Role of endoscopic ultrasound-guided fine needle aspiration in the diagnosis of mass lesions

Development and Validation of Contrast-Enhanced CT-Based Deep Transfer Learning and Combined Clinical-Radiomics Model to Discriminate Thymomas and Thymic Cysts: A Multicenter Study

RATIONALE AND OBJECTIVES

This study aims to evaluate the feasibility and effectiveness of deep transfer learning (DTL) and clinical-radiomics in differentiating thymoma from thymic cysts.

MATERIALS AND METHODS

Clinical and imaging data of 196 patients pathologically diagnosed with thymoma and thymic cysts were retrospectively collected from center 1. (training cohort: n = 137; internal validation cohort: n = 59). An independent external validation cohort comprised 68 thymoma and thymic cyst patients from center 2. Region of interest (ROI) delineation was performed on contrast-enhanced chest computed tomography (CT) images, and eight DTL models including Densenet 169, Mobilenet V2, Resnet 101, Resnet 18, Resnet 34, Resnet 50, Vgg 13, Vgg 16 were constructed. Radiomics features were extracted from the ROI on the CT images of thymoma and thymic cyst patients, and feature selection was performed using intra-observer correlation coefficient (ICC), Spearman correlation analysis, and least absolute shrinkage and selection operator (LASSO) algorithm. Univariate analysis and multivariable logistic regression (LR) were used to select clinical-radiological features. Six machine learning classifiers, including LR, support vector machine (SVM), k-nearest neighbors (KNN), Light Gradient Boosting Machine (LightGBM), Adaptive Boosting (AdaBoost), and Multilayer Perceptron (MLP), were used to construct Radiomics and Clinico-radiologic models. The selected features from the Radiomics and Clinico-radiologic models were fused to build a Combined model. Receiver operating characteristic curve (ROC), calibration curve, and decision curve analysis (DCA) were used to evaluate the discrimination, calibration, and clinical utility of the models, respectively. The Delong test was used to compare the AUC between different models. K-means clustering was used to subdivide the lesions of thymomas or thymic cysts into subregions, and traditional radiomics methods were used to extract features and compare the ability of Radiomics and DTL models to reflect intratumoral heterogeneity using correlation analysis.

RESULTS

The Densenet 169 based on DTL performed the best, with AUC of 0.933 (95% CI: 0.875-0.991) in the internal validation cohort and 0.962 (95% CI: 0.923-1.000) in the external validation cohort. The AdaBoost classifier achieved AUC of 0.965 (95% CI: 0.923-1.000) and 0.959 (95% CI: 0.919-1.000) in the internal and external validation cohorts, respectively, for the Radiomics model. The LightGBM classifier achieved AUC of 0.805 (95% CI: 0.690-0.920) and 0.839 (95% CI: 0.736-0.943) in the Clinico-radiologic model. The AUC of the Combined model in the internal and external validation cohorts was 0.933 (95% CI: 0.866-1.000) and 0.945 (95% CI: 0.897-0.994), respectively. The results of the Delong test showed that the Radiomics model, DTL model, and Combined model outperformed the Clinico-radiologic model in both internal and external validation cohorts (p-values were 0.002, 0.004, and 0.033 in the internal validation cohort, while in the external validation cohort, the p-values were 0.014, 0.006, and 0.015, respectively). But there was no statistical difference in performance among the three models (all p-values <0.05). Correlation analysis showed that radiomics performed better than DTL in quantifying intratumoral heterogeneity differences between thymoma and thymic cysts.

CONCLUSION

The developed DTL model and the Combined model based on radiomics and clinical-radiologic features achieved excellent diagnostic performance in differentiating thymic cysts from thymoma. They can serve as potential tools to assist clinical decision-making, particularly when endoscopic biopsy carries a high risk.

The New American Joint Committee on Cancer T staging system for stomach: increased complexity without clear improvement in predictive accuracy for endoscopic ultrasound

Background

The efficacy of endoscopic ultrasound (EUS) for determining the T category of gastric cancer is variable. The aim of this study was to evaluate the superiority of EUS by using the 6th edition American Joint Committee on Cancer (AJCC) staging system for stomach cancer compared to the new 7th/8th edition.

Methods

A retrospective analysis of clinical and EUS imaging features of 348 gastric carcinoma patients who underwent radical resection were retrospectively analyzed. Differences between the 6th and 7th/8th edition T staging systems for preoperative EUS evaluation were compared.

Results

The accuracy of EUS T staging was 72.4% for the 7th/8th edition and 78.4% for the 6th edition. T3 stage accuracy was significantly worse when the T3 group status was changed. The tumor location, echoendoscope type, and histological type were associated with inaccuracy. We further analyzed the EUS image features for each tumor T stage and found that an indistinctly visible muscularis propria (MP) or with obvious thickening was considered an indicator of lesions involved in the MP with a sensitivity of 81.3%; an MP completely disappeared and accompanied with a serosal layer intact may be a marker that the lesion invaded to the subserosa. We also found that irregularities in the outer edge of the gastric wall were markers of gastric serosal layer penetration with a positive predictive value of 92.2%.

Conclusions

The increased complexity of the 7th/8th edition T staging system is accompanied by worsening of the predictive accuracy for EUS as compared to the 6th edition. Furthermore, the tumor location, echoendoscope type, histological type, and EUS image features for each tumor T stage should warrant attention.

High Sensitivity of EUS-FNA and EUS-FNB in Lymphadenopathy Caused by Metastatic Disease: A Prospective Comparative Study

Background/Aims

The diagnostic work-up of lymphadenopathy is challenging but important to determine the correct therapy. Nevertheless, few studies have addressed the topic of endosonography (EUS)-guided tissue acquisition in lymphadenopathy. Therefore, we aimed to evaluate the accuracy and safety of EUS-guided fine-needle biopsy sampling (EUS-FNB) in intrathoracic and intraabdominal lymphadenopathy.

Methods

In a tertiary care center, patients with lymphadenopathy referred for EUS-guided sampling were included prospectively from 2014 to 2019 (NCT02360839). In all cases, EUS-FNB (22 gauge) and EUS-guided fine-needle aspiration (EUS-FNA) (25 gauge) were performed. The patients were randomized to the first needle pass with FNB or FNA. Study outcomes were the diagnostic accuracy and adverse event rate.

Results

Forty-eight patients were included (median age: 69 years [interquartile range, 59–76]; 24/48 females [50%]). The final diagnoses were metastasis (n=17), lymphoma (n=11), sarcoidosis (n=6), and inflammatory disease (n=14). The diagnostic performance of the two modalities was comparable, including a high sensitivity for metastatic nodes (EUS-FNB: 87% vs. EUS-FNA: 100%, p=0.5). The sensitivity for lymphoma was borderline superior in favor of EUS-FNB (EUS-FNB: 55% vs. EUS-FNA: 9%, p=0.06). No adverse events were recorded.

Conclusions

In lymphadenopathy, both EUS-FNB and EUS-FNA are safe and highly sensitive for metastatic lymph node detection. Lymphoma diagnosis is challenging regardless of the needle used.

Can computed tomography-based radiomics potentially discriminate between anterior mediastinal cysts and type B1 and B2 thymomas?

Background

Anterior mediastinal cysts (AMC) are often misdiagnosed as thymomas and undergo surgical resection, which caused unnecessary treatment and medical resource waste. The purpose of this study is to explore potential possibility of computed tomography (CT)-based radiomics for the diagnosis of AMC and type B1 and B2 thymomas.

Methods

A group of 188 patients with pathologically confirmed AMC (106 cases misdiagnosed as thymomas in CT) and thymomas (82 cases) and underwent routine chest CT from January 2010 to December 2018 were retrospectively analyzed. The lesions were manually delineated using ITK-SNAP software, and radiomics features were performed using the artificial intelligence kit (AK) software. A total of 180 tumour texture features were extracted from enhanced CT and unenhanced CT, respectively. The general test, correlation analysis, and LASSO were used to features selection and then the radiomics signature (radscore) was obtained. The combined model including radscore and independent clinical factors was developed. The model performances were evaluated on discrimination, calibration curve.

Results

Two radscore models were constructed from the unenhanced and enhanced phases based on the selected four and three features, respectively. The AUC, sensitivity, and specificity of the enhanced radscore model were 0.928, 89.3%, and 83.8% in the training dataset and 0.899, 84.6%, and 87.5% in the test dataset (higher than the unenhanced radscore model). The combined model of enhanced CT including radiomics features and independent clinical factors yielded an AUC, sensitivity and specificity of 0.941, 82.1%, and 94.6% in the training dataset and 0.938, 92.3%, and 87.5% in the test dataset (higher than the unenhanced combined model and enhanced radscore model).

Conclusions

The study suggested the possibility that the combined model in enhanced CT provided a potential tool to facilitate the differential diagnosis of AMC and type B1 and B2 thymomas.

Exploration of an effective training system for the diagnosis of pancreatobiliary diseases with EUS: A prospective study

Background and Objective:

There are limited data on multistage-based training programs focused on EUS. We aimed to explore an effective training system for diagnosing pancreaticobiliary diseases with EUS.

Materials and Methods:

Nine advanced endoscopy trainees (AETs) with less EUS experience from nine institutions were recruited. The training system consisted of multiple stages and multi-teaching methods, including biliopancreatic standard scanning, anatomy and imaging knowledge, simulator, hands-on operations, error correction, and case analysis over a 12-month training period. Grading for technical and cognitive skills was assessed using The EUS Skills Assessment Tool.

Results:

After training, the overall scores for radial (4.16 ± 0.21 vs. 1.46 ± 0.16, P < 0.01) and linear (4.43 ± 0.20 vs. 1.63 ± 0.23, P < 0.01) scanning were significantly improved. The aortopulmonary window/mediastinum station can be learned more easily by AETs compared with other stations (P = 0023). The scanning of the descending part of the duodenum seemed to improve the slowest after training (P = 0.0072), indicating that the descending part of the duodenum can be more difficult and should be the focus of training. Every teaching method heightened EUS competence, especially case analysis and hands-on operations. AETs achieved equivalent EUS competence after training despite their initial experience. Through a poststudy questionnaire, it was found that all AETs strongly agreed they were satisfied with the training system, and their confidence was greatly enhanced when EUS was performed independently.

Conclusions:

The current multistage and multi-methods training system showed efficient performance in the cognitive and technical competence of EUS. Descending part of duodenum scanning was difficult for beginners and should be the focus of training.

Endoscopic ultrasound-guided fine-needle aspiration and cytology for differentiating benign from malignant lymph nodes

BACKGROUND AND STUDY AIMS

Intra-abdominal and mediastinal lymphadenopathy are often difficult to diagnose, particularly in the absence of a primary lesion. Endosonography (EUS)-guided fine-needle aspiration and cytology (FNAC) has provided an easy and safe access to these lymph nodes, sparing the use of invasive and costly interventions. The main aim of this study is to assess the specificity, sensitivity, and predictive value of EUS-guided FNAC in the diagnosis of benign and malignant lymph nodes. In addition, the study aims to determine significant EUS features that could help in predicting lymph node malignancy.

PATIENTS AND METHODS

This prospective study included 142 patients with intra-abdominal or intrathoracic lymphadenopathy who were referred for EUS-guided FNAC because of inaccessibility by other imaging modalities. Ninety (63.3%) patients were found to have malignant lymph nodes, and 52 (36.6%) had lymphadenopathy of benign nature.

RESULTS

EUS-guided FNAC had a sensitivity and specificity of 92% and 100% respectively. It had positive and negative predictive values of 100% and 88% for malignancy, respectively. By logistic regression analysis, EUS features and shortest diameter were found to be potential predictors of malignancy with p-value of <0.0001.

CONCLUSION

EUS-guided FNAC is a powerful modality in the diagnosis of benign and malignant lymph nodes. Additional complementary EUS features could be added to this technique for definitive diagnosis.

Imaging modalities for characterising focal pancreatic lesions

BACKGROUND

Increasing numbers of incidental pancreatic lesions are being detected each year. Accurate characterisation of pancreatic lesions into benign, precancerous, and cancer masses is crucial in deciding whether to use treatment or surveillance. Distinguishing benign lesions from precancerous and cancerous lesions can prevent patients from undergoing unnecessary major surgery. Despite the importance of accurately classifying pancreatic lesions, there is no clear algorithm for management of focal pancreatic lesions.

OBJECTIVES

To determine and compare the diagnostic accuracy of various imaging modalities in detecting cancerous and precancerous lesions in people with focal pancreatic lesions.

SEARCH METHODS

We searched the CENTRAL, MEDLINE, Embase, and Science Citation Index until 19 July 2016. We searched the references of included studies to identify further studies. We did not restrict studies based on language or publication status, or whether data were collected prospectively or retrospectively.

SELECTION CRITERIA

We planned to include studies reporting cross-sectional information on the index test (CT (computed tomography), MRI (magnetic resonance imaging), PET (positron emission tomography), EUS (endoscopic ultrasound), EUS elastography, and EUS-guided biopsy or FNA (fine-needle aspiration)) and reference standard (confirmation of the nature of the lesion was obtained by histopathological examination of the entire lesion by surgical excision, or histopathological examination for confirmation of precancer or cancer by biopsy and clinical follow-up of at least six months in people with negative index tests) in people with pancreatic lesions irrespective of language or publication status or whether the data were collected prospectively or retrospectively.

DATA COLLECTION AND ANALYSIS

Two review authors independently searched the references to identify relevant studies and extracted the data. We planned to use the bivariate analysis to calculate the summary sensitivity and specificity with their 95% confidence intervals and the hierarchical summary receiver operating characteristic (HSROC) to compare the tests and assess heterogeneity, but used simpler models (such as univariate random-effects model and univariate fixed-effect model) for combining studies when appropriate because of the sparse data. We were unable to compare the diagnostic performance of the tests using formal statistical methods because of sparse data.

MAIN RESULTS

We included 54 studies involving a total of 3,196 participants evaluating the diagnostic accuracy of various index tests. In these 54 studies, eight different target conditions were identified with different final diagnoses constituting benign, precancerous, and cancerous lesions. None of the studies was of high methodological quality. None of the comparisons in which single studies were included was of sufficiently high methodological quality to warrant highlighting of the results. For differentiation of cancerous lesions from benign or precancerous lesions, we identified only one study per index test. The second analysis, of studies differentiating cancerous versus benign lesions, provided three tests in which meta-analysis could be performed. The sensitivities and specificities for diagnosing cancer were: EUS-FNA: sensitivity 0.79 (95% confidence interval (CI) 0.07 to 1.00), specificity 1.00 (95% CI 0.91 to 1.00); EUS: sensitivity 0.95 (95% CI 0.84 to 0.99), specificity 0.53 (95% CI 0.31 to 0.74); PET: sensitivity 0.92 (95% CI 0.80 to 0.97), specificity 0.65 (95% CI 0.39 to 0.84). The third analysis, of studies differentiating precancerous or cancerous lesions from benign lesions, only provided one test (EUS-FNA) in which meta-analysis was performed. EUS-FNA had moderate sensitivity for diagnosing precancerous or cancerous lesions (sensitivity 0.73 (95% CI 0.01 to 1.00) and high specificity 0.94 (95% CI 0.15 to 1.00), the extremely wide confidence intervals reflecting the heterogeneity between the studies). The fourth analysis, of studies differentiating cancerous (invasive carcinoma) from precancerous (dysplasia) provided three tests in which meta-analysis was performed. The sensitivities and specificities for diagnosing invasive carcinoma were: CT: sensitivity 0.72 (95% CI 0.50 to 0.87), specificity 0.92 (95% CI 0.81 to 0.97); EUS: sensitivity 0.78 (95% CI 0.44 to 0.94), specificity 0.91 (95% CI 0.61 to 0.98); EUS-FNA: sensitivity 0.66 (95% CI 0.03 to 0.99), specificity 0.92 (95% CI 0.73 to 0.98). The fifth analysis, of studies differentiating cancerous (high-grade dysplasia or invasive carcinoma) versus precancerous (low- or intermediate-grade dysplasia) provided six tests in which meta-analysis was performed. The sensitivities and specificities for diagnosing cancer (high-grade dysplasia or invasive carcinoma) were: CT: sensitivity 0.87 (95% CI 0.00 to 1.00), specificity 0.96 (95% CI 0.00 to 1.00); EUS: sensitivity 0.86 (95% CI 0.74 to 0.92), specificity 0.91 (95% CI 0.83 to 0.96); EUS-FNA: sensitivity 0.47 (95% CI 0.24 to 0.70), specificity 0.91 (95% CI 0.32 to 1.00); EUS-FNA carcinoembryonic antigen 200 ng/mL: sensitivity 0.58 (95% CI 0.28 to 0.83), specificity 0.51 (95% CI 0.19 to 0.81); MRI: sensitivity 0.69 (95% CI 0.44 to 0.86), specificity 0.93 (95% CI 0.43 to 1.00); PET: sensitivity 0.90 (95% CI 0.79 to 0.96), specificity 0.94 (95% CI 0.81 to 0.99). The sixth analysis, of studies differentiating cancerous (invasive carcinoma) from precancerous (low-grade dysplasia) provided no tests in which meta-analysis was performed. The seventh analysis, of studies differentiating precancerous or cancerous (intermediate- or high-grade dysplasia or invasive carcinoma) from precancerous (low-grade dysplasia) provided two tests in which meta-analysis was performed. The sensitivity and specificity for diagnosing cancer were: CT: sensitivity 0.83 (95% CI 0.68 to 0.92), specificity 0.83 (95% CI 0.64 to 0.93) and MRI: sensitivity 0.80 (95% CI 0.58 to 0.92), specificity 0.81 (95% CI 0.53 to 0.95), respectively. The eighth analysis, of studies differentiating precancerous or cancerous (intermediate- or high-grade dysplasia or invasive carcinoma) from precancerous (low-grade dysplasia) or benign lesions provided no test in which meta-analysis was performed.There were no major alterations in the subgroup analysis of cystic pancreatic focal lesions (42 studies; 2086 participants). None of the included studies evaluated EUS elastography or sequential testing.

AUTHORS' CONCLUSIONS

We were unable to arrive at any firm conclusions because of the differences in the way that study authors classified focal pancreatic lesions into cancerous, precancerous, and benign lesions; the inclusion of few studies with wide confidence intervals for each comparison; poor methodological quality in the studies; and heterogeneity in the estimates within comparisons.

Two unusual variants of pancreatic neuroendocrine tumor and their potential pitfalls on fine-needle aspiration cytology

Endoscopic ultrasound-guided fine-needle aspiration is increasingly utilized for the diagnosis of pancreatic lesions. Although operator dependent, the procedure has good overall performance characteristics and is minimally invasive; however, accuracy and sensitivity are reportedly lower for pancreatic neuroendocrine tumor (PanNET) compared with the more common pancreatic ductal adenocarcinoma (pACA). The underperformance is further exacerbated by the unusual cases of PanNET presenting with variant cytomorphology. We report two separate diagnostically challenging cases: a pigmented PanNET and a clear cell PanNET. We briefly review the literature and emphasize the importance of recognizing these uncommon variants when encountered in aspirate material. Diagn. Cytopathol. 2017;45:371-378. © 2016 Wiley Periodicals, Inc.