prognosis in early-stage LUAD (25,26). However, the limited sensitivity restricts clinical application (25,27). The integration of CTC and radiomics may be useful for enhancing prognostic prediction (28,29). However, no research has yet combined CT quantification with CTC analysis for early-stage LUAD.

In this study, we retrospectively analyzed the performance of AI-based CT measurement, CTC, and their combination for predicting outcomes in early-stage LUAD patients. We hypothesized that the combination of radiological tumor measurements and CTC can improve prognostic prediction, which is superior to any single method, including traditional clinical stage, radiological measurements alone, or CTC alone. We present this article in accordance with the TRIPOD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-aw-2342/rc).

Methods

Study population

The retrospective study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of The First Affiliated Hospital of Sun Yat-sen University (SYSUFA) (No. [2021]-531), and individual consent for this retrospective analysis was waived. Medical records of patients who underwent lobectomy surgery with clinical stage I peripheral LUAD at SYSUFA between April 2015 and April 2016 were reviewed. All patients underwent thin-slice chest CT examination and CTC detection within 2 weeks before surgery and prior to any anti-tumor treatment. Fifty-one patients with synchronous or metachronous LUAD, cystic lung cancer, or other malignancies were excluded. Finally, a total of 183 cases were included for analysis.

Relapse-free survival (RFS) is defined as the time to relapse or death after surgery, or the most recent date of follow-up. Overall survival (OS) is defined as the time interval between the date of surgery and the date of death.

Radiological tumor assessment

CT scan was performed using a 64-row multidetector CT scanner (Aquilion 64, Canon, Japan). Scan parameters were as follows: tube voltage 120 kV and tube current 200 mAs; beam pitch, 0.828; rotation time, 0.5 seconds; beam collimation, 64×0.5 mm. Axial thin-section multidetector CT images of the whole lung were reconstructed with a slice thickness and spacing of 1 mm using high-spatial frequency algorithm.

The longest diameters of the solid component and the total tumor were measured on the axial image by two chest radiologists blinded to clinical data (with 5 and 15 years of experience, respectively) using standard lung window setting (window level, − 500 HU; window width, 1,500 HU). The average of the measurements was then calculated for further analysis. Then maximal solid size to tumor ratio on axial image by radiologists (R-MSSA%) was calculated.

Three-dimensional solid mass (3D-SM), 3D-solid volume (3D-SV), total tumor mass and total tumor volume were measured by AI as we studied previously (20). A deep learning network named DenseSharp was used for lesion segmentation, and we applied a density level of 0 HU as the threshold for separating the solid part and the ground-glass opacity component. The volume of solid component to tumor ratio (3D-SV%) and the mass of solid component to tumor ratio (3D-SM%) were further calculated. Detailed methods were same as previous study (20).

CTC detection

Peripheral blood samples (7.5 mL) from patients were obtained up to one week before surgery. CTC was enriched by the aptamer-modified NanoVelcro system, and the details were consistent with the previous research (30,31). Two experienced investigators analyzed chips independently under fluorescence microscopy. Finally, the range of CTC was 0 to 13.

Statistical analysis

Statistical analyses were performed with R software version 4.3.1 (R Foundation for Statistical Computing). Time-dependent receiver operating characteristic (ROC) curve was analyzed by R package timeROC. All survival analyses were dependent on R packages survival and survminer. Package survminer was used to calculate the cutoff value. Survivals were plotted by Kaplan-Meier curves, and differences were compared by log-rank test. The Bonferroni method was used to adjust the P value in multiple comparisons. The variance inflation factor was used to assess multicollinearity among variables. Cox proportional hazards models were used, and then hazard ratios (HRs) and C-index were calculated. R package caret was used for internal cross-validation. R package CsChange was used to compare the difference between C-index, and R package survIDINRI was used to calculate integrated discrimination improvement (IDI). A two-sided P value less than 0.05 was considered statistically significant.

Results

Patients’ characteristics

The characteristics of all patients in the present study are summarized in Table 1. A total of 183 patients were diagnosed with peripheral LUAD, of whom 151 patients were clinical stage IA {median age, 60 [interquartile range (IQR), 52, 67] years; 73 men} and 32 patients were clinical stage IB [median age, 61 (IQR, 54, 70) years; 18 men]. The median 3D-SV was 984.60 (IQR, 195.02, 2,845.34) mm3, while the median 3D-SM was 1,038.31 (IQR, 205.29, 3,034.92) mm3. The median follow-up time for RFS was 1,591 days, and recurrences occurred in 57 patients (31.15%). The median follow-up time for OS was 1,743 days.

Table 1

Clinicopathological characteristics of patients

Characteristic Value
Age
   <60 years 86 (46.99)
   ≥60 years 97 (53.01)
Sex
   Female 92 (50.27)
   Male 91 (49.73)
Smoking status
   Nonsmoker 147 (80.33)
   smoker 36 (19.67)
CTC 1 (0, 3)
3D-SV (mm3) 984.60 (195.02, 2,845.34)
3D-SM (mm3) 1,038.31 (205.29, 3,034.92)
3D-SV% 0.28 (0.09, 0.44)
3D-SM% 0.38 (0.16, 0.52)
R-MSSA% 0.93 (0.50, 1.00)
Nodule type
   GGN 102 (55.74)
   Pure solid nodule 81 (44.26)
Clinical stage
   IA 151 (82.51)
   IB 32 (17.49)
T stage
   ≤10 mm 48 (26.23)
   >10 mm but ≤20 mm 62 (33.88)
   >20 mm but ≤30 mm 41 (22.40)
   >30 mm but ≤40 mm 32 (17.49)
Pathologic type
   AIS/MIA 3 (1.64)
   IASLC grade 1 21 (11.48)
   IASLC grade 2 123 (67.21)
   IASLC grade 3 30 (16.39)
   Mucinous adenocarcinoma 6 (3.28)
Pathologic pleural invasion
   Negative 156 (85.25)
   Positive 27 (14.75)
Pathologic vascular invasion
   Negative 176 (96.17)
   Positive 7 (3.83)
Pathologic lymph node metastasis
   Negative 181 (98.91)
   Positive 2 (1.09)

Data are presented as median (interquartile range) or n (%). 3D, three-dimensional; 3D-SM, 3D-solid mass; 3D-SM%, 3D-consolidation tumor ratio of mass; 3D-SV, 3D solid volume; 3D-SV%, 3D-consolidation tumor ratio of volume; AIS/MIA, adenocarcinomas in situ or minimally invasive adenocarcinomas; CTC, circulating cancer cells; GGN, ground-glass nodules; IASLC, International Association for the Study of Lung Cancer; R-MSSA%, maximal solid size on axial image measured by radiologists.

In terms of pathological features, 3 (1.64%) were adenocarcinomas in situ (AIS) or minimally invasive adenocarcinomas (MIA), 21 (11.48%) were the International Association for the Study of Lung cancer (IASLC) grade 1, 123 (67.21%) were IASLC grade 2, 30 (16.39%) were IASLC grade 3, and 6 (3.28%) were mucinous adenocarcinoma. Pleural invasion was observed in 27 patients (14.75%), vascular invasion was observed in 7 patients (3.83%), and lymph node metastasis was observed in 2 patients (1.09%).

Survival risk stratification

Time-dependent ROC analysis was performed to predict recurrence (Table 2). The 3- and 5-year areas under the curve (AUCs) for RFS of 3D-SM [3-year AUC: 0.701 (95% CI: 0.625–0.776); 5-year AUC: 0.688 (95% CI: 0.543–0.833)] showed similarity with those of 3D-SV [3-year AUC: 0.700 (95% CI: 0.624–0.775); 5-year AUC: 0.686 (95% CI: 0.541–0.832)]. In contrast, the 3- and 5-year AUC of CTC was 0.566 (95% CI: 0.449–0.683) and 0.615 (95% CI: 0.493–0.736), respectively, lower than those of radiologic measurements. The optimal cut-off values for 3D-SV and 3D-SM were determined to be 800 and 864 mm3, respectively, and the optimal cut-off value of solid volume was 4. Consequently, patients were stratified into two groups.

Table 2

Results of time-depending ROC analysis

Characteristic RFS OS
3-year AUC (95% CI) 5-year AUC (95% CI) 3-year AUC (95% CI) 5-year AUC (95% CI)
3D-SV 0.700 (0.624–0.775) 0.686 (0.541–0.832) 0.692 (0.582–0.801) 0.617 (0.493–0.741)
3D-SM 0.701 (0.625–0.776) 0.688 (0.543–0.833) 0.693 (0.583–0.802) 0.617 (0.493–0.741)
3D-SV% 0.696 (0.617–0.775) 0.714 (0.574–0.854) 0.709 (0.607–0.810) 0.633 (0.510–0.756)
3D-SM% 0.696 (0.617–0.775) 0.701 (0.561–0.841) 0.705 (0.603–0.807) 0.631 (0.507–0.754)
CTC 0.566 (0.449–0.683) 0.615 (0.493–0.736) 0.609 (0.448–0.770) 0.572 (0.444–0.701)

3D, three-dimensional; 3D-SM, 3D-solid mass; 3D-SM%, 3D-consolidation tumor ratio of mass; 3D-SV, 3D-solid volume; 3D-SV%, 3D-consolidation tumor ratio of volume; AUC, area under the ROC curve; CI, confidence interval; CTC, circulating cancer cells; OS, overall survival; RFS, relapse-free survival; ROC, receiver operating characteristic.

The predictive performance of survival outcomes was assessed in the Kaplan-Meier curves (Figure 1), indicating a significant difference in recurrent survival between patients with high and low 3D-SV, 3D-SM, and CTC (P<0.001). However, no significant difference was observed in recurrent survival between patients with clinical stage (P=0.32). Specifically, the 5-year RFS proportions were 79.62% and 49.11% for low 3D-SV (0< 3D-SV <800 mm3) and high 3D-SV (3D-SV ≥800 mm3), while they were 79.65% and 48.92% for low 3D-SM (3D-SM <864 mm3) and high 3D-SM (3D-SM ≥864 mm3), respectively. Moreover, the 5-year RFS proportions for patients with low CTC levels (CTC <4) and high CTC levels (CTC ≥4) were 73.94% and 24.10%, respectively.

Figure 1 The Kaplan-Meier curves of RFS for all patients according to 3D-SV (A), 3D-SM (B), R-MSSA% (C), 3D-SV% (D) and 3D-SM% (E), CTC (F), and clinical stage (G). Log-rank tests were used to compute P values. 3D, three-dimensional; 3D-SM, 3D-solid mass; 3D-SM%, 3D-consolidation tumor ratio of mass; 3D-SV, 3D-solid volume; 3D-SV%, 3D-consolidation tumor ratio of volume; CTC, circulating cancer cells; R-MSSA%, maximal solid size on axial image measured by radiologists; RFS, relapse-free survival.

Figure 1 also shows the Kaplan-Meier curves of RFS for CTRs, including 3D-SM%, 3D-SV%, and R-MSSA%, with 0.5 as the cut-off value. All 3D-SM% and 3D-SV% were less than 0.8. Patients with higher 3D-SM% and R-MSSA% exhibited a significantly worse prognosis compared to those with lower 3D-SM% (P=0.0068) and R-MSSA% (P<0.0001). However, no significant difference was observed for 3D-SV% based on the log-rank test (P=0.24).

Additionally, Figure S1 displays the Kaplan-Meier curves of OS. Significant differences were observed in clinical stage (P=0.03) and all measurements, including 3D-SV (P=0.00023), 3D-SM (P<0.0001), R-MSSA% (P=0.00088), 3D-SV% (P=0.031), 3D-SM% (P=0.0039), and CTC (P<0.0001).

Survival risk stratification in patients with stage IA

In this study, clinical staging was not predictive of RFS. Therefore, we further investigated the roles of 3D-SM, 3D-SV, 3D-SM%, R-MSSA%, and CTC in stage IA patients compared with stage IB patients (Figure 2). It shows that there was no significant difference between stage IA patients with high 3D-SV (P=0.51), 3D-SM (P=0.49), 3D-SM% (P=1.00) and R-MSSA% (P=1.00) and stage IB patients. Specifically, in stage IA, 66/151 cases with higher 3D-SM or 3D-SV, and 106/151 cases with higher R-MSSA%, exhibited no significant difference of recurrent outcomes compared with stage IB patients. Furthermore, no significant difference was shown in stage IA patients with low 3D-SM% and stage IB patients (P=0.292). In contrast, 3D-SV% did not show a significant difference among all groups (P=0.6). When comparing CTC subgroups, patients with high CTC in stage IA showed a trend of poor prognosis compared to stage IB group, although the difference was not significant (P=0.12).

Figure 2 The Kaplan-Meier curves of RFS for patients with clinical stage IB and patients with stage IA according to 3D-SV (A), 3D-SM (B), CTC (C), and 3D-SV% (D), 3D-SM% (E) and R-MSSA% (F). P value, computed by log-rank test for all groups; P1, the P value of difference between IB group and low measurements subgroup in IA stage; P2, the P value of difference between IB group and high measurements subgroup in IA stage. 3D, three-dimensional; 3D-SM, 3D-solid mass; 3D-SM%, 3D-consolidation tumor ratio of mass; 3D-SV, 3D-solid volume; 3D-SV%, 3D-consolidation tumor ratio of volume; CTC, circulating cancer cells; R-MSSA%, maximal solid size on axial image measured by radiologists; RFS, relapse-free survival.

Additionally, Figure S2 displayed the Kaplan-Meier curves of OS. No significant difference was observed in stage IA patients with high 3D-SM, 3D-SV, CTC, or 3D-SM% and stage IB patients.

Combination of CTC and radiological tumor measurement predicts clinical outcome

Given the established association between tumor size and recurrence, we investigated whether the combination of radiological measurements and CTC could improve the prediction of recurrence (Figure 3). Patients in stage IA were further stratified into three risk groups: low risk for patients with both low CTC and low radiological measurement, medium risk for patients with either high CTC or high radiological measurement, and high risk for patients with both high CTC and high radiological measurement.

Figure 3 The Kaplan-Meier curves of RFS for patients with clinical stage IB and patients with stage IA according to CTC combined with 3D-SM (A), 3D-SV (B), 3D-SM% (C), or R-MSSA% (D). Low risk: both low CTC and radiological measurement; medium risk: either high CTC or high radiological measurement; high risk: both high CTC and high radiological measurement. P value, computed by log-rank test for all groups; P1, the P value of difference between IB group and medium risk IA subgroup; P2, the P value of difference between IB group and high-risk stage IA subgroup. 3D, three-dimensional; 3D-SM, 3D-solid mass; 3D-SM%, 3D-consolidation tumor ratio of mass; 3D-SV, 3D-solid volume; 3D-SV%, 3D-consolidation tumor ratio of volume; CTC, circulating cancer cells; R-MSSA%, maximal solid size on axial image measured by radiologists; RFS, relapse-free survival.

When integrating CTC with 3D-SM, 3D-SV, 3D-SM% or R-MSSA%, low-risk group showed a significant difference compared with other groups. When investigating CTC combined with 3D-SM (P=0.013) or 3D-SV (P=0.013), high-risk stage IA group had a significantly worse prognosis compared to stage IB group, and no significant difference was observed between the medium-risk stage IA group and stage IB group (P3D-SM >0.99, P3D-SV >0.99). When investigating the combination of CTC with 3D-SM% or R-MSSA%, no significant difference was observed between the high-risk stage IA group and stage IB group (P3D-SM% =0.11, PR-MSSA% =0.12), and no significant difference between the medium-risk stage IA group and stage IB group (P3D-SM% >0.99, PR-MSSA% >0.99). The representative image of each group is shown in Figure 4.

Figure 4 The representative cases. Red indicates the opacity of GGOs, and green indicates the solid component. (A-D) Patients with stage IA lung adenocarcinoma; (E) a patient with stage IB lung adenocarcinoma. The CTC counts of (A) and (B) are both 0, while those of (C) and (D) are 5 and 6 respectively. Additionally, (A) and (C) show minimal solid components. During follow-up, no recurrence was observed in patients of (A-C), while (D) showed recurrence at 837 days and (E) at 1,767 days of follow-up. CTC, circulating cancer cells; GGOs, ground-glass opacities.

Additionally, Figure S3 displayed the Kaplan-Meier curves of OS. When integrating CTC with 3D-SM, 3D-SV, 3D-SM% or R-MSSA%, low-risk stage IA group showed a significant difference compared with other groups (P<0.05). However, no significant difference was observed between the high-risk stage IA group and stage IB group, or medium-risk stage IA group and stage IB group (P>0.05).

Prognostic value of parameters

Univariate and multivariate survival analyses were performed to evaluate these variables for RFS. In the univariate analysis (Table 3), CTC (P<0.001), 3D-SV (P<0.001), 3D-SM (P<0.001), 3D-SM% (P=0.008), R-MSSA% (P=0.002), age (P=0.01), smoking status (P=0.005), nodule type (P<0.001), pathologic type (P=0.003), vascular invasion (P=0.003), and lymph node metastasis (P=0.02) were identified as risk factors for recurrence. Since collinearity analysis revealed that 3D-SV and 3D-SM had collinearity (variance inflation factor >10), only 3D-SM was included in multivariate Cox regression analysis. In the multivariate survival analysis, age (HR: 2.04, P=0.017), 3D-SM (HR: 3.64, P=0.001), and CTC (HR: 2.61, P=0.001) were identified as independent prognostic factors for RFS (Table 3).

Table 3

Univariate and multivariate survival analysis for RFS

Characteristic Risk factor Univariate analysis Multivariate analysis
HR 95% CI P HR 95% CI P
Age >60 years 1.97 1.15, 3.40 0.01 2.04 1.13, 3.67 0.017*
Sex Male 1.96 1.13, 3.37 0.02 1.12 0.58, 2.18 0.7
Smoking status Positive 2.35 1.34, 4.12 0.003 1.57 0.80, 3.08 0.2
3D-SM ≥864 mm3 5.42 2.73, 10.74 <0.0001 3.64 1.66, 7.98 0.001**
3D-SV ≥800 mm3 5.38 2.72, 10.67 <0.0001 –a – –
3D-SM% >0.5 2.05 1.21, 3.48 0.008 0.71 0.34, 1.49 0.4
R-MSSA% >0.5 22.2 3.08, 161 0.002 6.53 0.80, 53.5 0.080
CTC ≥4 3.98 2.35, 6.75 <0.0001 2.61 1.46, 4.66 0.001**
Nodule type Solid 3.05 1.75, 5.29 <0.0001 1.47 0.77, 2.81 0.2
Pathologic type – 1.72 1.20, 2.44 0.003 1.05 0.65, 1.69 0.9
Vascular invasion Positive 4.00 1.58, 10.09 0.003 1.09 0.36, 3.26 0.9
Lymph node metastasis Positive 5.30 1.28, 22.04 0.022 2.68 0.58, 12.4 0.2

a, since 3D-SV and 3D-SM had collinearity, 3D-SV was not included in multivariate cox regression analysis. *, P value <0.05 in multivariate analysis; **, P value <0.01 in multivariate analysis. 3D, three-dimensional; 3D-SM, 3D-solid mass; 3D-SM%, 3D-consolidation tumor ratio of mass; 3D-SV, 3D-solid volume; CI, confidence interval; CTC, circulating cancer cells; HR, hazard ratio; R-MSSA%, maximal solid size to tumor ratio on axial image by radiologists; RFS, relapse-free survival.

Cox regression models for RFS were also conducted to evaluate the prediction effect of radiological measurements and CTC (Table 4). The combined model with 3D-SM and CTC obtained a C-index of 0.755 (95% CI: 0.702–0.808), which was significantly higher than that of 3D-SM alone (P=0.005) or CTC alone (P<0.001). Internal 10-fold cross-validation of 3D-SM combined CTC resulted in a C-index of 0.729. The IDI showed that the discrimination power was significantly improved by the combined model when compared with 3D-SM alone [IDI =9.5% (95% CI: 1.5–21.0%)] or CTC alone [IDI =15.0% (95% CI: 5.2–26.8%)].

Table 4

Performance of multivariate Cox regression models of RFS for all patients

Characteristic Models, odds ratio (95% CI)
3D-SM CTC 3D-SM + CTC
3D-SM
   0< 3D-SM <864 mm3 Reference – Reference
   3D-SM ≥864 mm3 5.42 (2.73, 10.74) – 4.96 (2.50, 9.87)
CTC
   CTC <4 – Reference Reference
   CTC ≥4 – 3.98 (2.35, 6.75) 3.52 (2.07, 5.99)
C-index (95% CI) 0.70 (0.65, 0.76) 0.61 (0.55, 0.68) 0.75 (0.70, 0.81)
   Difference 0.05 0.14 Reference
   P value 0.005 <0.001 –
C-index (internal 10-fold cross-validation) 0.717 0.682 0.729
   IDI (%) −9.5 (−21.0, −1.5) −15.0 (−26.8, −5.2) Reference
   P value 0.018 <0.001 –

3D, three-dimensional; 3D-SM, 3D-solid mass; CI, confidence interval; CTC, circulating cancer cells; IDI, integrated discrimination improvement; RFS, relapse-free survival.

Discussion

In this study, we aim to investigate the ability of CT measurements and CTC in predicting the prognosis and whether their combinations can improve prognostic stratification in early-stage LUAD. The results showed that CT measurements, including 3D-SM, 3D-SV, R-MSSA%, 3D-SM%, and their pairwise combinations with CTC, could re-stratify the risk of recurrence in stage IA. 3D-SM and CTC were independent risk factors for recurrence, and the combination has higher prediction ability than that of either individually.

In this study, 3D-SM, 3D-SV, and CTC help to stratify the risk of recurrence in patients with stage IA. Patients with high 3D-SM and CTC showed a higher risk of recurrence than patients with IB. Although, according to the current clinical stage, stage I LUAD has a better prognosis, a large degree of variability still exists (32,33). Therefore, further risk stratification is essential for treatment decisions (10,34). Several studies have attempted to stratify recurrent risk in stage IA patients depending on radiomics (35,36). However, standardization and interpretability are one of the challenges that radiomics must overcome for clinical deployment (37). The present study combines CT measurements with CTC, which have been standardized. It suggests that these patients with high 3D-SM and CTC are at higher risk of recurrence and metastasis, requiring more aggressive interventions, such as adjuvant therapy and surveillance.

It is hypothesized that there is a correlation between CTC and the size of the tumor (38). But both the higher CTC and higher tumor volume/mass were independent risk factors associated with recurrence in the present study. Several studies have also demonstrated that there is no simple linear correlation between CTC count and pathological stage (25). Furthermore, the combination of CTC and 3D-SM may increase the discriminant ability for recurrence prediction. Some researchers also provide evidence for the value of combined liquid biopsy and radiology. Integration of deep learning radiomics and CTC improving the indeterminate lung solid nodule diagnosis (28,29), and improves prediction of recurrence outcomes for patients who are treated with radiation therapy for early-stage NSCLC (39). The integration of radiographic tumor volume and circulating tumor DNA may help determine risk of recurrence in patients for early-stage NSCLC (40). Thus, combining CTC and radiologic measurements might further enhance personalized and accurate management of LUAD patients than traditional stage.

In the present study, 3D-SM and 3D-SV were significantly correlated and performed similarly in the prediction of RFS. However, while the cutoff value for CTR was taken as 0.5 based on previous studies (12,13), only 3D-SM% other than 3D-SV% could stratify patients with clinical stage IA. Since the mass measured based on CT is calculated by volume and attenuation, it can reflect the changes of both, showing potential in evaluating the growth rate of part-solid ground-glass nodules (41). A recent study also showed that for patients with stage T1 LUAD, the independent prognostic factor for some solid nodules is tumor size rather than CTR (42). Considering a minor ground glass opacity component given the favorable prognosis (43), 3D-SM% may perform more sensitively than 3D-SV%. Moreover, compared to R-MSSA%, 3D-SM% or 3D-SV% did not perform advantageously in the prediction of recurrence. Thus, further study is necessary to ensure the application of 3D-CTR in early-stage LUAD.

The present study still has some limitations. Since there is limited long-term follow-up data on CTC, our study has a small sample size and lacks external validation, and a prospective study and large sample size are needed to understand how tumor mass or volume and CTC integrate better. Furthermore, the influence of vessels or bronchioles was not avoided in the measurements. For all cases with 3D-SM% lower than 1, the ground-glass nodules and solid nodules were not investigated separately. In addition, as we only analyzed CTC count without further detection of CTC clusters, the prognostic predictive value of CTCs may have been underestimated.

Conclusions

In conclusion, the integration of CTC and 3D-SM, 3D-SV, 3D-SM% or R-MSSA%, could re-stratify the risk of recurrence for lung cancer patients at clinical stage IA. The integration of 3D-SM and CTC can improve the prognosis prediction for patients with clinical stage I of LUAD, compared with 3D-SM or CTC alone.

Acknowledgments

We would like to thank Jian-Cheng Yang and Li Zhang (Dianei Technology) for providing technical support for this study.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-aw-2342/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-aw-2342/dss

Funding: This work was supported by a grant from the National Natural Science Foundation of China (No. 82102109).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-aw-2342/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The retrospective study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of The First Affiliated Hospital of Sun Yat-sen University (SYSUFA) (No. [2021]-531), and individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

  1. Ettinger DS, Wood DE, Aisner DL, Akerley W, Bauman JR, Bharat A, et al. NCCN Guidelines® Insights: Non-Small Cell Lung Cancer, Version 2.2023. J Natl Compr Canc Netw 2023;21:340-50. [Crossref] [PubMed]
  2. Lee JM, McNamee CJ, Toloza E, Negrao MV, Lin J, Shum E, Cummings AL, Kris MG, Sepesi B, Bara I, Kurtsikidze N, Schulze K, Ngiam C, Chaft JE. Neoadjuvant Targeted Therapy in Resectable NSCLC: Current and Future Perspectives. J Thorac Oncol 2023;18:1458-77. [Crossref] [PubMed]
  3. Banna GL, Hassan MA, Signori A, Giunta EF, Maniam A, Anpalakhan S, Acharige S, Ghose A, Addeo A. Neoadjuvant Chemo-Immunotherapy for Early-Stage Non-Small Cell Lung Cancer: A Systematic Review and Meta-Analysis. JAMA Netw Open 2024;7:e246837. [Crossref] [PubMed]
  4. Nomori H, Mori T, Shiraishi A, Fujino K, Sato Y, Ito T, Suzuki M. Long-Term Prognosis After Segmentectomy for cT1 N0 M0 Non-Small Cell Lung Cancer. Ann Thorac Surg 2019;107:1500-6. [Crossref] [PubMed]
  5. Maeda R, Yoshida J, Hishida T, Aokage K, Nishimura M, Nishiwaki Y, Nagai K. Late recurrence of non-small cell lung cancer more than 5 years after complete resection: incidence and clinical implications in patient follow-up. Chest 2010;138:145-50. [Crossref] [PubMed]
  6. Van Schil PE, Asamura H, Nishimura KK, Rami-Porta R, Kim YT, Bertoglio P, Cangir AK, Donington J, Fang W, Giroux DJ, Lievens Y, Liu H, Lyons G, Sakai S, Travis WD, Ugalde P, Jeffrey Yang CF, Yotsukura M, Detterbeck F; . The International Association for the Study of Lung Cancer Lung Cancer Staging Project: Proposals for the Revisions of the T-Descriptors in the Forthcoming Ninth Edition of the TNM Classification for Lung Cancer. J Thorac Oncol 2024;19:749-65.
  7. Maeda R, Yoshida J, Ishii G, Aokage K, Hishida T, Nishimura M, Nishiwaki Y, Nagai K. Long-term outcome and late recurrence in patients with completely resected stage IA non-small cell lung cancer. J Thorac Oncol 2010;5:1246-50. [Crossref] [PubMed]
  8. Jovanoski N, Bowes K, Brown A, Belleli R, Di Maio D, Chadda S, Abogunrin S. Survival and quality-of-life outcomes in early-stage NSCLC patients: a literature review of real-world evidence. Lung Cancer Manag 2023;12:LMT60. [Crossref] [PubMed]
  9. Shin DW, Cho JH, Yoo JE, Cho J, Yoon DW, Lee G, Shin S, Kim HK, Choi YS, Kim J, Zo JI, Shim YM. Conditional Survival of Surgically Treated Patients with Lung Cancer: A Comprehensive Analyses of Overall, Recurrence-free, and Relative Survival. Cancer Res Treat 2021;53:1057-71. [Crossref] [PubMed]
  10. Saji H, Okada M, Tsuboi M, Nakajima R, Suzuki K, Aokage K, et al. Segmentectomy versus lobectomy in small-sized peripheral non-small-cell lung cancer (JCOG0802/WJOG4607L): a multicentre, open-label, phase 3, randomised, controlled, non-inferiority trial. Lancet 2022;399:1607-17. [Crossref] [PubMed]
  11. Zhang Y, Fu F, Wen Z, Deng L, Wang S, Li Y, Chen H. Segment Location and Ground Glass Opacity Ratio Reliably Predict Node-Negative Status in Lung Cancer. Ann Thorac Surg 2020;109:1061-8. [Crossref] [PubMed]
  12. Ito H, Suzuki K, Mizutani T, Aokage K, Wakabayashi M, Fukuda H, Watanabe SI; . Long-term survival outcome after lobectomy in patients with clinical T1 N0 lung cancer. J Thorac Cardiovasc Surg 2020;S0022-5223(20)30054-4.
  13. Wu Y, Song W, Wang D, Chang J, Wang Y, Tian J, Zhou S, Dong Y, Zhou J, Li J, Zhao Z, Che G. Prognostic value of consolidation-to-tumor ratio on computed tomography in NSCLC: a meta-analysis. World J Surg Oncol 2023;21:190. [Crossref] [PubMed]
  14. Fu F, Zhang Y, Wen Z, Zheng D, Gao Z, Han H, Deng L, Wang S, Liu Q, Li Y, Shen L, Shen X, Zhao Y, Zhao Z, Ye T, Xiang J, Zhang Y, Sun Y, Hu H, Chen H. Distinct Prognostic Factors in Patients with Stage I Non-Small Cell Lung Cancer with Radiologic Part-Solid or Solid Lesions. J Thorac Oncol 2019;14:2133-42. [Crossref] [PubMed]
  15. Yoon SH, Kim YJ, Doh K, Kim J, Lee KH, Lee KW, Kim J. Interobserver variability in Lung CT Screening Reporting and Data System categorisation in subsolid nodule-enriched lung cancer screening CTs. Eur Radiol 2021;31:7184-91. [Crossref] [PubMed]
  16. Hamanaka K, Takayama H, Koyama T, Matsuoka S, Takeda T, Agatsuma H, Yamada K, Hyogotani A, Kawakami S, Ito KI. Interobserver size measurement variability in part-solid lung adenocarcinoma using pre-operative computed tomography. J Thorac Dis 2019;11:2924-31. [Crossref] [PubMed]
  17. Kawaguchi Y, Nakao M, Omura K, Iwamoto N, Ozawa H, Kondo Y, Ichinose J, Matsuura Y, Okumura S, Mun M. The utility of three-dimensional computed tomography for prediction of tumor invasiveness in clinical stage IA lung adenocarcinoma. J Thorac Dis 2020;12:7218-26. [Crossref] [PubMed]
  18. Hong JH, Park S, Kim H, Goo JM, Park IK, Kang CH, Kim YT, Yoon SH. Volume and Mass Doubling Time of Lung Adenocarcinoma according to WHO Histologic Classification. Korean J Radiol 2021;22:464-75. [Crossref] [PubMed]
  19. Kawaguchi Y, Shimada Y, Murakami K, Omori T, Kudo Y, Makino Y, Maehara S, Hagiwara M, Kakihana M, Yamada T, Park J, Matsubayashi J, Ohira T, Ikeda N. Prognostic impact of artificial intelligence-based volumetric quantification of the solid part of the tumor in clinical stage 0-I adenocarcinoma. Lung Cancer 2022;170:85-90. [Crossref] [PubMed]
  20. Zhu Y, Chen LL, Luo YW, Zhang L, Ma HY, Yang HS, Liu BC, Li LJ, Zhang WB, Li XM, Xie CM, Yang JC, Wang DL, Li Q. Prognostic impact of deep learning-based quantification in clinical stage 0-I lung adenocarcinoma. Eur Radiol 2023;33:8542-53. [Crossref] [PubMed]
  21. Lou F, Sima CS, Rusch VW, Jones DR, Huang J. Differences in patterns of recurrence in early-stage versus locally advanced non-small cell lung cancer. Ann Thorac Surg 2014;98:1755-60; discussion 1760-1. [Crossref] [PubMed]
  22. Potter AL, Costantino CL, Suliman RA, Haridas CS, Senthil P, Kumar A, Mayne NR, Panda N, Martin LW, Yang CJ. Recurrence After Complete Resection for Non-Small Cell Lung Cancer in the National Lung Screening Trial. Ann Thorac Surg 2023;116:684-92. [Crossref] [PubMed]
  23. Gu X, Wei S, Lv X. Circulating tumor cells: from new biological insights to clinical practice. Signal Transduct Target Ther 2024;9:226. [Crossref] [PubMed]
  24. Lawrence R, Watters M, Davies CR, Pantel K, Lu YJ. Circulating tumour cells for early detection of clinically relevant cancer. Nat Rev Clin Oncol 2023;20:487-500. [Crossref] [PubMed]
  25. Chemi F, Rothwell DG, McGranahan N, Gulati S, Abbosh C, Pearce SP, et al. Pulmonary venous circulating tumor cell dissemination before tumor resection and disease relapse. Nat Med 2019;25:1534-9. [Crossref] [PubMed]
  26. Jin F, Zhu L, Shao J, Yakoub M, Schmitt L, Reißfelder C, Loges S, Benner A, Schölch S. Circulating tumour cells in patients with lung cancer universally indicate poor prognosis. Eur Respir Rev 2022;31:220151. [Crossref] [PubMed]
  27. Garitaonaindia Y, Aguado-Noya R, Garcia-Grande A, Cordoba M, Coronado Albi MJ, Campo Cañaveral JL, Calvo V, Clemente MB, Álvarez R, Peñas M, Chara L, Royuela A, Provencio M. Monitoring with circulating tumor cells in the perioperative setting of patients with surgically treated stages I-IIIA NSCLC. Transl Lung Cancer Res 2023;12:1414-24. [Crossref] [PubMed]
  28. Wan Z, He H, Zhao M, Ma X, Sun S, Wang T, Deng J, Zhong Y, She Y, Ma M, Wang H, Chen Q, Chen C. The development and validation of a circulating tumor cells-based integrated model for improving the indeterminate lung solid nodules diagnosis. Transl Lung Cancer Res 2023;12:566-79. [Crossref] [PubMed]
  29. Zhang Y, Sun B, Yu Y, Lu J, Lou Y, Qian F, Chen T, Zhang L, Yang J, Zhong H, Wu L, Han B. Multimodal fusion of liquid biopsy and CT enhances differential diagnosis of early-stage lung adenocarcinoma. NPJ Precis Oncol 2024;8:50. [Crossref] [PubMed]
  30. Ren L, Zhong X, Liu W, Xu D, Lei Y, Zhou J, Jiang W, He Q, Sun Y, Ke Z. Clinical Significance of a Circulating Tumor Cell-based Classifier in Stage IB Lung Adenocarcinoma: A Multicenter, Cohort Study. Ann Surg 2023;277:e439-48. [Crossref] [PubMed]
  31. He W, Xu D, Wang Z, Wu H, Xiang X, Tang B, Jiang W, Cui Y, Wang H, Jiang N, Sun Y, Chen Y, Li S, Hou M, Zhang Y, Wang L, Ke ZF. Three-dimensional nanostructured substrates enable dynamic detection of ALK-rearrangement in circulating tumor cells from treatment-naive patients with stage III/IV lung adenocarcinoma. J Transl Med 2019;17:32. [Crossref] [PubMed]
  32. Aokage K, Miyoshi T, Ishii G, Kusumoto M, Nomura S, Katsumata S, Sekihara K, Tane K, Tsuboi M. Influence of Ground Glass Opacity and the Corresponding Pathological Findings on Survival in Patients with Clinical Stage I Non-Small Cell Lung Cancer. J Thorac Oncol 2018;13:533-42. [Crossref] [PubMed]
  33. Murakami S, Ito H, Tsubokawa N, Mimae T, Sasada S, Yoshiya T, Miyata Y, Yokose T, Okada M, Nakayama H. Prognostic value of the new IASLC/ATS/ERS classification of clinical stage IA lung adenocarcinoma. Lung Cancer 2015;90:199-204. [Crossref] [PubMed]
  34. Altorki N, Wang X, Kozono D, Watt C, Landrenau R, Wigle D, et al. Lobar or Sublobar Resection for Peripheral Stage IA Non-Small-Cell Lung Cancer. N Engl J Med 2023;388:489-98. [Crossref] [PubMed]
  35. Wang T, She Y, Yang Y, Liu X, Chen S, Zhong Y, Deng J, Zhao M, Sun X, Xie D, Chen C. Radiomics for Survival Risk Stratification of Clinical and Pathologic Stage IA Pure-Solid Non-Small Cell Lung Cancer. Radiology 2022;302:425-34. [Crossref] [PubMed]
  36. Na KJ, Kim YT, Goo JM, Kim H. Clinical Utility of a CT-based AI Prognostic Model for Segmentectomy in Non-Small Cell Lung Cancer. Radiology 2024;311:e231793. [Crossref] [PubMed]
  37. Bera K, Braman N, Gupta A, Velcheti V, Madabhushi A. Predicting cancer outcomes with radiomics and artificial intelligence in radiology. Nat Rev Clin Oncol 2022;19:132-46. [Crossref] [PubMed]
  38. Avanzini S, Kurtz DM, Chabon JJ, Moding EJ, Hori SS, Gambhir SS, Alizadeh AA, Diehn M, Reiter JG. A mathematical model of ctDNA shedding predicts tumor detection size. Sci Adv 2020;6:eabc4308. [Crossref] [PubMed]
  39. Jiao Z, Li H, Xiao Y, Dorsey J, Simone CB, Feigenberg S, Kao G, Fan Y. Integration of Deep Learning Radiomics and Counts of Circulating Tumor Cells Improves Prediction of Outcomes of Early Stage NSCLC Patients Treated With Stereotactic Body Radiation Therapy. Int J Radiat Oncol Biol Phys 2022;112:1045-54. [Crossref] [PubMed]
  40. Tran HT, Heeke S, Sujit S, Vokes N, Zhang J, Aminu M, Lam VK, Vaporciyan A, Swisher SG, Godoy MCB, Cascone T, Sepesi B, Gibbons DL, Wu J, Heymach JV. Circulating tumor DNA and radiological tumor volume identify patients at risk for relapse with resected, early-stage non-small-cell lung cancer. Ann Oncol 2024;35:183-9. [Crossref] [PubMed]
  41. Song YS, Park CM, Park SJ, Lee SM, Jeon YK, Goo JM. Volume and mass doubling times of persistent pulmonary subsolid nodules detected in patients without known malignancy. Radiology 2014;273:276-84. [Crossref] [PubMed]
  42. Li Z, Xu W, Gu T, Cao X, Wu W, Chen L. Tumor size, but not consolidation-to-tumor ratio, is an independent prognostic factor for part-solid clinical T1 non-small cell lung cancer. Thorac Cancer 2023;14:602-11. [Crossref] [PubMed]
  43. Li M, Wang J, Bao X, Wang F, Zhou F, Song N, Wan Z, Li W, Zhang L. Minor (≤ 10%) Ground-Glass Opacity Component in Clinical Stage I Non-Small Cell Lung Cancer: Associations With Pathologic Characteristics and Clinical Outcomes. AJR Am J Roentgenol 2024;223:e2431283. [Crossref] [PubMed]
Cite this article as: Li L, Chen M, Zhu J, Ma L, Lin F, Li XM, Li Q, Zhu Y. Deep learning-based computed tomography quantification integrated with circulating tumor cells for prognostic evaluation in stage I lung adenocarcinoma. Quant Imaging Med Surg 2026;16(7):526. doi: 10.21037/qims-2025-aw-2342

Article Options

Download Citation

Share