RPS 2204 - Updates in thoracic intervention

Author Block: J. Kroschke¹, B. J. Kerber¹, J. Happe¹, C. Steinack², P. Baumgartner², R. Engeli², S. Ulrich², T. Frauenfelder¹, T. Gaisl²; ¹Zürich/CH, ²Zurich/CH
Purpose: Malignancy risk stratification of pulmonary nodules remains a major challenge in lung cancer screening and incidentally detected nodules. Patients referred for robotic-assisted bronchoscopy with integrated cone-beam CT (RAB+CBCT) at a tertiary hospital represent a distinct, high-risk cohort and it is unclear whether existing stratification tools could reduce biopsy needs. This study evaluated the performance of statistical and deep-learning risk models in this setting.
Methods or Background: We retrospectively analyzed 130 patients (54.5% male, mean age 68.1±9.6 years) with 176 nodules who underwent RAB+CBCT biopsy for suspected lung cancer. Biopsied nodules were visually matched to pre-procedural CT scans using bronchoscopy reports and intraprocedural CBCT. Clinical data were extracted from medical records. Automated segmentation was performed, and a deep learning algorithm (malignancy similarity index, mSI; RevealDx) was applied to CT-based thumbnails including perifocal parenchyma. Brock and Mayo risk scores were calculated.
Results or Findings: Histopathology confirmed lung cancer in 68.8% of nodules. Nodule characteristics were 53.4% solid (62.8% malignant), 21.0% part-solid (56.8% malignant), and 25.6% ground-glass (91.1% malignant). Median nodule size was 12.00mm (IQR 9.00-16.12mm) and volume 670.00mm³ (IQR 290.50-1710.75 mm³). Most nodules (62.5%) were located in upper lobes, without consistent differences in malignancy rate across lobes (e.g. right upper 64.9% vs. right lower 71.4%). ROC analysis demonstrated limited discrimination with Brock (AUC=0.62), Mayo (AUC 0.63), and mSI (AUC=0.50). In patients with prior scans, volumetric growth achieved higher accuracy (AUC=0.75).
Conclusion: Our RAB+CBCT cohort shows a distinct risk profile, especially unexpectedly high malignancy rates in ground-glass nodules. Existing risk models, including deep learning, underperform in this population, making biopsy currently unavoidable. Future research should target tailored stratification models to improve decision-making in high-risk cohorts.
Limitations: Single-center, single-vendor. Limited patient numbers. Selection bias. AI-model not trained for this application.
Funding for this study: No external funding was provided for this study.
Has your study been approved by an ethics committee?: Yes
Ethics committee - additional information: Approval for this study was granted by the local ethics committee (BASEC2025-D0037).

Author Block: N. Gasimli, V. Mammadlı, H. Ayyildiz, S. M. Ertürk, R. MAMMADZADA; Istanbul/TR
Purpose: Our aim in this study is to evaluate the predictability of pneumothorax, the most common complication of PTB, before TTNAB with deep learning algorithms, and to plan patient management.
Methods or Background: A total of 403 PTB procedures performed under CT guidance in our institution were included in our retrospective study. In our hospital system, images of patients who underwent TTIAB between 2018 and 2023 were downloaded and the data was processed with deep learning algorithms in anonymized form. Control CT images of the patients taken during the PTB planning phase before biopsy and after PTB were used. In our study, a deep learning model was created using the VGG-19 algorithm to classify DICOM images.
Results or Findings: Pneumothorax did not develop in 271 (67%) of the patients who underwent the procedure, while pneumothorax development was observed in 132 cases (33%). 76 of these 132 cases (58%) did not require hospitalization, but 54 cases (42%) required hospitalization. According to the results obtained as a result of the analysis process in our study, the VGG-19 model has 68.7% accuracy and 0.68 AUC values. However, the sensitivity value calculated as 52% as a result of the analysis shows that approximately half of the positive cases are predicted correctly. As a result of the application of the model, 73% of negative cases were classified correctly. This can be attributed to the fact that the chosen method has more effective success on negative cases.
Conclusion: These values, which express the general accuracy rate of the model, show that the selected method has an acceptable success in predicting the risk of pneumothorax.
Limitations: Our study is single-center and one of the limitations is that our sample is not larger.
Funding for this study: None.
Has your study been approved by an ethics committee?: Yes
Ethics committee - additional information: Ethics committe: 09.05.2023

Author Block: X. ZHANG, S. Thom R.G., G. J. De Jonge, M. Van Tuinen, W. Caroline van de, E. Michiel, W. F. A. Den Dunnen, G. De Bock, M. Dorrius; Groningen/NL
Purpose: In December 2020, our center transitioned from using a wire with a fish-hook shape to a wire with a spiral shape for CT-guided pulmonary nodule localization. This study compared the effectiveness and safety of both techniques.
Methods or Background: We retrospectively reviewed 157 consecutive patients who underwent CT-guided wire localization between November 2017 and December 2024 in University Medical Center Groningen (UMCG), Netherlands. Patient demographics, lesion characteristics, procedural and surgical details, and pathological findings were collected. Procedure-related complications were classified as minor (pneumothorax or hemoptysis without clinical consequences) or major (events requiring intervention, including air embolism). Logistic regression was performed to identify factors associated with localization success and procedure-related events.
Results or Findings: Fish-hook shaped wires were used in 79 patients and spiral shaped wires in 78 patients. Technical success was similar between groups (94.9% vs. 96.2%). Minor complications occurred more frequently with fish-hook shape wires (34.2% vs. 21.8%), while major complications were rare and comparable. Pneumothorax rates (minor and major combined) were 31.6% for fish-hook shape wire and 20.5% for spiral shape wires. Pathological representativeness did not differ significantly. In multivariable regression, longer procedure duration and fish-hook shape wire use were independently associated with an increased risk of minor complications.
Conclusion: Both wire types demonstrated high and comparable effectiveness and safety for CT-guided pulmonary nodule localization. However, spiral-shaped wires are associated with fewer minor complications, supporting their preferential use in clinical practice.
Limitations: This is a single-center study with a relatively small sample size, which may limit the generalizability of our findings. In addition, pain scores are not collected, precluding evaluation of this potentially relevant outcome.
Funding for this study: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Has your study been approved by an ethics committee?: Not applicable
Ethics committee - additional information:

Author Block: E. G. KAHRAMAN, Y. VAROL, A. Selver, O. Ozdemir, E. Hasbay, Y. EROL; Izmir/TR
Purpose: To develop and evaluate a novel biopsy-indication scoring system based on TorchXRayVision (TxRV), an open-source deep learning model trained on large chest radiograph datasets, aiming to support clinical decision-making in differentiating malignant from benign lung lesions.
Methods or Background: Chest radiographs of 300 patients were screened; 285 (206 malignant, 79 benign) were eligible after excluding anterior–posterior views and indeterminate pathology. TxRV outputs for 18 radiological findings were extracted. For biopsy indication, six core features (effusion, pneumonia, nodule, mass, lung lesion, opacity) were used to generate three handcrafted scores (simple sum, weighted, maximum). For malignancy prediction, extended 18-feature models were tested: simple sum, weighted sum, logistic regression, and random forest. Statistical analyses included Mann–Whitney U tests, ROC/AUC, confusion matrices, and feature importance mapping.
Results or Findings: Nodule, mass, and lung lesion scores were significantly higher in malignant cases (p < 0.01), while opacity showed borderline association (p = 0.06). Weighted biopsy scoring yielded the highest discriminatory capacity, with a malignancy prevalence of ~80% above the 75th percentile cut-off. Logistic regression improved interpretability, achieving AUC 0.71 with balanced sensitivity/specificity. Random forest demonstrated superior performance (AUC 0.94, accuracy 90%), but feature importance confirmed that classic oncologic signs (mass, nodule, lesion) remained the strongest predictors. Precision–recall analysis supported these findings, with Random Forest showing F1 = 0.92, PPV 0.94, and NPV 0.85, underscoring robust diagnostic value.
Conclusion: TxRV-derived scoring provides an interpretable, reproducible framework to guide biopsy indication and malignancy risk stratification. Weighted scoring improved diagnostic balance, and logistic regression offered stable performance suitable for clinical translation. Such modeling may help reduce unnecessary lung biopsies.
Limitations: This was a single-center retrospective study with modest sample size. Random forest results suggest possible overfitting, highlighting the need for external validation in larger, multi-center cohorts.
Funding for this study: None
Has your study been approved by an ethics committee?: Yes
Ethics committee - additional information: This study was approved by the Ethics Committee of Izmir City Hospital (Approval Number: 2025/364 )
All procedures were conducted in accordance with the Declaration of Helsinki.

Author Block: R. Ahmed, P. Beddy, J. Kavanagh, D. Murphy, V. Young, S. Nicholson, J. F. Meaney; Dublin/IE
Purpose: Anterior mediastinal masses are uncommon, with an estimated prevalence of 0.4-0.9%. Computed tomography (CT) guided core needle biopsy (CNB) facilitates a histological diagnosis and appropriate management. The mediastinum contains several vital structures which must be considered when deciding a trajectory for biopsy. The aim of this study is to review technique, yield, and complications to describe a safe and effective method for biopsy.
Methods or Background: All CT-guided CNB for anterior mediastinal masses (2015 -2025) at a single centre were included. All images were reviewed by two experienced radiologists. Univariable comparisons between the diagnostic yield and non-diagnostic yield groups were performed.
Results or Findings: 52 patients were evaluated. 37 (71.15%) were diagnostic and 15 were (28.85%) non-diagnostic. Of the diagnostic samples, 16 (43.24%) were thymic, 14 (37.85%) lymphomas, and 7 (18.92%) other diagnoses. Minor complications included 6 (11.54%) pneumothoraxes, 6 (11.54%) haematomas, 5 (9.62%) pneumomediastinum, 3 (5.77%) pleural effusions, 2 (3.85%) haemothoraxes, and 1 pulmonary haemorrhage (1.92%). Major complications include 1 (1.92%) haemopericardium requiring pericardiocentesis and 1 (1.92%) fatal pulmonary artery puncture. On univariable analysis, a larger anteroposterior dimension (p < 0.017), larger mediolateral dimension (p < 0.041), and a greater number of biopsy samples taken (p < 0.002) were associated with a lower risk of non-diagnostic yield.
Conclusion: CNB is effective for confirming the nature of anterior mediastinal masses, but complications can arise, particularly from inadvertent puncture of the many vascular structures within this area. To our knowledge, this is the first reported case of a biopsy related fatality. We highlight a method to avoid vascular puncture to ensure safe outcome.
Limitations: Retrospective cohort study.
Small sample size.
Funding for this study: No funding support was available for this study.
Has your study been approved by an ethics committee?: Yes
Ethics committee - additional information: Ethical approval was gained from the local ethics committee at St James's Hospital

Author Block: H. D. J. Rueda-Chaparro¹, A. Al Mutairi², F. Cadour³, S. Kandel³, P. Rogalla³; ¹Bogotá/CO, ²Riad/SA, ³Toronto, ON/CA
Purpose: To assess the effect of operator experience on diagnostic yield and complication rates in computed tomography (CT)-guided percutaneous chest biopsies.
Methods or Background: This retrospective single-center cohort study included 1,453 consecutive adult patients who underwent CT-guided chest biopsy between July 2022 and June 2024. Exclusion criteria were incomplete clinical/procedural data, ultrasound-guided biopsies, and operators with less than 20 procedures performed during the study period. Eight operators were stratified into three groups based on procedural volume: Group A (200/year, n=2). Outcomes included diagnostic yield, complication rates, and predictors of both, assessed using univariable and multivariable logistic regression.
Results or Findings: Adequate tissue for definitive histopathology was obtained in 1,392 cases (96%). Diagnostic yield differed across operator groups: 93% in Group A, 94% in Group B, and 98% in Group C (P=.002). Group C remained independently associated with higher diagnostic yield (OR 2.69; 95%CI 1.26–5.73; P=.011). Complications occurred in 27% of patients, including pneumothorax (20%), chest tube insertion (2%) and other events (5%). Complication rates were highest in Group A (16%) and lowest in Group B (9%), yet operator experience was not independently associated with complications. Independent predictors of higher complication risk included smaller lesion size (OR 0.98 per mm, 95%CI 0.97–0.99; P=.005), use of sedation (OR 1.55; 95%CI 1.09–2.21; P=.016), and lesion location in the right upper lobe (OR 1.73; 95%CI 1.03–2.91; P=.038).
Conclusion: Operator experience was independently associated with higher diagnostic yield in CT-guided chest biopsies, with the most experienced operators achieving superior diagnostic performance. Complication risk related chiefly to lesion size, sedation, and anatomical location.
Limitations: Limitations include the retrospective single-center design, reliance on a two-year procedural volume, and possible dilution of operator-specific effects due to trainee supervision.
Funding for this study: None
Has your study been approved by an ethics committee?: Yes
Ethics committee - additional information: The Research Ethics Board at University Health Network, Toronto, Canada approved this research.

Author Block: D. Saeed¹, N. D. Adroja¹, E. Obasi¹, C. Ridge²; ¹London/UK, ²London/IE
Purpose: Pleural drain insertion is frequently required in patients on dual antiplatelet therapy (DAPT) at our tertiary cardiothoracic centre. Departmental policy advises withholding P2Y12 inhibitors for seven days prior to drain placement, however this is often impractical in urgent settings. Evidence on bleeding risk in this group is limited. This audit evaluated complication rates in DAPT patients undergoing ultrasound (US)-guided pleural drain placement.
Methods or Background: A retrospective review was conducted of all US thorax and interventional radiology guided US drainage procedures from April 2024 to April 2025. Inclusion criteria were documented DAPT use at the time of procedure. Data collected included demographics, antiplatelet regimen, drug-hold duration, procedural details, operator grade, laboratory results, and complications.
Results or Findings: Of 368 drain requests, 33 (9%) were in patients on DAPT (median age 73 years; range 42-81; 64% male). Most received aspirin + clopidogrel (94%), with 70% continuing DAPT without interruption. Drains were predominantly 8 Fr (88%), left-sided (73%), and inserted in the radiology department (73%); 27% were performed in ICU. Operators included consultants (24%), fellows (21%), and registrars (55%). Mean interval from request to insertion was 0.6 days. No major or minor bleeding complications occurred. Two patients received transfusions for unrelated reasons. Referral documentation omitted complete antiplatelet / anticoagulant details in 68% of cases.
Conclusion: In this audit, US-guided small-bore pleural drain placement in patients on uninterrupted DAPT was not associated with bleeding complications, supporting British Society of Interventional Radiology guidance that classifies the procedure as low risk. The findings question the necessity of a seven-day P2Y12 inhibitor hold and highlight the need for improved documentation on referral requests.
Limitations: Small sample size, retrospective design, incomplete documentation, single-centre setting, and lack of long-term follow-up limit generalisability.
Funding for this study: No funding
Has your study been approved by an ethics committee?: Not applicable
Ethics committee - additional information:

Author Block: V. Gnanananthan, M. Morgan, S. Lam, H. Kolajian, A. El-Zeki, A. Metwalli, C. Peebles, A. Alzetani, L. Veres; Southampton/UK
Purpose: Navigational bronchoscopy (NB) is a recognized and increasingly available technique for tissue diagnosis of central pulmonary lesions unamenable to percutaneous CT guided approach. The use of intra-procedural 3D-CT has helped increase diagnostic yield and recent studies have demonstrated good safety profile with fewer complications compared to CT guided biopsy. However, NB remains resource intensive, requiring general anaesthetic and theatre access. Using CT lesion characteristics to predict features favouring diagnostic sampling will allow patient stratification in MDT, avoiding risks of anaesthetic and diagnostic delay.
Methods or Background: This retrospective study included 101 consecutive, eligible patients who underwent NB biopsy at University Hospital Southampton between June 2023 and February 2025. Patients with available histology reports and diagnostic CT imaging preceding NB were included. Diagnostic success was defined as an unequivocal malignant or benign histological diagnosis. Insufficient and equivocal results were classed as non-diagnostic. Pre-NB CT imaging was analysed for lesion characteristics by a team of radiologists.
Results or Findings: NB demonstrated good safety profile with 1% (n=1) post-procedure pneumothorax and no post-procedure haemorrhage . 60% (n=60) of procedures yielded diagnostic sample.
Greater lesion size showed weak but statistically significant correlation with diagnostic sampling rate (r=0.20, p=0.04). Lesions ≥2cm were significantly more likely to yield diagnostic sample (p=0.035) as were lesions adjacent to more central airways (Weibel airway generation ≤4 ) (p=0.008).
Solid vs subsolid lesions, lobar location and presence of a bronchus sign showed no significant correlation to diagnostic sampling.
Conclusion: CT lesion size cut-off (2cm) and assessment of proximity to central bronchial tree (airway generation ≤4) are significant predictors of success and will help stratify patients referred for NB and streamline cancer diagnostic pathways.
Limitations: Patients with unavailable diagnostic CT imaging were excluded
Funding for this study: No funding was obtained for this study.
Has your study been approved by an ethics committee?: Not applicable
Ethics committee - additional information:

Author Block: A. K. Geetha Virupakshappa; Bedford/UK
Purpose: Aim is to evaluate the histological adequacy and safety of the CT guided percutaneous lung core biopsy done in our radiology department
Methods or Background: Adequacy and safety parameters: (First cycle)
Diagnostic adequacy : 95 % of samples were adequate. Only one sample was inadequate, and this was lesion >2cm, done with an 20G needle and 5 passes obtained.
Sensitivity for malignancy 95% for lesions >2 cm
False positive rate =0%.
Complication rates:
Pneumothorax =20%
Pneumothorax needing drainage= 15% (3cases)
Haemoptysis =0%
Death 2 cm
False positive rate =0%.
Complication rates:
Pneumothorax =40% (10cases)
Pneumothorax needing drainage= 0%
Haemoptysis =16% (4cases)
Death < 0%
Results or Findings: We have a diagnostic accuracy rate of 95% in first cycle and 100% in second cycle, higher than the standard (90%).
We have a pneumothorax rate equal to the standard, 20% in first cycle and 40% in second cycle, but none of them required drainage (0%, standard is 3%).
16% of patients experienced haemoptysis after the procedure (Standard <5%), which were managed conservatively (none of them required hospitalization).
No death resulting from the procedure was recorded.
Conclusion: Documentation of the procedure (mode of imaging, needle size, core samples, number of passes) 100%. Diagnostic adequacy should be >90%. Sensitivity for malignancy > 85%. False positive rate <1%. complication rates: Pneumothorax <20%, pneumothorax requiring drainages <3%, haemoptysis<5%
Limitations: Individual radiologist variation
Funding for this study: No funding for the study
Has your study been approved by an ethics committee?: Not applicable
Ethics committee - additional information: