The application of artificial intelligence for Rapid On-Site Evaluation during flexible bronchoscopy

Background

Rapid On-Site Evaluation (ROSE) during flexible bronchoscopy (FB) can improve the adequacy of biopsy specimens and diagnostic yield of lung cancer. However, the lack of cytopathologists has restricted the wide use of ROSE.

Objective

To develop a ROSE artificial intelligence (AI) system using deep learning techniques to differentiate malignant from benign lesions based on ROSE cytological images, and evaluate the clinical performance of the ROSE AI system.

Method

6357 ROSE cytological images from 721 patients who underwent transbronchial biopsy were collected from January to July 2023 at the Tangdu Hospital, Air Force Medical University. A ROSE AI system, composed of a deep convolutional neural network (DCNN), was developed to identify whether there were malignant cells in the ROSE cytological images. Internal testing, external testing, and human-machine competition were used to evaluate the performance of the system.

Results

The ROSE AI system identified images containing lung malignant cells with the accuracy of 92.97% and 90.26% on the internal testing dataset and external testing dataset respectively, and its performance was comparable to that of the experienced cytopathologist. The ROSE AI system also showed promising performance in diagnosing lung cancer based on ROSE cytological images, with accuracy of 89.61% and 87.59%, and sensitivity of 90.57% and 94.90% on the internal testing dataset and external testing dataset respectively. More specifically, the agreement between the ROSE AI system and the experienced cytopathologist in diagnosing common types of lung cancer, including squamous cell carcinoma, adenocarcinoma, and small cell lung cancer, demonstrated almost perfect consistency in both the internal testing dataset (κ  =   0.930 ) and the external testing dataset (κ  =   0.932 ).

Conclusions

The ROSE AI system demonstrated feasibility and robustness in identifying specimen adequacy, showing potential enhancement in the diagnostic yield of FB. Nevertheless, additional enhancements, incorporating a more diverse range of training data and leveraging advanced AI models with increased capabilities, along with rigorous validation through extensive multi-center randomized control assays, are crucial to guarantee the seamless and effective integration of this technology into clinical practice.

Assessing ROSE for adequacy of EBUS-TBNA compared with a direct-to-cell block approach as a response to the COVID-19 pandemic

Introduction:

Rapid on-site evaluation (ROSE) has been used during the endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) procedure as standard practice. Because of the COVID-19 (coronavirus disease 2019) pandemic, our institute had had to discontinue ROSE and adopt a direct-to-cell block approach. In the present study, we aimed to determine whether this change has had significant effects on the cytopathology quality.

Materials and methods:

A total of 1903 EBUS-TBNA cases from 734 patients were collected (1097 cases with ROSE for 452 patients; 806 cases without ROSE but with direct-to-cell block for 282 patients). The clinical and cytology data were analyzed using SAS, version 9.4, software to render calculated standardized residuals and a fitted multivariate generalized linear model.

Results:

On average, a biopsy from a patient with ROSE was 0.936 (=exp −0.066) times less likely to be reported as satisfactory compared with a biopsy from a patient without ROSE, although the difference was not statistically significant (P = 0.785). The inadequacy rate of EBUS-TBNA was 6.4% higher on average for cases with ROSE compared with a direct-to-cell block approach. However, this difference was also not statistically significant. The proportions of biopsies reported as diagnostic for malignancy and other were significantly different between the ROSE and no-ROSE groups with a standardized residual of 1.80 (P = 0.036) and −2.27 (P = 0.012), respectively.

Conclusions

Discontinuing ROSE and using a direct-to-cell block approach had no negative effects on cytopathology quality. This practice can be considered acceptable during the COVID-19 pandemic when social distancing and the shortage of staff and supplies have resulted in challenges to delivering quality care to cancer patients whose treatment cannot be postponed.

Rapid on-site evaluation (ROSE) during endobronchial ultrasound bronchoscopy (EBUS) in the diagnosis of granulomatous diseases

AIM

We aimed to evaluate the role of Rapid On-Site Evaluation (ROSE) in improving the diagnostic contribution of EBUS-guided transbronchial needle aspiration (EBUS-TBNA) in the diagnosis of granulomatous lymphadenitis.

METHODS

We retrospectively included all patients in the study who underwent EBUS-TBNA (n = 405) from September 2013 to September 2017. All cases who underwent EBUS-TBNA and were found to have "granuloma" in the final pathological diagnosis (n = 143) were included in the research.

RESULTS

One hundred forty-three cases who underwent EBUS and whose final pathological diagnosis was granuloma were included in the study. A total of 231 lymph nodes were sampled. The final diagnosis was found to be compatible with the examination during the procedure for 55 (85.9%) out of 64 patients for whom ROSE was performed. In addition, while 79.7% of the 64 patients who underwent ROSE were tested positive for granuloma, 9.4% were suspected to have granuloma, and no diagnostic result was obtained for 10.9%. The number of lymph node needle aspirations was statistically significantly lower in patients for whom ROSE was performed (P < .001). The final diagnosis was found to be compatible with the examination during the procedure for 55 (85.9%) out of 64 patients for whom ROSE was performed. According to the evaluation made on the duration, the procedure was completed in the first 15 minutes in 95.4% of the patients for whom ROSE was performed, while the procedure took longer than 15 minutes in 93.6% of the patients for whom ROSE was not performed.

CONCLUSION

We found a high incidence of congruence between ROSE and the final cytological diagnosis if granulomas were identified in the final cytology. Performing ROSE shortens the duration of the procedure and reduced sampling thus preventing complications that may occur due to prolongation of the procedure.

Interobserver agreement between pathologist, pulmonologist and molecular pathologist to estimate the tumour burden in rapid on-site evaluation smears from endosonography and guided bronchoscopy

OBJECTIVE

A growing number of studies have suggested that non-pathologists can reliably assess the adequacy and malignancy in rapid on-site evaluation (ROSE) smears prepared during endoscopic sampling procedures. However, no study has verified whether they can also consistently estimate the tumour burden, which is critical for the molecular profiling of lung cancer. We aimed to assess the interobserver agreement (IOA) between a pathologist, a pulmonologist (previously trained in lung and lymph node cytopathology) and a molecular pathologist for the tumour burden in ROSE smears.

METHODS

The ROSE smears of consecutive patients with suspected lung cancer undergoing endosonography or guided bronchoscopy were assessed independently by a pathologist, a pulmonologist and a molecular pathologist (gold standard). The IOA for the tumour burden, assessed through k-statistics, was the primary outcome.

RESULTS

A total of 322 ROSE smears obtained from 162 patients were evaluated. The IOA between the molecular pathologist and pulmonologist was very good (moderate to substantial), although slightly inferior to the IOA between the molecular pathologist and pathologist in the whole slide set (k: 0.707, 95% confidence interval [CI]: 0.677-0.739 vs 0.793, 95% CI: 0.762-0.815), as well as in smears prepared from lymphadenopathy (k: 0.783, 95% CI: 0.760-0.855 vs 0.827, 95% CI: 0.728-0.892) or from pulmonary nodules/masses (k: 0.558, 95% CI: 0.416-0.686 vs 0.715, 95% CI: 0.621-0.767).

CONCLUSIONS

A professionally trained pulmonologist can reliably estimate the tumour burden in bronchoscopically derived ROSE smears, especially in the setting of lymphadenopathy. This can be particularly useful in institutions where a cytopathologist is not available regularly.

The role of rapid on site evaluation on touch imprint cytology and brushing during conventional bronchoscopy

BACKGROUND

The increase in immunohistochemical and molecular predictive tests in lung cancer requires new strategies for managing small samples taken during bronchoscopic procedures. The value of Rapid On Site Evaluation (ROSE) during conventional bronchoscopic procedures on endobronchial neoplasms in optimizing small biopsies and cytologlogical tissue specimens for diagnostic testing, and ancillary studies was evaluated.

METHOD

ROSE on touch imprint cytology (TIC) and brushing was performed on 690 consecutive cases of patients undergoing biopsies, using fiber optic bronchoscopy. Immunohistochemical assay for PD-L1, ALK, and ROS1 and molecular testing, via next generation technique for EGFR, KRAS, and BRAF, were performed.

RESULTS

The concordance between ROSE and final diagnoses was almost perfect for brushing (sensitivity: 0.84; specificity: 0.96), and less so for touch preparations (sensitivity: 0.77; specificity: 0.89). Immunohistochemical assay for PD-L1 was evaluated on 256 bioptic cases with only six unsuitable samples. Material available for immunohistochemistry for ALK was sufficient in 151 biopsies with no inadequate cases. ROS1 was evaluated in 132 biopsies, with only two unsuitable samples. Molecular analysis was performed on 128 biopsies, 29 TIC, and 17 brushing. Out of these, only ten were considered to be unsuitable.

CONCLUSIONS

ROSE is an effective procedure for monitoring the quality and quantity of material taken during conventional bronchoscopic procedures for evaluating the suitability of small samples that must undergo immunohistochemical and molecular assay.

Diagnostic value of rapid on-site evaluation during endobronchial ultrasound with a guide sheath for peripheral pulmonary lesions

OBJECTIVE

To evaluate the applied value of rapid on-site evaluation during endobronchial ultrasound (EBUS) with a guide sheath for peripheral pulmonary lesions (PPLs).

METHODS

Consecutive patients who underwent EBUS with a guide sheath for PPLs at our hospital from December 2015 to June 2017 in this retrospective study. The samples obtained from each operation were made rapid on-site evaluation at the same time. The results of rapid on-site evaluation were compared with the pathological diagnosis.

RESULTS

A total of 127 PPLs in 124 patients were included in the study. 70 lesions were malignancy in the final pathological diagnosis. The sensitivity, specificity, positive predictive value, negative predictive value and diagnostic accuracy of rapid on-site evaluation for malignancy during EBUS with a guide sheath for PPLs was 88.6%, 98.2%, 98.4%, 87.5% and 92.9%, respectively.

CONCLUSIONS

Rapid on-site evaluation during EBUS with a guide sheath has a high diagnostic value for malignant PPLs.

Diagnostic value of rapid on-site evaluation during transbronchial biopsy for peripheral lung cancer

BACKGROUND

An increasing number of pulmonary lesions-particularly, peripheral lung lesions-are identified with current technological advancements. Notably, the yield of traditional bronchoscopy for the diagnosis of peripheral lung lesions is low. This study evaluated the diagnostic value of rapid on-site evaluation (ROSE) during transbronchial biopsy for peripheral lung cancer.

METHODS

This study included 641 patients who underwent transbronchial biopsy for suspected lung cancer at the Respiratory Department of Tianjin Medical University General Hospital between January 2012 and December 2016. Based on whether ROSE was used, patients were assigned to the ROSE group (353 patients) or non-ROSE group (288 patients). In the ROSE group, several air-dried smears were processed with Diff-Quik staining; the remaining samples were placed in 10% formalin. Diagnostic yields for central and peripheral lung cancer were compared between the two groups. In addition, ROSE results were compared with final diagnoses.

RESULTS

Diagnostic yield for peripheral lung cancer, stratified by pathology, was significantly higher in the ROSE group than in the non-ROSE group (42.9% vs. 30.7%, P < 0.05). The sensitivity, specificity, positive predictive value, negative predictive value and diagnostic accuracy of ROSE for peripheral pulmonary lesions were 90.0%, 89.5%, 94.0%, 82.8% and 89.8%, respectively. Conformance was high between ROSE and final pathologic evaluations during transbronchial biopsy (Kappa = 0.780, P = 0.035). There were no procedure-related deaths.

CONCLUSIONS

ROSE during conventional transbronchial biopsy improves diagnostic yield, stratified by pathology, for patients with peripheral lung cancer via live feedback. Moreover, ROSE diagnosis correlates with final cytopathological diagnosis.

A Trained Pulmonologist Can Reliably Assess Endosonography-Derived Lymph Node Samples during Rapid On-Site Evaluation

BACKGROUND

The widespread use of rapid on-site evaluation is hampered by constraints related to time and resources, inadequate reimbursement, and evidence from randomized trials that show a lack of increase in diagnostic yield and specimen adequacy associated with its usage.

OBJECTIVE

We aimed to verify whether a pulmonologist can assess endosonography-derived lymph node samples after a comprehensive and reproducible training provided by a specialist pathologist.

METHODS

Prospective, observational trial structured in three phases. In the first (training) phase, a pathologist critically evaluated the smears from 150 archival endosonography cases with a pulmonologist. In the second (test) phase, the pulmonologist was asked to assess 50 archival endosonography-derived samples. In the last (real-life) phase, the pulmonologist classified the samples from 200 patients during the endosonography. The overall agreement between pulmonologist and pathologist (gold standard), assessed through κ-statistics, was the primary outcome. The agreement for the identification of specific cytological categories was the secondary outcome.

RESULTS

The overallagreement between pulmonologist and pathologist was 84% (κ0.765, 95% CI 0.732-0.826) in the test phase and 89.7% (κ 0.844, 95% CI 0.799-0.881) in the real-life phase. The agreement for specific cytological categories was 92.7% (95% CI 0.824-0.980) for inadequate samples, 90.3% (95% CI 84.5-94.5%) for reactive lymphadenopathies, 90.5% (95% CI 0.845-0.946) for malignancy, and 73% (95% CI 0.515-0.897) for granulomatous samples.

CONCLUSIONS

A trained pulmonologist can reliably assess adequacy and malignancy for endosonography-derived samples, which could be useful in institutions where a cytopathologist/cytotechnician is not available regularly.

Endobronchial ultrasound-guided transbronchial needle aspiration in sarcoidosis: Beyond the diagnostic yield

Endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) is the commonly used technique for pathological confirmation of clinically suspected sarcoidosis, mostly owing to its consistently high success rate in the detection of granulomas. However, other possible advantages, which are less appreciated and often poorly studied, may also contribute to the wider use of EBUS-TBNA in the future. These advantages include refinement of differential diagnoses through the study of lymph node characteristics during B-mode examination; reduction of complications associated with bronchoscopy, as well as improved triage of the specimen for ancillary studies with the use of rapid on-site evaluation; optimization of the quality of the sample through the selection of a target area for biopsy with minimal vascularity and absence of calcifications by using the colour Doppler and the B-mode; and prediction of the presence of extensive lymph node fibrosis by using the strain elastography module. Yet, limitations and possible clinical drawbacks should also be acknowledged. Indeed, due to the lack of specificity of the pathology findings in EBUS-derived samples, the diagnosis of sarcoidosis is one of the exclusion and should remain essentially clinical. The external validity of EBUS-TBNA results in sarcoidosis is questionable, as they mainly derive from studies in populations with a high disease prevalence. Finally, the risk exists that the low morbidity and high diagnostic yield of EBUS-TBNA may lead to its overuse in patients with clinical/radiological findings specific enough to secure a clinical diagnosis of sarcoidosis.

Competence in flexible bronchoscopy and basic biopsy technique

Diagnostic bronchoscopy and tissue sampling techniques using forceps (endobronchial biopsy [EBB] and transbronchial biopsies [TBB]) or needle aspiration (transbronchial needle aspiration-TBNA), all performed with a flexible bronchoscope, are the basic elements of any interventional procedure. The flexible fibrobronchoscopy allows the visualization of the airways and is used both for diagnostic and therapeutic purposes. The working channel of both fibrobronchoscopes with optical fibers and videobronchoscopes, even if of relatively small diameter, allows the insertion of various diagnostic and therapeutic accessories. Fiber optic systems have been widely replaced by video cameras using a miniaturized charge-coupled device camera positioned at the end of the scope that provides electronic transmission of images to a monitor. The indications for both diagnostic and therapeutic fibrobronchoscopy derive from a correct evaluation of symptoms and objective signs of the patient and from the correct interpretation of imaging methods. Although bronchoscopy techniques keep evolving at a rapid pace, basic procedures such as bronchoalveolar lavage, transbronchial lung biopsy, and transbronchial needle aspiration still play a key role in pulmonary disease diagnostics, and therefore, these methods must still be part of the training of interventional pulmonologists. Trainees will acquire a thorough knowledge of thoracic anatomy and become skilled in the interpretation of thoracic imaging, after which they will be given a theoretical and practical training course on virtual reality simulators, on animal or cadaver models, the effectiveness of which has been fully demonstrated by scientific studies. Specific DOPS tests have been developed for a qualitative evaluation of procedures on simulators, on animal models and on the patient.

Conventional Transbronchial Needle Aspiration Versus Endobronchial Ultrasound-guided Transbronchial Needle Aspiration, With or Without Rapid On-Site Evaluation, for the Diagnosis of Sarcoidosis: A Randomized Controlled Trial

BACKGROUND

Endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) as a standalone modality is superior to conventional TBNA (c-TBNA) for the diagnosis of sarcoidosis. However, the overall yield is not different if combined with endobronchial biopsy (EBB) and transbronchial lung biopsy (TBLB). The utility of rapid on-site evaluation (ROSE) in a comparative evaluation of EBUS-TBNA versus c-TBNA for the diagnosis of sarcoidosis has not been previously evaluated.

METHODS

Eighty patients with suspected sarcoidosis were randomized 1:1:1:1 into 4 groups: c-TBNA without ROSE (TBNA-NR), c-TBNA with ROSE (TBNA-R), EBUS-TBNA without ROSE (EBUS-NR), and EBUS-TBNA with ROSE (EBUS-R). EBB and TBLB were performed in all patients. Primary objective was detection of granulomas for combined procedure. Secondary objectives were individual procedure yields, sedation dose, and procedure duration. Patients without a diagnosis following c-TBNA subsequently underwent EBUS-TBNA.

RESULTS

A total of 74 patients were finally diagnosed with sarcoidosis. Overall, granuloma detection was not significantly different between the 4 groups (68% in TBNA-NR, 89% in TBNA-R, 84% in EBUS-NR, and 83% in EBUS-R groups, P=0.49). The yield of c-TBNA in the TBNA-NR group was lower compared with that in TBNA-R group and EBUS-TBNA in EBUS groups (32%, 72%, 68%, and 67% for TBNA-NR, TBNA-R, EBUS-NR, and EBUS-R groups, respectively, P=0.04). Additional 20% patients were diagnosed when EBUS-TBNA was performed following a nondiagnostic bronchoscopy procedure in the TBNA-NR group. Sedation requirement and procedure duration were significantly lower with c-TBNA as compared with EBUS-TBNA (P<0.001).

CONCLUSION

When performing TBNA in the setting of suspected sarcoidosis, we found c-TBNA with ROSE and EBUS-TBNA (with or without ROSE) to be superior to c-TBNA alone. Whether c-TBNA with ROSE is equivalent to EBUS-TBNA cannot be determined from our study due to small sample size/low power.

Advanced medical interventions in pleural disease

The burden of a number of pleural diseases continues to increase internationally. Although many pleural procedures have historically been the domain of interventional radiologists or thoracic surgeons, in recent years, there has been a marked expansion in the techniques available to the pulmonologist. This has been due in part to both technological advancements and a greater recognition that pleural disease is an important subspecialty of respiratory medicine. This article summarises the important literature relating to a number of advanced pleural interventions, including medical thoracoscopy, the insertion and use of indwelling pleural catheters, pleural manometry, point-of-care thoracic ultrasound, and image-guided closed pleural biopsy. We also aim to inform the reader regarding the latest updates to more established procedures such as chemical pleurodesis, thoracentesis and the management of chest drains, drawing on contemporary data from recent randomised trials. Finally, we shall look to explore the challenges faced by those practicing pleural medicine, especially relating to training, as well as possible future directions for the use and expansion of advanced medical interventions in pleural disease.

Direct comparison of liquid-based and smear-based cytology with and without rapid on-site evaluation for fine needle aspirates of thoracic tumors

BACKGROUND

Liquid-based cytology (LBC) and rapid on-site evaluation (ROSE) are proposed to improve the quality of fine needle aspirates (FNA) and their diagnostic yield compared with conventional smear cytology (CSC). This prospective study directly compared outcomes of sonar-guided FNA of thoracic tumors supported by LBC, CSC, or CSC with ROSE.

METHODS

Three aspirates each for both LBC and CSC with separate 22G spinal needles in a randomized, alternating sequence during 64 transthoracic FNA of thoracic tumors were collected. Smears were prepared by cytology staff on site but evaluated with ROSE only when all six samples had been collected. If no diagnostic material was found on the first three CSC additional needle passes guided by ROSE were performed.

RESULTS

Final diagnoses were non-small cell lung cancer in 50 (78.1%), small cell lung cancer in 11 (17.2%), mesothelioma in 1 (1.6%), and inflammation in 2 cases (3.1%), respectively. LBC and CSC were diagnostic in 42 (65.6%) and 49 (76.6%) cases, respectively (P = 0.039), with both methods diagnostic in 41 cases (64.1%). Fifteen cases (23.4%) remained undiagnosed following three passes for CSC but 9 (14.1%) of these were diagnosed using FNA and ROSE with a total yield of 58 cases (90.6%; P < 0.001).

CONCLUSION

The diagnostic yield of transthoracic FNA submitted for LBC is significantly lower than with CSC when slides are prepared professionally. ROSE significantly increases the yield of transthoracic FNA.