CT radiomics may predict the grade of pancreatic neuroendocrine tumors: a multicenter study

Pancreatic Cancer Imaging: A New Look at an Old Problem

Computed tomography is the most commonly used imaging modality to detect and stage pancreatic cancer. Previous advances in pancreatic cancer imaging have focused on optimizing image acquisition parameters and reporting standards. However, current state-of-the-art imaging approaches still misdiagnose some potentially curable pancreatic cancers and do not provide prognostic information or inform optimal management strategies beyond stage. Several recent developments in pancreatic cancer imaging, including artificial intelligence and advanced visualization techniques, are rapidly changing the field. The purpose of this article is to review how these recent advances have the potential to revolutionize pancreatic cancer imaging.

Computed Tomography-Based Radiomics Signature for the Preoperative Differentiation of Pancreatic Adenosquamous Carcinoma From Pancreatic Ductal Adenocarcinoma

PURPOSE

The purpose was to assess the predictive ability of computed tomography (CT)-based radiomics signature in differential diagnosis between pancreatic adenosquamous carcinoma (PASC) and pancreatic ductal adenocarcinoma (PDAC).

MATERIALS AND METHODS

Eighty-one patients (63.6 ± 8.8 years old) with PDAC and 31 patients (64.7 ± 11.1 years old) with PASC who underwent preoperative CE-CT were included. A total of 792 radiomics features were extracted from the late arterial phase (n = 396) and portal venous phase (n = 396) for each case. Significantly different features were selected using Mann-Whitney U test, univariate logistic regression analysis, and minimum redundancy and maximum relevance method. A radiomics signature was constructed using random forest method, the robustness and the reliability of which was validated using 10-times leave group out cross-validation (LGOCV) method.

RESULTS

Seven radiomics features from late arterial phase images and three from portal venous phase images were finally selected. The radiomics signature performed well in differential diagnosis between PASC and PDAC, with 94.5% accuracy, 98.3% sensitivity, 90.1% specificity, 91.9% positive predictive value (PPV), and 97.8% negative predictive value (NPV). Moreover, the radiomics signature was proved to be robust and reliable using the LGOCV method, with 76.4% accuracy, 91.1% sensitivity, 70.8% specificity, 56.7% PPV, and 96.2% NPV.

CONCLUSION

CT-based radiomics signature may serve as a promising non-invasive method in differential diagnosis between PASC and PDAC.

Radiologists with MRI-based radiomics aids to predict the pelvic lymph node metastasis in endometrial cancer: a multicenter study

OBJECTIVE

To construct a MRI radiomics model and help radiologists to improve the assessments of pelvic lymph node metastasis (PLNM) in endometrial cancer (EC) preoperatively.

METHODS

During January 2014 and May 2019, 622 EC patients (age 56.6 ± 8.8 years; range 27-85 years) from five different centers (A to E) were divided into training set, validation set 1 (351 cases from center A), and validation set 2 (271 cases from centers B-E). The radiomics features were extracted basing on T2WI, DWI, ADC, and CE-T1WI images, and most related radiomics features were selected using the random forest classifier to build a radiomics model. The ROC curve was used to evaluate the performance of training set and validation sets, radiologists based on MRI findings alone, and with the aid of the radiomics model. The clinical decisive curve (CDC), net reclassification index (NRI), and total integrated discrimination index (IDI) were used to assess the clinical benefit of using the radiomics model.

RESULTS

The AUC values were 0.935 for the training set, 0.909 and 0.885 for validation sets 1 and 2, 0.623 and 0.643 for the radiologists 1 and 2 alone, and 0.814 and 0.842 for the radiomics-aided radiologists 1 and 2, respectively. The AUC, CDC, NRI, and IDI showed higher diagnostic performance and clinical net benefits for the radiomics-aided radiologists than for the radiologists alone.

CONCLUSIONS

The MRI-based radiomics model could be used to assess the status of pelvic lymph node and help radiologists improve their performance in predicting PLNM in EC.

KEY POINTS

• A total of 358 radiomics features were extracted. The 37 most important features were selected using the random forest classifier. • The reclassification measures of discrimination confirmed that the radiomics-aided radiologists performed better than the radiologists alone, with an NRI of 1.26 and an IDI of 0.21 for radiologist 1 and an NRI of 1.37 and an IDI of 0.24 for radiologist 2.

Anatomic and Functional Imaging of Neuroendocrine Tumors

OPINION STATEMENT

Neuroendocrine tumors (NETs) can occur in a wide variety of organs and display a spectrum of pathologic behavior. Accurate and effective imaging is paramount to the diagnosis, staging, therapy, and surveillance of patients with NET. There have been continuous advancements in the imaging of NET which includes anatomic and functional techniques.

CT-Based Radiomics Score for Distinguishing Between Grade 1 and Grade 2 Nonfunctioning Pancreatic Neuroendocrine Tumors

OBJECTIVE. The objective of our study was to explore the relationship between a CT-based radiomics score and grade of nonfunctioning pancreatic neuroendocrine tumors (PNETs) and to evaluate the ability of a calculated CT radiomics score to distinguish between grade 1 and grade 2 nonfunctioning PNETs. MATERIALS AND METHODS. This retrospective study assessed 102 patients with surgically resected, pathologically confirmed nonfunctioning PNETs who underwent MDCT from January 2014 to December 2017. Radiomic methods were used to extract features from portal venous phase CT scans, and the least absolute shrinkage and selection operator (LASSO) method was used to select the features. Multivariate logistic regression models were used to analyze the association between the CT radiomics score and nonfunctioning PNET grades. The performance of the CT radiomics score was assessed on the basis of its discriminative ability and clinical usefulness. RESULTS. The CT radiomics score, which consisted of four selected features, was significantly associated with nonfunctioning PNET grades. Every 1-point increase in radiomics score was associated with a 57% increased risk of grade 2 disease. The score also showed high accuracy (AUC = 0.86 for all PNETs; AUC = 0.81 for PNETs ≤ 2 cm). The best cutoff point for maximal sensitivity and specificity was a CT radiomics score of -0.134. Decision curve analysis showed that the CT radiomics score is clinically useful. CONCLUSION. The CT radiomics score shows a significant association with the grade of nonfunctioning PNETs and provides a potentially valuable noninvasive tool for distinguishing between different grades of nonfunctioning PNET, especially among patients with tumors 2 cm or smaller.

The combination of hepatobiliary phase with Gd-EOB-DTPA and DWI is highly accurate for the detection and characterization of liver metastases from neuroendocrine tumor

OBJECTIVES

To compare the diagnostic accuracy of dynamic contrast-enhanced phases, hepatobiliary phase (HBP), and diffusion-weighted imaging (DWI) for the detection of liver metastases from neuroendocrine tumor (NET).

METHODS

Sixty-seven patients with suspected NET liver metastases underwent gadoxetic acid-enhanced MRI. Three radiologists read four imaging sets separately and independently: DWI, T2W+dynamic, T2WI+HBP, and DWI+HBP. Reference standard included all imaging, histological findings, and clinical data. Sensitivity and specificity were calculated and compared for each imaging set. Interreader agreement was evaluated by intraclass correlation coefficient (ICC). Univariate logistic regression was performed to evaluate lesion characteristics (size, ADC, and enhancing pattern) associated to false positive and negative lesions.

RESULTS

Six hundred twenty-five lesions (545 metastases, 80 benign lesions) were identified. Detection rate was significantly higher combining DWI+HBP than the other imaging sets (sensitivity 86% (95% confidence interval (CI) 0.845-0.878), specificity 94% (95% CI 0.901-0.961)). The sensitivity and specificity of the other sets were 82% and 65% for DWI, 88% and 69% for T2WI, and 90% and 82% for HBP+T2WI, respectively. The interreader agreement was statistically higher for both HBP sets (ICC = 0.96 (95% CI 0.94-0.97) for T2WI+HBP and ICC = 0.91 (95% CI 0.87-0.94) for DWI+HBP, respectively) compared with that for DWI (ICC = 0.76 (95% CI 0.66-0.83)) and T2+dynamic (ICC = 0.85 (95% CI 0.79-0.9)). High ADC values, large lesion size, and hypervascular pattern lowered the risk of false negative.

CONCLUSION

Given the high diagnostic accuracy of combining DWI+HBP, gadoxetic acid-enhanced MRI is to be considered in NET patients with suspected liver metastases. Fast MRI protocol using T2WI, DWI, and HBP is of interest in this population.

KEY POINTS

• The combined set of diffusion-weighted (DW) and hepatobiliary phase (HBP) images yields the highest sensitivity and specificity for neuroendocrine liver metastasis (NELM) detection. • Gadoxetic acid should be the contrast agent of choice for liver MRI in NET patients. • The combined set of HBP and DWI sequences could also be used as a tool of abbreviated MRI in follow-up or assessment of treatment such as somatostatin analogs.

Diagnostic accuracy of unenhanced CT texture analysis to differentiate mass-forming pancreatitis from pancreatic ductal adenocarcinoma

PURPOSE

To investigate the value of texture analysis on unenhanced computed tomography (CT) to potentially differentiate mass-forming pancreatitis (MFP) from pancreatic ductal adenocarcinoma (PDAC).

METHODS

A retrospective study consisting of 109 patients (30 MFP patients vs 79 PDAC patients) who underwent preoperative unenhanced CT between January 2012 and December 2017 was performed. Synthetic minority oversampling technique (SMOTE) algorithm was adopted to reconstruct and balance MFP and PDAC samples. A total of 396 radiomic features were extracted from unenhanced CT images. Mann-Whitney U test and minimum redundancy maximum relevance (MRMR) methods were used for the purpose of dimension reduction. Predictive models were constructed using random forest (RF) method, and were validated using leave group out cross-validation (LGOCV) method. Diagnostic performance of the predictive model, including sensitivity, specificity, accuracy, positive predicting value (PPV), and negative predicting value (NPV), was recorded.

RESULTS

We applied 200% of SMOTE to MFP and PDAC patients, resulting in 90 MFP patients compared with 120 PDAC patients. Dimension reduction steps yielded 30 radiomic features using Mann-Whitney U test and MRMR methods. Ten radiomic features were retained using RF method. Four most predictive parameters, including GreyLevelNonuniformity_angle90_offset1, VoxelValueSum, HaraVariance, and ClusterProminence_AllDirection_offset1_SD, were used to generate the predictive model with preferable 92.2% sensitivity, 94.2% specificity, 93.3% accuracy, 92.2% PPV, and 94.2% NPV. Finally, in LGOCV analysis, a high pooled mean sensitivity, specificity, and accuracy (82.6%, 80.8%, and 82.1%, respectively) indicate a relatively reliable and stable predictive model.

CONCLUSIONS

Unenhanced CT texture analysis can be a promising noninvasive method in discriminating MFP from PDAC.

Comparison between neuroendocrine carcinomas and well-differentiated neuroendocrine tumors of the pancreas using dynamic enhanced CT

OBJECTIVES

To identify CT features distinguishing neuroendocrine carcinomas (NECs) of pancreas from well-differentiated neuroendocrine tumors (NETs) according to the World Health Organization 2017 and 2019 classification systems.

METHODS

This retrospective study included 69 patients with pathologically confirmed pancreatic neuroendocrine neoplasms who underwent dynamic CT (17, 17, 18, and 17 patients for well-differentiated grade 1, 2, 3 NET and NEC, respectively). CT was used to perform qualitative analysis (component, homogeneity, calcification, peripancreatic infiltration, main pancreatic ductal dilatation, bile duct dilatation, intraductal extension, and vascular invasion) and quantitative analysis (interface between tumor and parenchyma [delta], arterial enhancement ratio [AER], portal enhancement ratio [PER], and dynamic enhancement pattern). Uni- and multivariate logistic regression analyses were performed to identify features indicating NEC. Optimal cutoff values for enhancement ratios were determined.

RESULTS

NECs demonstrated significantly higher frequencies of main pancreatic ductal dilatation, bile duct dilatation, vascular invasion, and significantly lower delta (i.e., lower conspicuity), AER, and PER than well-differentiated NET (p < 0.05). On multivariate analysis, PER was the only independent factor selected by the model for differentiation of NEC from well-differentiated NET (odds ratio, < 0.001; 95% confidence interval [CI], < 0.001-0.012). PER < 0.8 showed the sensitivity of 94.1% (95% CI, 71.3-99.9) and the specificity of 88.5% (95% CI, 76.6-95.6). When three significant CT features were combined, the sensitivity and specificity for diagnosing NEC were 88.2% and 88.5%, respectively.

CONCLUSIONS

Tumor-parenchyma enhancement ratio in portal phase is a useful CT feature to distinguish NECs from well-differentiated NETs. Combining qualitative and quantitative CT features may aid in achieving good diagnostic accuracy in the differentiation between NEC and well-differentiated NET.

KEY POINTS

• Neuroendocrine carcinoma of the pancreas should be distinguished from well-differentiated neuroendocrine tumor in line with the revised grading and staging system. • Neuroendocrine carcinoma of the pancreas can be differentiated from well-differentiated neuroendocrine tumor on dynamic CT based on assessment of the portal enhancement ratio, arterial enhancement ratio, tumor conspicuity, dilatation of the main pancreatic duct or bile duct, and vascular invasion. • Tumor-parenchyma enhancement ratio in portal phase of dynamic CT is a useful feature, which may help to distinguish neuroendocrine carcinoma from well-differentiated neuroendocrine tumor of the pancreas.

Noncontrast Radiomics Approach for Predicting Grades of Nonfunctional Pancreatic Neuroendocrine Tumors

BACKGROUND

Endoscopic ultrasound-guided fine-needle aspiration is associated with the accurate determination of tumor grade. However, because it is an invasive procedure there is a need to explore alternative noninvasive procedures.

PURPOSE

To develop and validate a noncontrast radiomics model for the preoperative prediction of nonfunctional pancreatic neuroendocrine tumor (NF-pNET) grade (G).

STUDY TYPE

Retrospective, single-center study.

SUBJECTS

Patients with pathologically confirmed PNETs (139) were included.

FIELD STRENGTH/SEQUENCE

3T/breath-hold single-shot fast-spin echo T₂ -weighted sequence and unenhanced and dynamic contrast-enhanced T₁ -weighted fat-suppressed sequences.

ASSESSMENT

Tumor features on contrast MR images were evaluated by three board-certified abdominal radiologists.

STATISTICAL TESTS

Multivariable logistic regression analysis was used to develop the clinical model. The least absolute shrinkage and selection operator method and linear discriminative analysis (LDA) were used to select the features and to construct a radiomics model. The performance of the models was assessed using the training cohort (97 patients) and the validation cohort (42 patients), and decision curve analysis (DCA) was applied for clinical use.

RESULTS

The clinical model included 14 imaging features, and the corresponding area under the curve (AUC) was 0.769 (95% confidence interval [CI], 0.675-0.863) in the training cohort and 0.729 (95% CI, 0.568-0.890) in the validation cohort. The LDA included 14 selected radiomics features that showed good discrimination-in the training cohort (AUC, 0.851; 95% CI, 0.758-0.916) and the validation cohort (AUC, 0.736; 95% CI, 0.518-0.874). In the decision curves, if the threshold probability was 0.17-0.84, using the radiomics score to distinguish NF-pNET G1 and G2/3, offered more benefit than did the use of a treat-all-patients or treat-none scheme.

DATA CONCLUSION

The developed radiomics model using noncontrast MRI could help differentiate G1 and G2/3 tumors, to make the clinical decision, and screen pNETs grade.

LEVEL OF EVIDENCE

4 TECHNICAL EFFICACY STAGE: 2 J. Magn. Reson. Imaging 2020;52:1124-1136.

Computed Tomography-Based Tumor Heterogeneity Analysis Reveals Differences in a Cohort with Advanced Pancreatic Carcinoma under Palliative Chemotherapy

Purpose

Imaging in pancreatic cancer is a challenge, especially regarding therapy response evaluation. Tumor size, attenuation, and perfusion are widely used as parameters for computed tomography (CT) examinations, but are often limited due to blurry tumor borders and missing qualitative parameters. To improve monitoring of therapy response, we tested a new CT-based approach of tumor heterogeneity feature analysis.

Methods

A total of 13 patients with pancreatic adenocarcinoma undergoing abdominal CT according to standard as baseline imaging with clinical follow-up and imaging (median time span 64 days) under systematic therapy (FOLFIRINOX/gemcitabine) were retrospectively analyzed. Progression was defined as new lesions and local tumor spread. Tumor heterogeneity analysis was performed using mintLesion®. Seven different image features referring to image heterogeneity were analyzed. Statistical analysis was performed with Spearman's rank correlation and Mann-Whitney U test.

Results

During follow-up, tumor volume did not significantly change between our groups with overall progression (local and systemic) and progression-free patients (p = 0.661). Mean positivity of pixel values were significantly higher in patients without progression compared to patients with progression (p = 0.030). There was a significant negative correlation between changes in kurtosis and time to local tumor spread (p = 0.008) or systemic progression (p = 0.017).

Conclusions

Results suggest that analysis of tumor heterogeneity might provide valuable information from routine-acquired images regarding therapy response evaluation. This might help adjusting therapy regimes and could be easily integrated in clinical workflows. Furthermore, this procedure might possibly predict therapy response and, hence could lead the way to find a potential marker for progression-free survival.

Detection of insulinoma: one-stop pancreatic perfusion CT with calculated mean temporal images can replace the combination of bi-phasic plus perfusion scan

OBJECTIVE

To evaluate the feasibility of one-stop pancreatic perfusion CT with mean temporal (MT) imaging replacing the combination of a bi-phasic scan plus a perfusion scan to detect insulinoma.

MATERIAL AND METHODS

Forty-five patients with suspected insulinoma, who underwent both biphasic and perfusion CT, were enrolled in this retrospective study. MT datasets including images for different delineation purposes were generated by averaging 3 dynamic datasets from perfusion CT, which are MTA for arterial, MTPV for portal vein and MTO for lesions. Two readers assessed the image quality and diagnostic performance separately for biphasic and MT datasets. Radiation doses were also assessed. Paired t tests, Wilcoxon signed-rank tests and McNemar's tests were applied for comparison.

RESULTS

Compared with bi-phasic CT images, image noise, SNR and CNR of the MTA and MTPV datasets were all non-inferior (noise and CNR of the portal vein, p = 0.565 and p = 0.227, respectively) or superior (p ≤ 0.001). The subjective image quality was better in the MTA and MTPV images (p < 0.001 to p = 0.004). The sensitivity and NPV of MT images were also better (95% vs 75% and 75% vs 37.5% for reader 1; 97.5% vs 72.5% and 85.7% vs 35.3% for reader 2). Omitting the bi-phasic scan resulted in a dose reduction of 25% ± 4%.

CONCLUSION

MT imaging can allow pancreatic perfusion CT to be used alone without the need for an additional bi-phasic CT in the detection of insulinoma.

KEY POINTS

• Mean temporal images reconstructed from perfusion CT with an averaging technique reproduce usual bi-phasic images (arterial and portal phases). • The image quality of mean temporal images is non-inferior or superior to native bi-phasic CT. The sensitivity and NPV for the diagnosis of insulinoma are better for mean temporal images than for traditional bi-phasic CT. • Mean temporal imaging can allow pancreatic perfusion CT to be used alone without the need for an additional bi-phasic CT in the detection of insulinoma. Radiation dose saving is important.

Performance of CT-based radiomics in diagnosis of superior mesenteric vein resection margin in patients with pancreatic head cancer

OBJECTIVES

To accurately identify the relationship between a portal radiomics score (rad-score) and pathologic superior mesenteric vein (SMV) resection margin and to evaluate the diagnostic performance in patients with pancreatic head cancer.

MATERIALS AND METHODS

A total of 181 patients with postoperatively and pathologically confirmed pancreatic head cancer who underwent multislice computed tomography within one month of resection between January 2016 and December 2018 were retrospectively investigated. For each patient, 1029 radiomics features of the portal phase were extracted, which were reduced using the least absolute shrinkage and selection operator (LASSO) logistic regression algorithm. Multivariate logistic regression models were used to analyze the association between the portal rad-score and SMV resection margin.

RESULTS

Patients with negative (R0) and positive (R1) margins accounted for 70.17% (127) and 29.83% (54) of the cohort, respectively. The rad-score was significantly associated with the SMV resection margin status (p < 0.05). Multivariate analyses confirmed a significant and independent association between the portal rad-score and SMV resection margin (OR 4.62; 95% CI 2.19-9.76; p < 0.0001). The portal rad-score had high accuracy (area under the curve = 0.750). The best cut point based on maximizing the sum of sensitivity and specificity was - 0.741 (sensitivity = 64.8%; specificity = 74.0%; accuracy = 71.3%). Decision curve analysis indicated the clinical usefulness of radiomics score.

CONCLUSIONS

The portal rad-score is significantly associated with the pathologic SMV resection margin, and it can accurately and noninvasively predict the SMV resection margin in patients with pancreatic cancer.

Evaluation of Texture Analysis for the Differential Diagnosis of Mass-Forming Pancreatitis From Pancreatic Ductal Adenocarcinoma on Contrast-Enhanced CT Images

Purpose: To investigate the potential of computed tomography (CT) imaging features and texture analysis to differentiate between mass-forming pancreatitis (MFP) and pancreatic ductal adenocarcinoma (PDAC). Materials and Methods: Thirty patients with pathologically proved MFP and 79 patients with PDAC were included in this study. Clinical data and CT imaging features of the two lesions were evaluated. Texture features were extracted from arterial and portal phase CT images using commercially available software (AnalysisKit). Multivariate logistic regression analyses were used to identify relevant CT imaging and texture parameters to discriminate MFP from PDAC. Receiver operating characteristic curves were performed to determine the diagnostic performance of predictions. Results: MFP showed a larger size compared to PDAC (p = 0.009). Cystic degeneration, pancreatic ductal dilatation, vascular invasion, and pancreatic sinistral portal hypertension were more frequent and duct penetrating sign was less frequent in PDAC compared to MFP. Arterial CT attenuation, arterial, and portal enhancement ratios of MFP were higher than PDAC (p < 0.05). In multivariate analysis, arterial CT attenuation and pancreatic duct penetrating sign were independent predictors. Texture features in arterial phase including SurfaceArea, Percentile40, InverseDifferenceMoment_angle90_offset4, LongRunEmphasis_angle45_offset4, and uniformity were independent predictors. Texture features in portal phase including LongRunEmphasis_angle135_offset7, VoxelValueSum, LongRunEmphasis_angle135_offset4, and GLCMEntropy_angle45_offset1 were independent predictors. Areas under the curve of imaging feature-based, texture feature-based in arterial and portal phases, and the combined models were 0.84, 0.96, 0.93, and 0.98, respectively. Conclusions: CT texture analysis demonstrates great potential to differentiate MFP from PDAC.