Dataset Information

Prediction of malignant lymph nodes in NSCLC by machine-learning classifiers using EBUS-TBNA and PET/CT.

ABSTRACT: Accurate determination of lymph-node (LN) metastases is a prerequisite for high precision radiotherapy. The primary aim is to characterise the performance of PET/CT-based machine-learning classifiers to predict LN-involvement by endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) in stage-III NSCLC. Prediction models for LN-positivity based on [¹⁸F]FDG-PET/CT features were built using logistic regression and machine-learning models random forest (RF) and multilayer perceptron neural network (MLP) for stage-III NSCLC before radiochemotherapy. A total of 675 LN-stations were sampled in 180 patients. The logistic and RF models identified SUV_max, the short-axis LN-diameter and the echelon of the considered LN among the most important parameters for EBUS-positivity. Adjusting the sensitivity of machine-learning classifiers to that of the expert-rater of 94.5%, MLP (P = 0.0061) and RF models (P = 0.038) showed lower misclassification rates (MCR) than the standard-report, weighting false positives and false negatives equally. Increasing the sensitivity of classifiers from 94.5 to 99.3% resulted in increase of MCR from 13.3/14.5 to 29.8/34.2% for MLP/RF, respectively. PET/CT-based machine-learning classifiers can achieve a high sensitivity (94.5%) to detect EBUS-positive LNs at a low misclassification rate. As the specificity decreases rapidly above that level, a combined test of a PET/CT-based MLP/RF classifier and EBUS-TBNA is recommended for radiation target volume definition.

SUBMITTER: Guberina M

PROVIDER: S-EPMC9584941 | biostudies-literature | 2022 Oct

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Prediction of malignant lymph nodes in NSCLC by machine-learning classifiers using EBUS-TBNA and PET/CT.

Guberina Maja M Herrmann Ken K Pöttgen Christoph C Guberina Nika N Hautzel Hubertus H Gauler Thomas T Ploenes Till T Umutlu Lale L Wetter Axel A Theegarten Dirk D Aigner Clemens C Eberhardt Wilfried E E WEE Metzenmacher Martin M Wiesweg Marcel M Schuler Martin M Karpf-Wissel Rüdiger R Santiago Garcia Alina A Darwiche Kaid K Stuschke Martin M

Scientific reports 20221020 1

Accurate determination of lymph-node (LN) metastases is a prerequisite for high precision radiotherapy. The primary aim is to characterise the performance of PET/CT-based machine-learning classifiers to predict LN-involvement by endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) in stage-III NSCLC. Prediction models for LN-positivity based on [<sup>18</sup>F]FDG-PET/CT features were built using logistic regression and machine-learning models random forest (RF) and multi ...[more]

PMID: 36266403

Similar Datasets

Project description:BackgroundStaging of non-small-cell lung cancer (NSCLC) is a multidisciplinary process involving imaging, endoscopic and surgical techniques. This study aims at investigating the diagnostic accuracy of 18F-FDG PET/CT, CT scan, and endobronchial ultrasound/transbronchial needle aspirate (EBUS/TBNA) in preoperative mediastinal lymph nodes (MLNs) staging of NSCLC.MethodsWe identified all patients who were diagnosed with NSCLC at the King Hussein Cancer Center in Amman, Jordan, between July 2011 and December 2017. We collected their relevant clinical, radiological, and histopathological findings. The per-patient analysis was performed on all patients (N = 101) and then on those with histopathological confirmation (N = 57), followed by a per-lymph-node-station basis overall, and then according to distinct N-stage categories.Results18F-FDG PET/CT, in comparison to CT, had a better sensitivity (90.5% vs. 75%, p = 0.04) overall and in patients with histopathological confirmation (83.3% vs. 54.6%), and better specificity (60.5% vs. 43.6%, p = 0.01) overall and in patients with histopathological confirmation in MLN staging (60.6% vs. 38.2%). Negative predictive value of mediastinoscopy, EBUS/TBNA, and 18F-FDG PET/CT were (87.1%), (90.91%), and (83.33%) respectively. The overall accuracy was highest for mediastinoscopy (88.6%) and EBUS/TBNA (88.2%), followed by 18F-FDG PET/CT (70.2%). Dividing patients into N1 disease vs. those with N2/N3 disease yielded similar findings. Comparison between 18F-FDG PET/CT and EBUS/TBNA in patients with histopathological confirmation shows 28 correlated true positive and true negative findings with final N-staging. In four patients, 18F-FDG PET/CT detected metastatic MLNs that would have otherwise remained undiscovered by EBUS/TBNA alone. Lymph nodes with a maximal standardized uptake value (SUVmax) more than 3 were significantly more likely to be true-positive.ConclusionMultimodality staging of the MLNs in NSCLC is essential to provide accurate staging and the appropriate treatment. 18F-FDG PET/CT has better overall diagnostic utility when compared to the CT scan. The NPV of 18F-FDG PET/CT in MLNs is reliable and comparable to the NPV of EBUS/TBNA. SUVmax of MLNs can help in predicting metastases, but nevertheless, a positive 18F-FDG PET/CT MLNs particularly if such a result would change the treatment plan, should be verified histopathologically.

Project description:PurposeThe aim of this study was to compare the pattern of intra-patient spread of lymph-node (LN)-metastases within the mediastinum as assessed by 18F-FDG PET/CT and systematic endobronchial ultrasound-guided transbronchial-needle aspiration (EBUS-TBNA) for precise target volume definition in stage III NSCLC.MethodsThis is a single-center study based on our preceding investigation, including all consecutive patients with initial diagnosis of stage IIIA-C NSCLC, receiving concurrent radiochemotherapy (12/2011-06/2018). Inclusion criteria were curative treatment intent, 18F-FDG PET/CT and EBUS-TBNA prior to start of treatment. The lymphatic drainage was classified into echelon-1 (ipsilateral hilum), echelon-2 (ipsilateral LN-stations 4 and 7) and echelon-3 (rest of the mediastinum, contralateral hilum). The pattern of spread was classified according to all permutations of echelon-1, echelon-2, and echelon-3 EBUS-TBNA findings.ResultsIn total, 180 patients were enrolled. Various patterns of LN-spread could be identified. Skip lesions with an involved echelon distal from an uninvolved one were detected in less than 10% of patients by both EBUS-TBNA and PET. The pattern with largest asymmetry was detected in cases with EBUS-TBNA- or PET-positivity at all three echelons (p < 0.0001, exact symmetry test). In a multivariable logistic model for EBUS-positivity at echelon-3, prognostic factors were PET-positivity at echelon-3 (Hazard ratio (HR) = 12.1; 95%-CI: 3.2-46.5), EBUS-TBNA positivity at echelon-2 (HR = 6.7; 95%-CI: 1.31-31.2) and left-sided tumor location (HR = 4.0; 95%-CI: 1.24-13.2). There were significantly less combined ipsilateral upper (LN-stations 2 and 4) and lower (LN-station 7) mediastinal involvements (16.8% of patients) with EBUS-TBNA than with PET (38.9%, p < 0.0001, exact symmetry test). EBUS-TBNA detected a lobe specific heterogeneity between the odds ratios of LN-positivity in the upper versus lower mediastinum (p = 0.0021, Breslow-Day test), while PET did not (p = 0.19).ConclusionFrequent patterns of LN-metastatic spread could be defined by EBUS-TBNA and PET and discrepancies in the pattern were seen between both methods. EBUS-TBNA showed more lobe and tumor laterality specific patterns of LN-metastases than PET and skipped lymph node stations were rare. These systematic relations offer the opportunity to further refine multi-parameter risk of LN-involvement models for target volume delineation based on pattern of spread by EBUS-TBNA and PET.

Project description:PURPOSE/INTRODUCTION: [18F]FDG-PET/CT is the standard imaging-technique for radiation treatment (RT) planning in locally advanced non-small cell lung cancer (NSCLC). The purpose of this study was to examine the additional value of endobronchial-ultrasound transbronchial needle aspiration (EBUS-TBNA) to standard PET/CT for mediastinal lymph-node (LN) staging and its impact on clinical target volume (CTV).Materials and methodsAll consecutive patients with primary stage III NSCLC who underwent [18F]FDG-PET/CT and EBUS-TBNA prior to RT were analyzed from 12/2011 to 06/2018. LN-stations were assessed by an expert-radiologist and a nuclear medicine-physician. CTV was evaluated by two independent radiation oncologists. LNs were grouped with increasing distance along the lymphatic chains from primary tumor into echelon-1 (ipsilateral hilum), echelon-2 (LN-station 7 and ipsilateral 4), and echelon-3 (remaining mediastinum and contralateral hilum).ResultsA total of 675 LN-stations of which 291 were positive for tumor-cells, were sampled by EBUS-TBNA in 180 patients. The rate of EBUS-positive LNs was 43% among all sampled LNs. EBUS-positivity in EBUS-probed LNs decreased from 85.8% in echelon-1 LNs to 42.4%/ 9.6% in echelon-2/ -3 LNs, respectively (p < 0.0001, Fisher's exact test). The false discovery rate of PET in comparison with EBUS results rose from 5.3% in echelon-1 to 32.9%/ 69.1% in echelon-2/ -3 LNs, respectively (p < 0.0001, Fisher's exact test). Sensitivity and specificity of FDG-PET/CT ranged from 85 to 99% and 67 to 80% for the different echelons. In 22.2% patients, EBUS-TBNA finding triggered changes of the treated CTV, compared with contouring algorithms based on FDG-avidity as the sole criterion for inclusion. CTV was enlarged in 6.7% patients due to EBUS-positivity in PET-negative LN-station and reduced in 15.5% by exclusion of an EBUS-negative but PET-positive LN-station.ConclusionThe false discovery rate of [18F]FDG-PET/CT increased markedly with distance from the primary tumor. Inclusion of systematic mediastinal LN mapping by EBUS-TBNA in addition to PET/CT has the potential to increase accuracy of target volume definition, particularly in echelon-3 LNs. EBUS-TBNA is recommended as integral part of staging for radiochemotherapy in stage III NSCLC.

Project description:BackgroundPatients with mediastinal lymph node enlargement (MLNE) are diagnosed depending on lymph node biopsy. Whereas, how to obtain larger tissue masses from mediastinal lymph nodes and improve the diagnostic yield of the disease remains to be investigated.ObjectivesAiming to assess the diagnostic value of endobronchial ultrasound-guided intranodal forceps biopsy via transbronchial laser photoablation (EBUS-IFB-TLP) in patients with MLNE.DesignA prospective, self-controlled study.MethodsThis study was conducted on 67 MLNE patients requiring a lymph node biopsy for diagnosis at the Henan Provincial People's Hospital and the Fuwai Central China Cardiovascular Hospital in China, from January 2020 to December 2022. Each patient underwent endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA group) and EBUS-IFB-TLP (EBUS-IFB-TLP group) on the same mediastinal lymph node for biopsies. The operation time, diagnostic efficiency, and complication rates of the two biopsy methods were compared.ResultsThe number of diagnosed patients in the EBUS-IFB-TLP and the EBUS-TBNA groups was 65 (97.0%) and 57 (85.1%), respectively (p = 0.021). In the EBUS-IFB-TLP group, 28 cases (96.6%) were diagnosed with lung cancer and were classified into different epithelial types. In the EBUS-TBNA group, there were 27 cases (93.1%) diagnosed with lung cancer, of which 26 (89.7%) were classified into different epithelial types. There were 37 (97.4%) and 30 (78.9%) non-lung cancer patients diagnosed in the EBUS-IFB-TLP and EBUS-TBNA groups, respectively (p = 0.039), while 27 cases (96.4%) of sarcoidosis in the EBUS-IFB-TLP group and 20 cases (71.4%) of sarcoidosis in the EBUS-TBNA group were diagnosed (p = 0.016). The percentages of intraoperative mild to moderate bleeding complications were 23.9% (16/67) and 14.9% (10/67) in the EBUS-IFB-TLP and in the EBUS-TBNA groups, respectively (p = 0.109).ConclusionThis study demonstrated that EBUS-IFB-TLP could be a feasible and effective method in the diagnosis of patients with MLNE, presenting an analogous safety profile compared with EBUS-TBNA. Further studies are needed to verify the diagnostic performance of EBUS-IFB-TLP for MLNE.

Dataset Information

Prediction of malignant lymph nodes in NSCLC by machine-learning classifiers using EBUS-TBNA and PET/CT.

Publications

Prediction of malignant lymph nodes in NSCLC by machine-learning classifiers using EBUS-TBNA and PET/CT.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets