RPS 1816 - Modern strategies in thoracic oncologic imaging

Author Block: A. Cicchetti¹, P. Vallerio¹, A. Catalano¹, C. Sangalli¹, L. Marrazzo², R. Tummineri¹, A. Botti³, F. Dionisi⁴, E. Gioscio¹; ¹Milan/IT, ²Firenze/IT, ³Reggio Emilia/IT, ⁴Rome/IT
Purpose: To assess whether baseline CT-derived cardiovascular and metabolic biomarkers can stratify intermediate-risk patients with locally advanced non-small cell lung cancer (LA-NSCLC) treated with chemoradiotherapy+immunotherapy, thereby supporting personalized follow-up strategies and identifying candidates for advanced cardiological imaging.
Methods or Background: Baseline CT-scans from 305 LA-NSCLC patients enrolled in four Italian centers were retrospectively analyzed. Two-year overall survival (OS2Y) was recorded. Image-derived features included the Agatston score, body composition metrics, aorta diameters, and percent Emphysema. Gross tumor volume (GTV) was contoured by experts and added to the analysis together with chemo/immune information. Automatic segmentation and feature extraction was performed using open-source tools and locally developed scripts.
A Random Forest classifier was trained following feature selection; SHAP values were used for feature importance interpretation, and UMAP+HDBSCAN clustering was applied to identify patient subgroups.
Results or Findings: OS2Y was 62%. Patients receiving chemoradiotherapy (66%) combined with immunotherapy were 34%.
Four distinct survival clusters were identified based on selected features(GTV, Visceral Fat, Hepatic Fat, Agatston score, and Emphysema):

Cluster1 (n=69) – OS2Y 83.6%: characterized by small tumor volumes and absence of imaging risk factors above the median.

Cluster2 (n=24) – OS2Y 62.5%: tumor volumes comparable to Cluster 1, but high torso fat, very-high Agatston score and hepatic fat.

Cluster3 (n=201) – OS2Y 59.7%: features around median values, representing the average population group.

Cluster4 (n=11) – OS2Y 11.1%: characterized by large GTV volumes and moderate-to-high emphysema.
Conclusion: This unsupervised clustering approach demonstrated the ability to identify patient subgroups with shared clinical characteristics but distinct risk profiles. Importantly, it highlighted a significant survival impact (cl1-cl2=21.1%) for patients with baseline cardiac calcifications, visceral and hepatic fat, underlining their potential role as prognostic imaging biomarkers to guide follow-up strategies.
Limitations: Lack of systematic cardiological test, retrospective analysis
Funding for this study: The study LOCATION MATTERS was funded by AIRC MFAG 27480
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Num INT 36/23

Author Block: F. Sparascio¹, E. Versienti¹, L. Cereser¹, T. Nadarević², C. Ciancimino¹, A. P. Pace¹, G. Como¹, R. Girometti¹, C. Zuiani¹; ¹Udine/IT, ²Rijeka/HR
Purpose: To assess intra- and inter-reader agreement for Node-RADS v1.0 in mediastinal lymph node evaluation on chest CT in stage I–III non-small cell lung cancer (NSCLC) and determine its diagnostic performance.
Methods or Background: This retrospective, single-center study included 46 patients (38 adenocarcinomas, 8 squamous cell carcinomas) with 158 pathologically confirmed mediastinal lymph nodes (22 malignant, 136 benign). A contrast- enhanced chest CT scan was available for all patients. Four radiologists (two experts, two juniors) independently assigned Node-RADS scores and descriptors (“size” and
“configuration”) in two sessions, three weeks apart. Intra- and inter-reader agreement were assessed using Gwet’s AC2. Diagnostic performance was assessed by ROC analysis; sensitivity, specificity, and predictive values were calculated at a Node-RADS score ≥3 threshold.
Results or Findings: Inter-reader agreement for Node-RADS scores was almost perfect for experts (Gwet’s AC2 = 0.97; 95% CI: 0.96–0.99) and juniors (Gwet’s AC2 = 0.95; 95% CI: 0.93–0.97). Intra-reader agreement Gwet’s AC2 values ranged from 0.95–0.99. Descriptor agreement was similarly high (Gwet’s AC2 ≥ 0.85). ROC AUCs ranged from 0.71–0.76 for experts and 0.68–0.84 for juniors. At the ≥3 threshold, specificity and negative predictive value were consistently ≥90%, while sensitivity remained limited (<64%) for all readers.
Conclusion: Node-RADS v1.0 shows excellent reproducibility across radiologists with different expertise for mediastinal lymph node assessment on CT in stage I–III NSCLC. Its high specificity and negative predictive value suggest a supportive role in excluding malignancy, although limited sensitivity warrants cautious interpretation and complementary diagnostic assessment.
Limitations: Single-center and retrospective study.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This retrospective study has been approved by the Institutional Review Board (IRB) of the Department of Medicine, University of Udine. Study protocol code: 019/2025. Approval date: 15/01/2025.

Author Block: W. Yang, Q. Feng, X. Chen, X. Xin; Nanjing/CN
Purpose: This study evaluated the potential of spectral image-based histogram features for early assessment of immunotherapy response in advanced lung cancer.
Methods or Background: Thirty-five patients who underwent baseline and follow-up spectral contrast-enhanced CT scans before and during immunotherapy were retrospectively analyzed. Treatment response at the 4th follow-up was determined using RECIST 1.1 and categorized as response (CR, PR) or non-response (SD, PD). Spectral image series-including conventional images, 40/70 keV virtual monoenergetic images (VMI), iodine density, effective atomic number, electron density, and water/iodine-based maps-were reconstructed in arterial and venous phases. VMI-40 keV images were used for 3D semi-automatic lesion segmentation, and first-order histogram features were extracted. Features were standardized with Z-scores, and significant predictors were identified by Mann-Whitney U-test. Logistic regression models were built, and discriminatory ability was evaluated with ROC AUC; AUC differences were compared with the DeLong test.
Results or Findings: Based on RECIST 1.1, 2 patients achieved CR, 23 PR, 9 SD, and 1 PD. From baseline spectral data, three histogram features distinguished response from non-response with an AUC of 0.796. When combining baseline and first follow-up data, three features achieved an improved AUC of 0.852. No predictive features were identified from conventional images.
Conclusion: Histogram features derived from spectral CT, particularly when incorporating both baseline and early follow-up data, show promise for early prediction of immunotherapy response in advanced lung cancer patients.
Limitations: Sample size is small.
Funding for this study: Not applicable
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by the Institutional Ethics Committee. The requirement for informed consent was waived due to the retrospective nature of the analysis.

Author Block: Y. Zhou, Z. Wang, Y. Guo, L. Lei, X. Guo; Zhengzhou/CN
Purpose: To compare the benefits of ultra-high-resolution photon-counting detector CT (PCD-CT) using low radiation and low contrast agent doses versus conventional energy-integrating detector CT (EID-CT) protocols in contrast-enhanced chest CT imaging for patients with lung cancer.
Methods or Background: This prospective study enrolled a total of 200 patients with lung cancer (male: 119) who underwent ultra-high-resolution PCD-CT or EID-CT. Participants were matched for age, sex,BMI, and TNM stage. Two protocols were applied: a low-dose, ultra-high-resolution PCD-CT protocol (2.0 mL/s injection rate, 1.0 mL/kg ) and a standard-dose EID-CT protocol (3.0 mL/s, 1.2 mL/kg). Radiation exposure, adverse events, and baseline lung cancer characteristics were recorded. Quantitative assessments of lesions, lymph nodes, and vessels included standard deviation (SD), signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR), and tumor-to-lung ratio (TLR). Two radiologists independently evaluated subjective image quality, with a particular focus on enhancement-related imaging features. Meanwhile, BMI- and size-dependent subgroup were analyzed.
Results or Findings: The ultra-high-resolution PCD-CT using a low-dose protocol reduced radiation exposure by 66.08% compared to EID-CT (P < 0.001), while also decreasing contrast agent volume by 19.4%, effectively lowering the incidence of CI-AKI and adverse reactions. The PCD-CT 1 mm demonstrated superior SD, SNR, and CNR of tumors, lymph nodes, vessels(P < 0.001). The PCD-CT 0.4 mm provided comparable objective parameters to EID-CT 1 mm, while offering improved overall subjective image quality and detection of contrast-enhanced related imaging features (P < 0.001). The benefits were particularly pronounced in BMI subgroups and in lesions <3 cm.
Conclusion: Ultra-high-resolution PCD-CT with low-dose protocol reduces radiation exposure and incidence of CI-AKI, and enhances overall image quality and improves the detection of imaging features in lung cancer, making it suitable for patients of various BMI and those with small lesions.
Limitations: Not applicable
Funding for this study: the Key Scientific Research Project of Colleges and Universities in Henan Province (20B320047)
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study has been reviewed by the ethics committee

Author Block: H. W. Phillips¹, P. Marra¹, A. Smedile², M. Bertuletti², C. Gargiulo¹, K. D. Martins De Mattos¹, P. A. Bonaffini², G. Muscogiuri², S. Sironi¹; ¹Milan/IT, ²Bergamo/IT
Purpose: Lung cancer screening of at-risk individuals remains under investigation, with heterogeneous results being expected due to local environmental and epidemiological factors. We present the preliminary experience of the experimental low-dose chest CT screening campaign at a single Italian centre.
Methods or Background: Eligibility criteria include age (55-75) and positive smoking-history (≥30 pack years, active or cessation within the last 15 years) with follow-up imaging at variable intervals based on Lung-RADS 1.1 (2019) risk stratification. Suspect nodules underwent further evaluation following specialist review, often 18FDG-PET, biopsy and/or surgical excision. A review of all lung and collateral oncological findings in the study population was performed.
Results or Findings: From 11/02/23, 796 individuals have been imaged, with 593 having a second scan by 30/06/25. 16 diagnoses of NSCLC (2% detection rate) have been made, all categorised as LR4 (sensitivity 100%, specificity 89.5%, PPV 16.3%, NPV 100%), 15/16 at initial CT. All underwent 18FDG-PET (seven positive, nine nonspecific), 12 were biopsied (nine CT-guided, three by EBUS-TBNA) whilst four proceeded directly to surgery. Initial CT also identified 56 others with LR4 nodules, 19 of whom were evaluated by 18FDG-PET. Four had EBUS-TBNA biopsies whilst three were directly excised. Follow-up scans showed evolution to LR4 in 26 others, resulting in six 18FDG-PET scans, one EBUS-TBNA biopsy and one excision. During the project, 17 participants were diagnosed with extra-pulmonary malignancies, seven as a result of their screening CT: two thyroid, two breast, one tonsillar, one lymphoma and one thymoma.
Conclusion: Lung cancer screening at a national level and its potential to provide earlier detection seems valuable, though false positives and resultant investigations deserve consideration.
Limitations: Limitations include provisional data use, cohort factors with strict eligibility criteria, compliance and imaging interpretation, particularly of slow growing malignancies.
Funding for this study: None.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Data obtained from a national multicentre prospective study.

Author Block: Y. Li¹, X. Yang¹, P. Wang², W. Meng¹; ¹Harbin/CN, ²Beijing/CN
Purpose: This study aimed to quantify intratumoural heterogeneity (ITH) to preoperatively predict the spread through air spaces (STAS) status of lung adenocarcinoma (LUAD) and further explore the potential biological basis underlying the prediction model.
Methods or Background: STAS is an aggressive pattern of primary LUAD that affects both prognosis and treatment strategies for patients. In our study, conventional radiomics features and habitat features were extracted from intratumoural and peritumoural regions on preoperative CT images. A new index, the ITH score, was developed to quantify ITH. Clinical-radiologic characteristics associated with STAS were identified by multivariable logistic regression analyses. Additionally, intratumoural-peritumoural habitat features, ITH score, and clinical-radiologic characteristics were integrated into a combined model by various machine learning algorithms. Finally, 24 patients with RNA sequencing data were utilised for gene expression analysis.
Results or Findings: A total of 1268 patients (median age, 60 years; IQR, 53.8–66.0 years; 850 female) were divided into the training set (n=943), validation set (n=236), and external test set (n=89). Using the Light Gradient Boosting Machine classifier, the combined model demonstrated the highest predictive performance for STAS, achieving an AUC value of 0.97 in the training, 0.98 in the validation, and 0.91 in the external test set. Differentially expressed genes in high probability group were associated with monocarboxylic acid transport and metabolism.
Conclusion: The combined model demonstrated superior performance in predicting STAS in primary LUAD.
Limitations: First, its retrospective design may introduce bias, and further prospective studies are needed to validate the model's accuracy. Second, manual tumor delineation by different radiologists could affect the consistency of radiomic features, so future research should prioritize automated segmentation methods. Third, the limited RNA-seq sample size may weaken biological validation, and larger, more diverse cohorts are needed in future studies.
Funding for this study: This study was supported by the Scientific and Technological Innovation 2030-New Generation Artificial Intelligence Project of the National Key Research and Development Program of China, the National Natural Science Foundation of China, and the Climbing Fund of the National Cancer Center.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ky2024-104

Author Block: P. Eckwolf, D. Kifjak, H. Prosch, L. Beer; Wien/AT
Purpose: To assess whether quantitative 18F-fluoro-2-deoxy-D-glucose (FDG) positron emission tomography/computed tomography (PET/CT) parameters improve preoperative thoracic lymph node (LN) staging accuracy in non-small-cell lung cancer (NSCLC).
Methods or Background: Thirty-eight treatment-naïve patients (mean age 67 ± 9 years) with confirmed NSCLC (29 adenocarcinoma [76.3%], 8 squamous cell carcinoma [21.1%], 1 large cell carcinoma [2.6%]) underwent pre-interventional long field-of-view 18F-FDG PET/CT. A total of 149 thoracic LN were sampled intraoperatively, by transbronchial needle aspiration (TBNA), or both. PET metrics included maximum, peak, and mean standardized uptake value (SUVmax, SUVpeak, SUVmean), metabolic tumor volume (MTV), and total lesion glycolysis (TLG).
Results or Findings: Twelve patients (32%) had metastatic thoracic LN (19/149, 12.8%). Metastatic LN were significantly larger (P=0.013) and more metabolically active (P<0.001) than non-metastatic nodes. ROC analysis showed the best diagnostic performance for SUVmax and SUVpeak (AUC 0.86 and 0.862). Using a sensitivity-optimized SUVmax cutoff of 2.82 yielded a negative predictive value (NPV) of 98%, moderate positive predictive value (PPV, 38%), and overall accuracy of 80%. False positives included reactive LN with SUVmax up to 12, limiting PPV despite high NPV.
Conclusion: Quantitative 18F-FDG PET/CT, particularly SUVmax and SUVpeak, enables highly reliable exclusion of thoracic LN metastasis in NSCLC, with excellent NPV for nodes showing low FDG uptake. However, positive findings remain nonspecific and require histopathological confirmation.
Limitations: Retrospective single-center design and modest sample size.
Funding for this study: The financial support by the Austrian Federal Ministry for Digital and Economic Affairs, the National Foundation for Research, Technology and Development and the Christian Doppler Research Association is gratefully acknowledged.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethic committee number: 1187/2025

Author Block: P. Huang¹, Z. Lin², Y. Liao³, W. Deng³, B. Fan¹; ¹Nanchang/CN, ²Hubei/CN, ³Shanghai/CN
Purpose: The 2020 International Association for the Study of Lung Cancer (IASLC) grading system for invasive lung adenocarcinoma (LUAD) provides significant prognostic value and helps identify patients likely to benefit from adjuvant chemotherapy. This study investigated whether quantitative parameters from dual-layer detector spectral CT (DLCT) could preoperatively predict these IASLC grades in LUAD.
Methods or Background: In this retrospective, two-center study, we analyzed DLCT parameters from 263 patients with pathologically confirmed LUAD. Parameters included conventional CT values (arterial/venous phases), virtual monoenergetic image CT values (40/70/100 keV), iodine concentration (IC), effective atomic number (Zeff), electron density (ED), and the arterial enhancement fraction (AEF). Based on IASLC criteria, patients were classified as low-grade (Grades I-II; n=193) or high-grade (Grade III; n=70). Patients from center 1 were randomly divided into a training set and an internal validation set (7:3 ratio), while patients from center 2 formed an external validation set. Statistically significant parameters from univariable analysis in the training set underwent dimensionality reduction via LASSO regression. Subsequent backward stepwise multivariable logistic regression was used to build a predictive model, with performance assessed by the area under the receiver operating characteristic curve (AUC).
Results or Findings: The training, internal validation, and external validation sets comprised 131 (97 low-grade, 34 high-grade), 55 (41 low-grade, 14 high-grade), and 77 (55 low-grade, 22 high-grade) patients, respectively. Multivariable analysis identified venous phase ED and Zeff as independent predictors for IASLC grading. The prediction model based on these factors achieved AUCs of 0.826, 0.754, and 0.808 in the training, internal, and external validation sets, respectively.
Conclusion: Venous phase ED and Zeff derived from DLCT are effective, non-invasive predictors for preoperatively determining the IASLC grade of lung adenocarcinoma.
Limitations: A retrospective study with a limited sample size
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by our hospital's Ethics Committee with a waiver of informed consent.

Author Block: E. Suderland, G. J. Schmid; Neuss/DE
Purpose: To evaluate feasibility, safety, and local tumour control after CT-guided percutaneous cryoablation of primary and metastatic lung cancer.
Methods or Background: This retrospective single-center study included 29 patients (11 men, 18 women; mean age 67.8 years) with 33 pulmonary lesions ≤3 cm treated between 2012 and 2025. Mean lesion size was 17.3 mm (range 7-30 mm). Eight lesions represented primary lung cancer (1 SCLC, 1 SCC, 6 Adenocarcinomas), and 25 metastases (14 colorectal, 4 breast, 3 uterine, 4 others). Cryoablation was performed under general anaesthesia with CT-guidance. The cryoablation protocol usually consisted of 3 freezing cycles of 3/7/10 minutes, followed by passive and active thawing. Depending on tumor size and location 1-3 cryoprobes were used. Chest tubes were placed during or after treatment in cases of pneumothorax. Treatment decisions were approved by a multidisciplinary tumour board and oncologic consultation.
Results or Findings: Initial technical success was 100%. No major adverse events occured. Minor complications included 18 peri-interventional pneumothoraces and 6 post-interventional pneumothoraces; in total, 18 chest tubes were placed. Additional minor events were perifocal bleeding, post-procedural haemoptysis and pneumonia. Mean hospital stay was 4.2 days (SD 2.1). After a mean follow-up of 14.7 months (SD 13.6), 21 lesions showed local tumor control, 6 showed progression and for 6 lesions follow-up was not available.
Conclusion: CT-guided cryoablation is safe and effective for selected primary and secondary lung tumours, achieving high technical success with low complication rates and good intermediate local tumor control.
Limitations: The study is limited by its retrospective, single-centre design and the relatively small patient cohort. In 6 cases follow-up was missing to fully assess local tumour control.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: J. Tu, L. Lai, W. Li; Lishui/CN
Purpose: To investigate the safety and effectiveness of drug-eluting transarterial chemoembolization plus anlotinib of non-small cell lung cancer in the elderly.
Methods or Background: Patients with at least 75 years old, refusal or unresectable, driver gene-negative, refusal or inability to tolerate radiotherapy, chemotherapy and immunotherapy, non-small cell lung cancer, were enrolled in this retrospective, multicenter study (January 2018–December 2023). DE-TACE consisted cisplatin (75 mg/m2) and gemcitabine (600 mg/m2) via feeding arteries and then embolization using drug-eluting beads carrying gemcitabine (400 mg). Start anlotinib 12mg daily 1week post-operative, 2 weeks on/1 week off (21-day cycle). The primary objectives were overall survival (OS) and safety. Overall survival (OS) was analyzed using Kaplan-Meier analysis.
Results or Findings: The final analysis included 90 patients (median age, 80 years, 74 men). The technical success rate of intratumoral drug delivery was 100%. Within the 26-month median follow-up, the median OS was 25.6 months (95% CI 21.1-29.7) and the median PFS was 8.6 months (95% CI 7.8-9.3). Grade 3 or higher treatment-emergent adverse events was 18.9% (17/90).
Conclusion: DE-TACE plus anlotinib was a successful and effective treatment for non-small cell lung cancer in the elderly, with no severe adverse events.
Limitations: The sample size (particularly the PSM cohort) may not be sufficient to detect differences of AEs. And the treatment was gemcitabine-based. Whether the results apply to other chemotherapeutic regimens requires further investigation.
Funding for this study: There is no funding for this study
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Approval Number: Research Ethics Review 2025 (I) No. 043

Author Block: C. Fang, Y. Cui, X. Yang; Taiyuan,Shanxi/CN
Purpose: To evaluate the predictive value of dual-layer spectral CT (DLCT) combined with machine learning models for predicting spread through air spaces (STAS) in lung adenocarcinoma.
Methods or Background: This dual-center study retrospectively analyzed 439 patients with pathologically confirmed lung adenocarcinoma. Clinical-imaging features and DLCT parameters were evaluated. Independent predictors were identified via logistic regression. We built two predictive models: a DLCT-parameter model (Model-DLCT) and a combined model (Model-COM). Five ML algorithms (RF, XGBoost, SVM, LR, LightGBM) were trained and compared.
Results or Findings: Multivariate analysis identified CTR as a significant clinical predictor, and ED-VP and NIC-VP as key DLCT predictors. The XGBoost algorithm outperformed others. Model-DLCT achieved AUCs of 0.833 (training) and 0.829 (test). Model-COM demonstrated superior performance, with AUCs of 0.862 and 0.832, respectively. The model showed excellent calibration and clinical utility on decision curve analysis.
Conclusion: The XGBoost-based model integrating DLCT parameters and clinical features enables accurate, non-invasive prediction of STAS preoperatively. This tool holds significant promise for improving individualized surgical planning and risk stratification in lung adenocarcinoma.
Limitations: Not applicable.
Funding for this study: This work was supported by the National Natural Science Foundation of China (NSFC, Grant No. 82572216).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This retrospective study was approved by the Institutional Review Board (IRB) of Shanxi Cancer Hospital (Ethics Approval Number: GZ2025014), and the requirement for written informed consent was waived due to the retrospective nature of the analysis, in accordance with the Declaration of Helsinki.