Preoperative localization of pulmonary nodules using ultra-low-dose CT based on artificial intelligence iterative reconstruction

Introduction

Lung cancer ranks as the most common malignancy globally and first in mortality worldwide (1,2). Early diagnosis and timely surgical resection are fundamental for prolonging patient survival. Video-assisted thoracoscopic surgery (VATS) wedge resection demonstrates comparable 5-year disease-free survival to more invasive segmentectomy or lobectomy for small peripheral nodules, while better preserving pulmonary function and enabling faster postoperative recovery—establishing it as a standard approach for select early-stage lung cancers (3-6).

Successful wedge resection hinges on precise nodule localization. However, small solid nodules or ground-glass opacities (GGOs) are often undetectable by visual inspection or manual palpation during VATS (7). Computed tomography (CT)-guided hook-wire placement has emerged as an effective localization method (8), but requires ≥3 scans (pre-/post-needle insertion, post-hook-wire deployment) (9). Repeated scanning increases radiation exposure, conflicting with the as low as reasonably achievable (ALARA) principle for radiation protection (10).

Although current guidelines lack specific dose recommendations for CT-guided pulmonary nodule localization, reducing radiation exposure remains a critical clinical priority. Multiple dose-reduction strategies have been implemented: iterative reconstruction-based low-dose protocols are most widely adopted, whereas tin filtration techniques achieve up to 73.2% dose reduction compared to standard protocols—albeit limited to select platforms such as Siemens scanners (11-14). Moreover, quiet breathing during localization reduces lesion-to-lung contrast compared to deep-inspiration scans, and metallic localization grids/needles introduce streak artifacts. These technical challenges compound the difficulty of maintaining nodule conspicuity on low-dose CT (LDCT) images. Deep learning-based approaches, such as artificial intelligence iterative reconstruction (AIIR), represents a breakthrough by integrating convolutional neural networks with iterative algorithms to preserve image textures while suppressing noise (15). This technology enables diagnostic-quality chest CT imaging at ultra-low doses comparable to radiography, demonstrating non-inferiority to standard-dose CT in nodule detection and malignancy characterization (15,16).

We hypothesized that ultra-low-dose CT (ULDCT) with AIIR may overcome noise limitations in localization imaging. This prospective study compared ULDCT-AIIR against conventional LDCT for nodule localization accuracy, potentially establishing radiography-dose CT as a viable option for preoperative localization. We present this article in accordance with the STARD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1544/rc).

Methods

Patient cohort

From August to December 2024, we enrolled candidates for preoperative hook-wire localization of pulmonary nodules using CT guidance at The Fifth Affiliated Hospital of Sun Yat-sen University. The inclusion criteria were as follows: (I) nodules with high malignancy suspicion on diagnostic CT; (II) size ≤20 mm located in the peripheral third of the lung suitable for wedge resection; (III) requirement for localization if: diameter <1 and >1.5 cm from pleural surface, or pure ground-glass nodules (GGNs), or anticipated intraoperative localization difficulty; (IV) age ≥18 years. The exclusion criteria were as follows: coagulopathy; proximity to cardiac structures; osseous obstruction of needle path; and body mass index (BMI) >30 kg/m². The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of The Fifth Affiliated Hospital of Sun Yat-sen University (approval No. K181-1). Written informed consent was provided by all participants. The trial was registered with the Chinese Clinical Trial Registry (Registration No. ChiCTR2400088626).

CT scanning protocol

All patients underwent pulmonary nodule localization using a novel 320-detector row CT system (uCT 960+, United Imaging Healthcare, Shanghai, China). The localization procedure employed a three-phase scanning sequence: pre-needle insertion, post-needle placement, and post-hook-wire deployment. At each phase, an LDCT scan was immediately followed by a ULDCT scan in direct succession, ensuring minimal procedural delay while generating dose-matched LDCT-ULDCT image pairs for comparative analysis. The LDCT protocol, serving as the reference for localization, was performed with 120 kVp and 30 mAs (17,18), whereas the ULDCT protocol employed 100 kVp and 5 mAs (15,19). Identical gantry rotation time (0.5 s) and pitch (0.956) were applied in both protocols. Notably, the cumulative radiation dose for both LDCT and ULDCT complied with established protocols (20-22), remaining below the routine-dose level (120 kVp/110 mAs) specified in our prior clinical practice.

To evaluate procedural safety, we quantified the total procedure duration and radiation exposure metrics, including the volume CT dose index (CTDI_vol, mGy), dose-length product (DLP, mGy·cm), and effective dose (ED). ED was calculated as ED = DLP × k, where the conversion coefficient k=0.014 mSv/(mGy·cm). Radiation exposure metrics were compared between the LDCT and ULDCT protocols.

AIIR and CT image reconstruction

Technically, AIIR is an advanced deep learning-based reconstruction algorithm featuring a newly designed backbone for CT image reconstruction. It integrates the advantages of deep learning and model-based iterative reconstruction (MBIR) by replacing the regularization term in MBIR with a convolutional neural network. Therefore, this approach effectively suppresses image noise without introducing the “plastic” image appearance that is easily caused by the regularization term, while retaining the ability of characterizing image detail provided by MBIR (23-26).

Both LDCT and ULDCT scans were reconstructed using filtered back projection (FBP) and hybrid iterative reconstruction (HIR, Karl 3D, level 5; United Imaging Healthcare). ULDCT scans were additionally processed with AIIR (United Imaging Healthcare). For each patient, five image sequences were ultimately obtained: LDCT-FBP, LDCT-HIR, ULDCT-FBP, ULDCT-HIR, and ULDCT-AIIR. Since AIIR has not yet been commercially implemented in uCT960+, LDCT-HIR images were utilized for clinical localization procedures, whereas other image sequences served exclusively for comparative assessment of AIIR-reconstructed ultra-low-dose images’ potential feasibility in future practice.

Positioning procedure

The needle puncture localization was performed by one attending thoracic surgeon (14 years of experience). Initially, the puncture path and estimated skin entry point were planned based on prior standard-dose whole-lung CT images. With the patient positioned supine, prone, or laterally as preoperatively planned in the scanning table, a custom wire-made localization grid was placed at the estimated skin entry point. Subsequently, a limited LDCT scan centered on the target nodule was performed during quiet respiration.

The LDCT-HIR images were then used to precisely determine the skin entry point and plan the needle trajectory and depth. After routine local disinfection, layered infiltration anesthesia was administered using 2% lidocaine hydrochloride. Following percutaneous needle insertion (Needle brand: Kangqing SA-10, Changzhou, China), a confirmatory scan was performed to assess the distance between the needle tip and the lesion. A distance <2 cm was deemed indicative of satisfactory needle placement, at which point the hook-wire was deployed. Final scanning was performed to evaluate the hook-wire position and assess the complications such as pneumothorax or hemothorax. Hook-wire placement within 2 cm of the nodule was considered an appropriate localization marker.

Objective image analysis

A radiologist with 2 years of chest imaging experience measured the attenuation values [Hounsfield units (HU)] of pulmonary nodules, lung tissue, and the aorta. On axial images showing the largest nodule diameter, circular regions of interest (ROIs) were manually placed at identical levels: within the aortic lumen, in the nodule, and in avascular/bronchus-free parenchyma surrounding the nodule. The ROI covered at least 80% of the nodule area, with all measurements utilizing the copy-paste function to ensure perfect consistency in ROI size, shape, and position across different images. The standard deviation (SD) of attenuation values in perinodular parenchyma and that of the aorta were defined as lung parenchymal noise and aortic noise, respectively. We calculated the nodule-specific contrast-to-noise ratio (CNR) using the formula: CNR_nodule-lung = (ROI_nodule − ROI_lung)/Noise_lung, where Noise_lung was defined as the SD of attenuation values in the lung parenchyma. Subsequently, the global image CNR was computed using lung parenchymal and aortic attenuation values: CNR_aorta-lung = (ROI_aorta − ROI_lung)/Noise_aorta, where Noise_aorta was defined as the SD of attenuation values in the aorta (20,21).

Subjective image analysis

After anonymization and randomization of all image datasets, one attending thoracic radiologist (17 years of diagnostic experience) and the same thoracic surgeon who had performed the positioning procedure independently evaluated pulmonary nodule image quality using the following 5-point scale: 1= non-diagnostic (severe artifacts; indistinct lesion and surrounding structures, unfeasible for path planning); 2= poor (significant artifacts; blurred lesion/adjacent tissues; difficult path selection); 3= adequate (sufficient for procedure; mild blurring but feasible path selection); 4= good (mild artifacts; well-visualized lesion/tissues; appropriate path selection); 5= excellent (no artifacts; sharp anatomical delineation, optimal path visualization). Using an established 5-point scale modeled after similar criteria, the following elements were assessed: localization markers (grid wires, needles, and hook-wires) and complications (hemorrhage and pneumothorax).

Distance measurements: pulmonary nodules to localization references

To evaluate the potential impacts of scan dosage and reconstruction algorithms on localization accuracy, distances were measured across different image sequences between: pulmonary nodules and the pleura, nodules and needle tips, and nodules and hook-wires.

Statistical analysis

All statistical analyses were performed using GraphPad Prism (Version 9.3.1; GraphPad Software, San Diego, CA, USA). Inter-rater agreement between the two radiologists’ subjective scores was assessed with Cohen’s kappa coefficient, interpreted as follows: κ≥0.75: excellent agreement; 0.4≤κ<0.75: moderate agreement; κ<0.4: poor agreement. Using Kruskal-Wallis test (non-parametric omnibus test) and Dunn’s post hoc analysis (multiple comparisons), statistical comparisons were performed across different image reconstructions for the following metrics measured: objective noise levels, CNR values (both nodule-specific and global), nodule-to-reference distances (pleura, needles, hook-wires), and subjective quality scores. Statistical significance was defined as p<0.05.

Results

Patient characteristics

From August to December 2024, 40 patients with suspected early-stage lung cancer requiring percutaneous localization were prospectively enrolled (Table S1). The average age of the patients was 56.70±9.85 years, comprising 14 males (35%) and 26 females (65%). Regarding BMI categories, 5 patients (12.5%) had BMI <18.5 kg/m², 27 patients (67.5%) had BMI 18.5–24.9 kg/m², and 8 patients (20%) had BMI ≥25 kg/m². Nodule types comprised 3 solid (7.5%), 17 part-solid (42.5%), and 20 GGN (50%), with a mean diameter of 11.06±3.80 mm (5.57–19.60 mm). Post-procedural complications included minor pulmonary hemorrhage in 14 patients (35%), minimal pneumothorax in 10 (25%), and concurrent hemorrhage with pneumothorax in 5 (12.5%), yielding an overall complication rate of 72.5%. Total procedural time averaged 485.0±222.3 seconds. Following localization, wedge resection was performed in 23 patients (57.5%), whereas others underwent segmentectomy or lobectomy based on intraoperative pathological findings (e.g., invasive adenocarcinoma).

Compared with LDCT, the ULDCT protocol achieved an approximate 90% reduction in CTDI_vol, DLP, and ED (Table 1).

Objective image quality analysis

For pulmonary tissue noise and nodule-to-lung contrast, ULDCT-AIIR demonstrated 31.61% to 49.91% reduction in pulmonary noise compared to ULDCT-FBP, ULDCT-HIR, and LDCT-FBP groups (Figures 1,2, Figure S1), with corresponding significantly improved CNR_nodule-lung (P<0.01). However, no significant difference was observed between ULDCT-AIIR and LDCT-HIR (P>0.05) (Figure 2). Regarding aorta noise and aorta-to-lung contrast, ULDCT-AIIR achieved 24.92–81.81% noise reduction versus ULDCT-FBP, ULDCT-HIR, LDCT-FBP, and LDCT-HIR groups (Figures S1,S2) (P<0.01), accompanied by significantly enhanced CNR_aorta-lung (P<0.05) (Figure S3).

Figure 1 Measured lung parenchyma image noise from different scans during needle insertion. Box plots (A-C) display the lung tissue noise measured in images across experimental groups during the puncture procedure. The central line indicates the median, the box boundaries correspond to the first and third quartiles, and the whiskers represent the maximum and minimum values of the dataset respectively. ****, P<0.0001; ns, non-significant. AIIR, artificial intelligence iterative reconstruction; FBP, filtered-back projection; HIR, hybrid iterative reconstruction; HU, Hounsfield units; LDCT, low-dose computed tomography; ULDCT, ultra-low-dose computed tomography.

Figure 2 Contrast-to-noise ratio of pulmonary nodule-to-lung tissue measured in each group’s images during the puncture procedure. Box plots (A-C) display the nodule-to-parenchyma CNR measured in images across experimental groups. The central line indicates the median, the box boundaries correspond to the first and third quartiles, and the whiskers represent the maximum and minimum values of the dataset respectively. ****, P<0.0001; ***, P<0.001; **, P<0.01; ns, non-significant. AIIR, artificial intelligence iterative reconstruction; CNR, contrast-to-noise ratio; FBP, filtered-back projection; HIR, hybrid iterative reconstruction; LDCT, low-dose computed tomography; ULDCT, ultra-low-dose computed tomography.

Subjective image quality analysis

Cohen’s kappa coefficient for evaluating pulmonary nodules, localization markers, and complications clarity were 0.85–0.97, 0.71–1, and 0.86–0.97, respectively, indicating excellent agreement in subjective image quality scoring. The 1–5 point scoring data are summarized in the bar charts presented in Figures S4-S7.

As image noise decreased, subjective scores for pulmonary nodule clarity progressively increased. Except for being comparable to LDCT-HIR (P>0.05), ULDCT-AIIR demonstrated significantly higher subjective scores than ULDCT-FBP, ULDCT-HIR, and LDCT-FBP (P<0.0001) (Figure S4). The AIIR algorithm enhanced the clarity of nodular margins and surrounding vasculobronchial structures across all nodule types, with the most pronounced improvement in part-solid and GGNs (Figure 3, Figure S5).

Figure 3 Comparison of visualization performance across different reconstruction algorithms for solid nodules, part-solid nodules, and ground-glass nodules. AIIR, artificial intelligence iterative reconstruction; FBP, filtered-back projection; HIR, hybrid iterative reconstruction; LDCT, low-dose computed tomography; ULDCT, ultra-low-dose computed tomography.

Regarding localization grids, needles, and hook-wires, FBP and HIR exhibited varying degrees of metal artifacts and structural blurring, which were particularly severe in ULDCT-FBP and ULDCT-HIR. In contrast, ULDCT-AIIR significantly mitigated these artifacts, showing optimal enhancement for localization grids while maintaining comparable needle/hook-wire clarity to LDCT-HIR and LDCT-FBP (Figure 4, Figure S6). In detecting post-procedural pulmonary hemorrhage and pneumothorax, ULDCT-AIIR demonstrated comparable performance to LDCT-HIR and outperformed most other imaging sequences (Figure 5, Figure S7).

Figure 4 Clarity of localization grid, needle, and hook-wire under different radiation doses and reconstruction methods. AIIR, artificial intelligence iterative reconstruction; FBP, filtered-back projection; HIR, hybrid iterative reconstruction; LDCT, low-dose computed tomography; ULDCT, ultra-low-dose computed tomography.

Figure 5 Visualization of pulmonary hemorrhage and pneumothorax on images from different reconstruction methods. AIIR, artificial intelligence iterative reconstruction; FBP, filtered-back projection; HIR, hybrid iterative reconstruction; LDCT, low-dose computed tomography; ULDCT, ultra-low-dose computed tomography.

Distance measurements between nodules and localization references

As shown in Table 2, there were no statistically significant differences in measurements of nodule-to-pleura distance, nodule-to-needle tip distance, and nodule-to-hook-wire distance across all imaging protocols (P>0.05), indicating that ULDCT-AIIR provides accurate spatial measurements for preoperative localization planning.

Discussion

Preoperative localization of pulmonary nodules typically requires at least three localized CT scans, necessitating effective radiation dose reduction while ensuring safety. Given that new ULDCT techniques has been clinically validated in lung cancer visualization, coronary computed tomography angiography (CTA), and pulmonary infection studies—demonstrating diagnostic efficacy comparable to standard-dose CT while achieving radiation levels approaching chest radiography (15,16,19-21), we conducted a paired comparison with LDCT to evaluate the feasibility of ULDCT for nodule localization. To the best of our knowledge, the application of artificial intelligence (AI)-denoised ULDCT for preoperative nodule localization has not been extensively explored. Using our ULDCT protocol (100 kVp, 5 mAs), we achieved 90% reductions in ED compared to LDCT. Regarding radiation metrics, the CTDI_vol for our AIIR-based ULDCT protocol was 0.22 mGy, representing 68.8% of tin filtration-based techniques (0.32 mGy), 44.9% of conventional iterative reconstruction (0.49 mGy), and was comparable to GE’s TrueFidelity™ (GE HealthCare, Chicago, IL, USA) AI denoising (0.13–0.27 mGy) (15). This constitutes the lowest reported CTDI_vol for CT-guided localization to date (22,27,28). AI-based denoising enables transformative radiation reduction for CT-guided thoracic nodule localization, significantly alleviating radiation concerns—particularly crucial for multiple-nodule localization, extended procedures (e.g., biopsy + ablation) and radiation-sensitive cohorts (29). Furthermore, AI-based denoising—validated across major CT platforms (Siemens, Philips, GE, United Imaging Healthcare)—demonstrates superior clinical adaptability compared to hardware-dependent solutions such as tin filtration, which are available only on select platforms (e.g., third-generation dual-source Siemens scanners).

The primary concern regarding the application of ULDCT in puncture localization is the clarity of pulmonary nodules. Since patients maintain quiet breathing during the puncture localization procedure, the contrast between lesions and lung tissue is not as pronounced as in conventional scans with deep inspiration, making the need for image noise reduction particularly urgent. As anticipated, the image quality of ULDCT-FBP was severely compromised. Even after iterative algorithm optimization, its image noise remained significantly higher than it did LDCT-FBP. In contrast, AIIR was trained on a diverse dataset encompassing variations in technical parameters, anatomy, and pathology. Comparative studies across a wide range of clinical scenarios have validated its robust performance, demonstrating its superiority over conventional methods (24-26). In our study, these advantages resulted in significantly improved contrast for pulmonary nodules and higher subjective image quality scores, achieving levels comparable to those of LDCT-HIR. Our study further revealed that ULDCT-AIIR demonstrated notably more pronounced improvements in image quality for GGNs and part-solid nodules compared to solid nodules. This phenomenon primarily stems from the minimal density difference between the ground-glass components of nodules and surrounding lung tissue. Under the extreme noise conditions of ULDCT-FBP/HIR reconstructions, the signal contrast fails to meet diagnostic requirements. After AIIR-enhanced denoising, the clarity of both GGNs and part-solid nodules reached near-parity with solid nodules. Considering that part-solid nodules and GGNs constitute the predominant subtypes encountered in wedge resection procedures (30)—where thoracoscopic localization proves particularly challenging—the clinical significance of AIIR in ULDCT-guided localization for these nodule types is especially pronounced.

Another critical determinant for successful nodule localization is the clarity of metallic reference devices. Prior studies have rarely thoroughly evaluated the image quality of reference markers, likely due to their substantially higher contrast with surrounding tissues compared to the contrast between lung tissue and nodules. Nevertheless, the metal artifacts from these markers, which can compromise localization accuracy, cannot be overlooked (14). As this study was designed to specifically evaluate AIIR’s clinical value, the scanner’s built-in metal artifact correction (MAC) algorithm was not activated. However, AIIR’s underlying MBIR technique inherently reduces streak artifacts by preserving fine anatomical details, thus improving marker visibility. Notably, ULDCT-AIIR demonstrated particularly pronounced noise reduction for localization grids, with significantly superior clarity compared to LDCT-HIR. Although the visualization of needles and hook-wires surpassed that of ULDCT-FBP and ULDCT-HIR, no statistically significant difference emerged versus LDCT-HIR or LDCT-FBP. The enhanced visualization of metallic localization grids under AIIR may be attributed to their parallel alignment with the patient’s longitudinal axis—optimizing photon flux per unit length—while also potentially reflecting differential artifact profiles arising from distinct metallic compositions of the two localization markers. Given the extensive well-established research on AI algorithms for metal artifact reduction, the synergistic integration of AIIR-based noise suppression with AI-driven artifact correction will enable more precise characterization of spatial relationships between pulmonary nodules and fiducial markers (31).

Our measurements of distances from pulmonary nodules to the reference markers demonstrated no significant differences across imaging groups—theoretically meeting fundamental requirements for nodule localization. However, these measurements were derived from experimental analyses rather than actual clinical settings. In practical nodule localization procedures, suboptimal visualization compromises operator confidence, prolongs procedural time, and elevates misinterpretation risks. Based on the significantly enhanced clarity of both nodules and localization markers, AIIR may effectively enhance the efficiency and safety of ULDCT-based localization.

Another safety-related concern involves complications from puncture localization, yet prior studies have rarely evaluated ULDCT’s capability in detecting pneumothorax and pulmonary hemorrhage. Although most localization complications are relatively minor, substantial pneumothorax or pulmonary hemorrhage—requiring prompt intervention—can significantly disrupt surgical scheduling. Our study demonstrates that ULDCT-AIIR clearly delineates these complications, thereby providing robust assessment for post-localization and pre-surgical safety evaluation.

Previous studies on ULDCT had typically employed either randomized patient allocation to LDCT or ULDCT scans, or generated virtual LDCT/ULDCT reconstructions from standard-dose CT data (32-34). In contrast, this study subjected the same cohort to both LDCT and ULDCT scanning protocols—which may raise ethical concerns regarding potential radiation exposure risks. Our dual-phase scanning protocol achieved a 58% reduction in cumulative radiation dose compared to conventional standard-dose CT (data not shown). Crucially, ULDCT acquisition was completed within the image reconstruction downtime of the LDCT scans, effectively eliminating any net increase in procedural time. Furthermore, compared to both our institution’s prior standard-dose scans and protocols reported by other groups, our combined LDCT + ULDCT approach demonstrated non-inferiority in total procedural duration and complication rates, thereby confirming its clinical safety and operational feasibility (35-37).

This study has several limitations. First, the generalizability of our findings is constrained by the modest sample size and the involvement of only a single operator for nodule localization. Future studies involving larger patient cohorts—including obese individuals (BMI ≥30 kg/m²)—and multiple operators are warranted to strengthen the robustness of the conclusions. The second limitation is that the AIIR algorithm was developed and validated exclusively on United Imaging uCT 960 scanners. Since it was trained on proprietary imaging data, its generalizability to other platforms may be limited. Therefore, further training, validation, and optimization using diverse datasets from multiple CT vendors will be essential to enable broader adaptation and widespread clinical application of this algorithm. Finally, since the AIIR algorithm is not yet commercially available on this CT scanner, we could not use LDCT-AIIR or ULDCT-AIIR as the reference standard for direct preoperative nodule localization. With the ongoing development and expected clinical implementation of this algorithm, we plan to conduct randomized controlled trials comparing ULDCT-AIIR with LDCT-HIR and LDCT-AIIR, to further validate the reliability of AIIR-based ULDCT in guiding preoperative localization.

Conclusions

Our research indicates that ULDCT-AIIR can achieve image quality comparable to that of LDCT-HIR with significantly reduced radiation dose, suggesting its potential as a substitute for preoperative pulmonary nodule localization in LDCT. Further prospective studies in real-world clinical practice are warranted, particularly with the imminent commercialization of AIIR technology.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STARD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1544/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1544/dss

Funding: This work was supported by the National Natural Science Foundation of China (grant No. 82372077) and the Natural Science Foundation of Guangdong Province (grant No. 2024A1515013100).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-1544/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 study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of The Fifth Affiliated Hospital of Sun Yat-sen University (approval No. K181-1). Written informed consent was obtained from all participants.

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. Bray F, Laversanne M, Sung H, Ferlay J, Siegel RL, Soerjomataram I, Jemal A. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. Luo G, Zhang Y, Rumgay H, Morgan E, Langselius O, Vignat J, Colombet M, Bray F. Estimated worldwide variation and trends in incidence of lung cancer by histological subtype in 2022 and over time: a population-based study. Lancet Respir Med 2025;13:348-63. [Crossref] [PubMed]
3. Bendixen M, Jørgensen OD, Kronborg C, Andersen C, Licht PB. Postoperative pain and quality of life after lobectomy via video-assisted thoracoscopic surgery or anterolateral thoracotomy for early stage lung cancer: a randomised controlled trial. Lancet Oncol 2016;17:836-44. [Crossref] [PubMed]
4. Suzuki K, Watanabe SI, Wakabayashi M, Saji H, Aokage K, Moriya Y, Yoshino I, Tsuboi M, Nakamura S, Nakamura K, Mitsudomi T, Asamura HWest Japan Oncology Group and Japan Clinical Oncology Group. A single-arm study of sublobar resection for ground-glass opacity dominant peripheral lung cancer. J Thorac Cardiovasc Surg 2022;163:289-301.e2. [Crossref] [PubMed]
5. Potter AL, Senthil P, Keshwani A, McCleery S, Haridas C, Kumar A, Mathey-Andrews C, Martin LW, Yang CJ. Long-term Survival After Lung Cancer Resection in the National Lung Screening Trial. Ann Thorac Surg 2024;117:734-42. [Crossref] [PubMed]
6. Seder CW, Chang SC, Towe CW, Puri V, Blasberg JD, Bonnell L, Fernandez FG, Habib RH, Kozower BD. Anatomic Lung Resection Is Associated With Improved Survival Compared With Wedge Resection for Stage IA (≤2 cm) NSCLC. J Thorac Oncol 2025;20:1075-85. [Crossref] [PubMed]
7. Ichinose J, Kohno T, Fujimori S, Harano T, Suzuki S. Efficacy and complications of computed tomography-guided hook wire localization. Ann Thorac Surg 2013;96:1203-8. [Crossref] [PubMed]
8. Yao F, Wang J, Yao J, Xu L, Wang J, Gao L. Reevaluation of the efficacy of preoperative computed tomography-guided hook wire localization: A retrospective analysis. Int J Surg 2018;51:24-30. [Crossref] [PubMed]
9. Zhong Y, Xu XQ, Pan XL, Zhang W, Xu H, Yuan M, Kong LY, Pu XH, Chen L, Yu TF. Retrospective Evaluation of Safety, Efficacy and Risk Factors for Pneumothorax in Simultaneous Localizations of Multiple Pulmonary Nodules Using Hook Wire System. Cardiovasc Intervent Radiol 2017;40:1408-14. [Crossref] [PubMed]
10. Smith-Bindman R, Chu PW, Azman Firdaus H, Stewart C, Malekhedayat M, Alber S, Bolch WE, Mahendra M, Berrington de González A, Miglioretti DL. Projected Lifetime Cancer Risks From Current Computed Tomography Imaging. JAMA Intern Med 2025;185:710-9. [Crossref] [PubMed]
11. Mohammadinejad P, Mileto A, Yu L, Leng S, Guimaraes LS, Missert AD, Jensen CT, Gong H, McCollough CH, Fletcher JG CT. Noise-Reduction Methods for Lower-Dose Scanning: Strengths and Weaknesses of Iterative Reconstruction Algorithms and New Techniques. Radiographics 2021;41:1493-508. [Crossref] [PubMed]
12. Li Z, Liu Y, Zhang P, Lu J, Gui Z. Decomposition iteration strategy for low-dose CT denoising. J Xray Sci Technol 2024;32:493-512. [Crossref] [PubMed]
13. Greffier J, Fitton I, Ngoc Ty CV, Frandon J, Beregi JP, Dabli D. Impact of tin filter on the image quality of ultra-low dose chest CT: A phantom study on three CT systems. Diagn Interv Imaging 2023;104:506-12. [Crossref] [PubMed]
14. Zhang J, Liu M, Liu D, Li X, Lin M, Tan Y, Luo Y, Zeng X, Yu H, Shen H, Wang X, Liu L, Tan Y, Zhang J. Low-dose CT with tin filter combined with iterative metal artefact reduction for guiding lung biopsy. Quant Imaging Med Surg 2022;12:1359-71. [Crossref] [PubMed]
15. Jiang B, Li N, Shi X, Zhang S, Li J, de Bock GH, Vliegenthart R, Xie X. Deep Learning Reconstruction Shows Better Lung Nodule Detection for Ultra-Low-Dose Chest CT. Radiology 2022;303:202-12. [Crossref] [PubMed]
16. Yang L, Liu H, Han J, Xu S, Zhang G, Wang Q, Du Y, Yang F, Zhao X, Shi G. Ultra-low-dose CT lung screening with artificial intelligence iterative reconstruction: evaluation via automatic nodule-detection software. Clin Radiol 2023;78:525-31. [Crossref] [PubMed]
17. Aberle DR, Adams AM, Berg CD, Black WC, Clapp JD, Fagerstrom RM, Gareen IF, Gatsonis C, Marcus PM, Sicks JD. Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med 2011;365:395-409. [Crossref] [PubMed]
18. Rodrigues JCL, Pierre AF, Hanneman K, Cabanero M, Kavanagh J, Waddell TK, Chung TB, Pakkal M, Keshavjee S, Cypel M, Yasufuku K, Nguyen ET. CT-guided Microcoil Pulmonary Nodule Localization prior to Video-assisted Thoracoscopic Surgery: Diagnostic Utility and Recurrence-Free Survival. Radiology 2019;291:214-22. [Crossref] [PubMed]
19. Sui X, Meinel FG, Song W, Xu X, Wang Z, Wang Y, Jin Z, Chen J, Vliegenthart R, Schoepf UJ. Detection and size measurements of pulmonary nodules in ultra-low-dose CT with iterative reconstruction compared to low dose CT. Eur J Radiol 2016;85:564-70. [Crossref] [PubMed]
20. Zou LM, Xu C, Xu M, Xu KT, Zhao ZC, Wang M, Wang Y, Wang YN. Ultra-low-dose coronary CT angiography via super-resolution deep learning reconstruction: impact on image quality, coronary plaque, and stenosis analysis. Eur Radiol 2025;35:4674-84. [Crossref] [PubMed]
21. Yan C, Liang C, Xu J, Wu Y, Xiong W, Zheng H, Xu Y. Ultralow-dose CT with knowledge-based iterative model reconstruction (IMR) in evaluation of pulmonary tuberculosis: comparison of radiation dose and image quality. Eur Radiol 2019;29:5358-66. [Crossref] [PubMed]
22. Ma Y, Cheng S, Li J, Yuan W, Song Z, Zhang H. Preoperative CT-guided localization of pulmonary nodules with low-dose radiation. Quant Imaging Med Surg 2023;13:4295-304. [Crossref] [PubMed]
23. Thibault JB, Sauer KD, Bouman CA, Hsieh J. A three-dimensional statistical approach to improved image quality for multislice helical CT. Med Phys 2007;34:4526-44. [Crossref] [PubMed]
24. Gong H, Peng L, Du X, An J, Peng R, Guo R, Ma X, Xiong S, Ma Q, Zhang G, Ma J. Artificial Intelligence Iterative Reconstruction in Computed Tomography Angiography: An Evaluation on Pulmonary Arteries and Aorta With Routine Dose Settings. J Comput Assist Tomogr 2024;48:244-50. [Crossref] [PubMed]
25. Li W, You Y, Zhong S, Shuai T, Liao K, Yu J, Zhao J, Li Z, Lu C. Image quality assessment of artificial intelligence iterative reconstruction for low dose aortic CTA: A feasibility study of 70 kVp and reduced contrast medium volume. Eur J Radiol 2022;149:110221. [Crossref] [PubMed]
26. Li J, Meng T, Zhang G, Yu X, Lu Z, Zhang W. Artificial intelligence iterative reconstruction in abdominal CT of patients with irregular arm positioning: a case-by-case evaluation. Acta Radiol 2024;65:907-12. [Crossref] [PubMed]
27. Wei W, Wang SG, Zhang JY, Togn XY, Li BB, Fang X, Pu RW, Zhou YJ, Liu YJ. Implementation of Individualized Low-Dose Computed Tomography-Guided Hook Wire Localization of Pulmonary Nodules: Feasibility and Safety in the Clinical Setting. Diagnostics (Basel) 2023;13:3235. [Crossref] [PubMed]
28. Yuan H, Li D, Zhang Y, Xie X, Shen L. Value of low-dose and optimized-length computed tomography (CT) scan in CT-guided percutaneous transthoracic needle biopsy of pulmonary nodules. J Interv Med 2021;4:143-8. [Crossref] [PubMed]
29. Liu S, Liang B, Li Y, Xu J, Qian W, Lin M, Xu M, Niu L. CT-Guided Percutaneous Cryoablation in Patients with Lung Nodules Mainly Composed of Ground-Glass Opacities. J Vasc Interv Radiol 2022;33:942-8. [Crossref] [PubMed]
30. Chen H, Kim AW, Hsin M, Shrager JB, Prosper AE, Wahidi MM, Wigle DA, Wu CC, Huang J, Yasufuku K, Henschke CI, Suzuki K, Tailor TD, Jones DR, Yanagawa J. The 2023 American Association for Thoracic Surgery (AATS) Expert Consensus Document: Management of subsolid lung nodules. J Thorac Cardiovasc Surg 2024;168:631-647.e11. [Crossref] [PubMed]
31. Kleber CEJ, Karius R, Naessens LE, Van Toledo CO. A C van Osch J, Boomsma MF, Heemskerk JWT, van der Molen AJ. Advancements in supervised deep learning for metal artifact reduction in computed tomography: A systematic review. Eur J Radiol 2024;181:111732. [Crossref] [PubMed]
32. Ye K, Chen M, Zhu Q, Lu Y, Yuan H. Effect of adaptive statistical iterative reconstruction-V (ASiR-V) levels on ultra-low-dose CT radiomics quantification in pulmonary nodules. Quant Imaging Med Surg 2021;11:2344-53. [Crossref] [PubMed]
33. Peters AA, Munz J, Klaus JB, Macek A, Huber AT, Obmann VC, Alsaihati N, Samei E, Valenzuela W, Christe A, Heverhagen JT, Solomon JB, Ebner L. Impact of Simulated Reduced-Dose Chest CT on Diagnosing Pulmonary T1 Tumors and Patient Management. Diagnostics (Basel) 2024;14:1586. [Crossref] [PubMed]
34. You Y, Zhong S, Zhang G, Wen Y, Guo D, Li W, Li Z. Exploring the Low-Dose Limit for Focal Hepatic Lesion Detection with a Deep Learning-Based CT Reconstruction Algorithm: A Simulation Study on Patient Images. J Imaging Inform Med 2024;37:2089-98. [Crossref] [PubMed]
35. Klinkenberg TJ, Dinjens L, Wolf RFE, van der Wekken AJ, van de Wauwer C, de Bock GH, Timens W, Mariani MA, Groen HJM. CT-guided percutaneous hookwire localization increases the efficacy and safety of VATS for pulmonary nodules. J Surg Oncol 2017;115:898-904. [Crossref] [PubMed]
36. Zhang H, Li Y, Yimin N, He Z, Chen X. CT-guided hook-wire localization of malignant pulmonary nodules for video assisted thoracoscopic surgery. J Cardiothorac Surg 2020;15:307. [Crossref] [PubMed]
37. Chu S, Wei N, Lu D, Chai J, Liu S, Lv W. Comparative study of the effect of preoperative hookwire and methylene blue localization techniques on post-operative hospital stay and complications in thoracoscopic pulmonary nodule surgery. BMC Pulm Med 2022;22:336. [Crossref] [PubMed]