Computer-assisted analysis of lymph nodes detected by EUS in patients with esophageal carcinoma

A new descriptor for computer-aided diagnosis of EUS imaging to distinguish autoimmune pancreatitis from chronic pancreatitis

BACKGROUND AND AIMS

Computer-aided diagnosis of EUS images was quite useful in differentiating pancreatic cancer from normal tissue and chronic pancreatitis. This study investigated the feasibility of using computer-aided diagnostic techniques to extract EUS image parameters to distinguish autoimmune pancreatitis from chronic pancreatitis.

METHODS

A new descriptor, local ternary pattern variance, was introduced to improve the performance of the classification model. Patients with autoimmune pancreatitis (n = 81) or chronic pancreatitis (n = 100) were recruited for this study. Representative EUS images were selected, and 115 parameters from 10 categories were extracted from the region of interest. Distance-between-class and sequential forward selection algorithms were used for their ideal combination of features that allowed a support vector machine predictive model to be built, trained, and validated. The accuracy, sensitivity, specificity, positive predictive values (PPVs), and negative predictive values (NPVs) were used to evaluate the performance of experimental results.

RESULTS

Fourteen parameters from 3 categories were selected as an ideal combination of features. The sample set was randomly divided into a training set and a testing set by using two different algorithms-the leave-one-out algorithm and the half-and-half method. The half-and-half method yielded an average (± standard deviation) accuracy of 89.3 ± 2.7%, sensitivity of 84.1 ± 6.4%, specificity of 92.5 ± 3.3%, PPV of 91.6 ± 3.7%, and NPV of autoimmune pancreatitis of 88.0 ± 4.1%.

CONCLUSIONS

This study shows that, with the local ternary pattern variance textural feature, computer-aided diagnosis of EUS imaging may be valuable to differentiate autoimmune pancreatitis from chronic pancreatitis. Further refinement of such models could generate tools for the clinical diagnosis of autoimmune pancreatitis.

Computerized quantification of ultrasonic heterogeneity in thyroid nodules

To test whether computerized quantification of ultrasonic heterogeneity can be of help in the diagnosis of thyroid malignancy, we evaluated ultrasonic heterogeneity with an objective and quantitative computerized method in a prospective setting. A total of 400 nodules including 271 benign thyroid nodules and 129 malignant thyroid nodules were evaluated. Patient clinical data were collected, and the grading of heterogeneity on conventional gray-scale ultrasound images was retrospectively reviewed by a thyroid specialist. Quantification of ultrasonic heterogeneity (heterogeneity index, HI) was performed by a proprietary program implemented with methods proposed in this article. HI values differed significantly between benign and malignant nodules, diagnosed by a combination of fine-needle aspiration and surgical pathology results (p < 0.001, area under the curve = 0.714). The ultrasonic heterogeneity of these samples, as assessed by an experienced clinician, could not significantly differentiate between benign and malignant thyroid nodules. However, nodules with marked ultrasonic heterogeneity had higher HI values than nodules with homogeneous nodules. These results indicate that the new computer-aided diagnosis method for evaluation of the ultrasonic heterogeneity of thyroid nodules is an objective and quantitative method that is correlated with conventional ultrasonic heterogeneity assessment, but can better aid in the diagnosis of thyroid malignancy.

Quantification of heterogeneity as a biomarker in tumor imaging: a systematic review

BACKGROUND

Many techniques are proposed for the quantification of tumor heterogeneity as an imaging biomarker for differentiation between tumor types, tumor grading, response monitoring and outcome prediction. However, in clinical practice these methods are barely used. This study evaluates the reported performance of the described methods and identifies barriers to their implementation in clinical practice.

METHODOLOGY

The Ovid, Embase, and Cochrane Central databases were searched up to 20 September 2013. Heterogeneity analysis methods were classified into four categories, i.e., non-spatial methods (NSM), spatial grey level methods (SGLM), fractal analysis (FA) methods, and filters and transforms (F&T). The performance of the different methods was compared.

PRINCIPAL FINDINGS

Of the 7351 potentially relevant publications, 209 were included. Of these studies, 58% reported the use of NSM, 49% SGLM, 10% FA, and 28% F&T. Differentiation between tumor types, tumor grading and/or outcome prediction was the goal in 87% of the studies. Overall, the reported area under the curve (AUC) ranged from 0.5 to 1 (median 0.87). No relation was found between the performance and the quantification methods used, or between the performance and the imaging modality. A negative correlation was found between the tumor-feature ratio and the AUC, which is presumably caused by overfitting in small datasets. Cross-validation was reported in 63% of the classification studies. Retrospective analyses were conducted in 57% of the studies without a clear description.

CONCLUSIONS

In a research setting, heterogeneity quantification methods can differentiate between tumor types, grade tumors, and predict outcome and monitor treatment effects. To translate these methods to clinical practice, more prospective studies are required that use external datasets for validation: these datasets should be made available to the community to facilitate the development of new and improved methods.

Digital image analysis is a useful adjunct to endoscopic ultrasonographic diagnosis of subepithelial lesions of the gastrointestinal tract

OBJECTIVE

The purpose of this study was to explore the role of digital image analysis in differentiating endoscopic ultrasonographic (EUS) features of potentially malignant gastrointestinal subepithelial lesions (SELs) from those of benign lesions.

METHODS

Forty-six patients with histopathologically confirmed gastrointestinal stromal tumors (GISTs), carcinoids, and lipomas who had undergone EUS evaluation were identified from our database. Representative regions of interest (ROIs) were selected from the EUS images, and features were extracted by texture analysis. On the basis of these features, an artificial neural network (ANN) was built, trained, and internally validated by unsupervised learning followed by supervised learning. Outcomes were the performance characteristics of the ANN.

RESULTS

A total of 106, 111, and 124 ROIs were selected from EUS images of 8, 10, and 28 patients with lipomas, carcinoids, and GISTs, respectively. For each ROI, 228 statistical parameters were extracted and later reduced to the 11 most informative features by principal component analysis. After training with 50% of the data, the remainder of the data were used to validate the ANN. The model was "good" in differentiating carcinoids and GISTs, with area under the receiver operating characteristic curve (AUC) values of 0.86 and 0.89, respectively. The model was "excellent" in identifying lipomas correctly, with an AUC of 0.92.

CONCLUSIONS

Digital image analysis of EUS images is a useful noninvasive adjunct to EUS evaluation of SELs.

In vivo characterization of pancreatic and lymph node tissue by using EUS spectrum analysis: a validation study

BACKGROUND

Quantitative spectral analysis of the radiofrequency (RF) signals that underlie grayscale EUS images can be used to provide additional, objective information about tissue state.

OBJECTIVE

Our purpose was to validate RF spectral analysis as a method to distinguish between (1) benign and malignant lymph nodes and (2) normal pancreas, chronic pancreatitis, and pancreatic cancer.

DESIGN AND SETTING

A prospective validation study of eligible patients was conducted to compare with pilot study RF data.

PATIENTS

Forty-three patients underwent EUS of the esophagus, stomach, pancreas, and surrounding intra-abdominal and mediastinal lymph nodes (19 from a previous pilot study and 24 additional patients).

MAIN OUTCOME MEASUREMENTS

Midband fit, slope, intercept, and correlation coefficient from a linear regression of the calibrated RF power spectra were determined.

RESULTS

Discriminant analysis of mean pilot-study parameters was then performed to classify validation-study parameters. For benign versus malignant lymph nodes, midband fit and intercept (both with t test P < .058) provided classification with 67% accuracy and area under the receiver operating curve (AUC) of 0.86. For diseased versus normal pancreas, midband fit and correlation coefficient (both with analysis of variance P < .001) provided 93% accuracy and an AUC of 0.98. For pancreatic cancer versus chronic pancreatitis, the same parameters provided 77% accuracy and an AUC of 0.89. Results improved further when classification was performed with all data.

LIMITATIONS

Moderate sample size and spatial averaging inherent to the technique.

CONCLUSIONS

This study confirms that mean spectral parameters provide a noninvasive method to quantitatively discriminate benign and malignant lymph nodes as well as normal and diseased pancreas.

Digital image analysis of EUS images accurately differentiates pancreatic cancer from chronic pancreatitis and normal tissue

BACKGROUND

Concomitant changes of chronic pancreatitis markedly degrade the performance of EUS in diagnosing pancreatic adenocarcinoma (PC). Digital image analysis (DIA) of the spatial distribution of pixels in a US image has been used as an effective approach to tissue characterization.

OBJECTIVE

We applied the techniques of DIA to EUS images of the pancreas to develop a classification model capable of differentiating pancreatic adenocarcinoma from non-neoplastic tissue.

DESIGN

Representative regions of interest were digitally selected in EUS images of 3 groups of patients with normal pancreas (group I), chronic pancreatitis (group II), and pancreatic adenocarcinoma (group III). Texture analyses were then performed by using image analysis software. Principal component analysis (PCA) was used for data reduction, and, later, a neural-network-based predictive model was built, trained, and validated.

SETTING

Tertiary academic medical center.

PATIENTS

Patients undergoing EUS of the pancreas.

RESULTS

A total of 110, 99, and 110 regions of interest in groups I, II, III, respectively, were available for analysis. For each region, a total of 256 statistical parameters were extracted. Eleven parameters were subsequently retained by PCA. A neural network model was built, trained by using these parameters as input variables for prediction of PC, and then validated in the remainder of the data set. This model was very accurate in classifying PC with an area under the receiver operating characteristic curve of 0.93.

LIMITATION

Exploratory study with a small number of patients.

CONCLUSIONS

DIA of EUS images is accurate in differentiating PC from chronic inflammation and normal tissue. With the potential availability of real-time application, DIA can develop into a useful clinical diagnostic tool in pancreatic diseases and in certain situations may obviate EUS-guided FNA.

Contrast-enhanced endoscopic ultrasound in discrimination between benign and malignant mediastinal and abdominal lymph nodes

BACKGROUND

Enlarged lymph nodes in the mediastinum reflect neoplastic, infectious or other diseases. The classification of these nodes is crucial in the management of the patient. Currently, only invasive measures obtaining tissue samples reach satisfying specificity. Contrast-enhanced endoscopic ultrasound (EUS) may offer a non-invasive alternative.

MATERIALS AND METHODS

A total of 122 patients (age: 63 +/- 15 years, 92 males, 30 females) with enlarged mediastinal and/or paraaortic lymph nodes diagnosed by CT scan were included in the study. EUS-guided fine needle aspiration was performed and cytologic specimens were diagnosed as representing a malignant or benign process in case of Papanicolau IV and V, or Papanicolau I and II, respectively.

RESULTS

Based on cytology results, the investigated lymph nodes were classified as neoplastic (n = 48) or non-neoplastic lymph nodes. Using the B-mode criteria the preliminary diagnosis was confirmed in 64 out of 74 benign lymph nodes (specificity 86%). Regarding malignant lymph nodes 33 of 48 were confirmed (sensitivity 68%). Using the advanced contrast-enhanced EUS criteria the diagnosis was confirmed in 68 of 74 benign lymph nodes (specificity 91%). However, in case of malignant lymph nodes the number of correct diagnoses dropped to 29 of 48 lymph nodes (sensitivity 60%). The contrast-enhanced EUS criteria to identify benign lymph nodes and node enlargement in malignant lymphoma do not differ. If those ten patients with malignant lymphoma are excluded, the sensitivity of the contrast enhanced EUS for malignant lymph nodes rises to 73%.

CONCLUSION

Contrast-enhanced EUS improves the specificity in diagnosing benign lymph nodes as compared to B-mode EUS. It does not improve the correct identification of malignant lymph nodes and cannot replace EUS-guided fine-needle aspiration.

Preoperative TN staging of esophageal cancer: Comparison of miniprobe ultrasonography, spiral CT and MRI

AIM: To evaluate the value of miniprobe sonography (MPS), spiral CT and MR imaging (MRI) in the tumor and regional lymph node staging of esophageal cancer.

METHODS: Eight-six patients (56 men and 30 women; age range of 39-73 years, mean 62 years) with esophageal carcinoma were staged preoperatively with imaging modalities. Of them, 81 (94%) had squamous cell carcinoma, 4 (5%) adenocarcinoma, and 1 (1%) adenoacanthoma. Eleven patients (12%) had malignancy of the upper one third, 41 (48%) of the mid-esophagus and 34 (40%) of the distal one third. Forty-one were examined by spiral CT in whom 13 were co-examined by MPS, and forty-five by MRI in whom 18 were also co-examined by MPS. These imaging results were compared with the findings of the histopathologic examination for resected specimens.

RESULTS: In staging the depth of tumor growth, MPS was significantly more accurate (84%) than spiral CT and MRI (68% and 60%, respectively, P < 0.05). The specificity and sensitivity were 82% and 85% for MPS; 60% and 69% for spiral CT; and 40% and 63% for MRI, respectively. In staging regional lymph nodes, spiral CT was more accurate (78%) than MPS and MRI (71% and 64%, respectively), but the difference was not statistically significant. The specificity and sensitivity were 79% and 77% for spiral CT; 75% and 68% for MPS; and 68% and 62% for MRI, respectively.

CONCLUSION: MPS is superior to spiral CT or MRI for T staging, especially in early esophageal cancer. However, the three modalities have the similar accuracy in N staging. Spiral CT or MRI is helpful for the detection of far-distance metastasis in esophageal cancer.