2026

DOI: 10.1016/j.ctarc.2025.101080

This study demonstrates that an early decrease in spleen volume between baseline and a 30-day follow-up is a significant predictor of improved progression-free survival (PFS) and overall survival (OS) in patients with relapsed or refractory non-Hodgkin lymphoma (NHL) - including LBCL, MCL, and FL - treated with CD19-specific CAR T-cell therapy. Researchers utilized the Lugano criteria for response assessment and identified spleen volume change as a potential novel dynamic biomarker.

mint Lesion was used to perform the imaging analyses, specifically for the 3D segmentation of the spleen and the evaluation of up to six target lesions, and to categorize treatment response in line with the Lugano criteria.

DOI: 10.1158/1078-0432.CCR-25-2274

In this study, researchers demonstrated that the absence of detectable circulating tumor DNA (ctDNA) at baseline is a strong predictor of improved survival for patients with metastatic uveal melanoma. The quantitative level of ctDNA also proved critical, as a mutant allele fraction (MAF) greater than 5% at any point correlated with significantly poorer outcomes compared to lower levels. Ultimately, the findings suggest that both the presence of ctDNA before treatment and its persistence within the first three months of therapy serve as powerful noninvasive markers for monitoring disease progression.

In this study, researchers found that the number of pulmonary lesions (specifically ≥ 5) on baseline CT scans independently predicted shorter overall survival for patients with soft tissue and bone sarcoma. Conversely, measurements of tumor burden showed no significant prognostic value for survival or disease progression. Ultimately, while lesion count serves as a predictor for overall survival, no specific imaging parameter was found to independently predict progression-free survival.

DOI: 10.1007/s00259-025-07550-2

This study evaluates the feasibility of using [18F]SiTATE PET/CT to determine eligibility and monitor treatment response in patients with metastatic pheochromocytoma and paraganglioma (PPGL) undergoing peptide receptor radionuclide therapy (PRRT). Researchers confirmed PRRT eligibility based on high somatostatin receptor expression using the Krenning score, finding that while all patients achieved stable disease per RECIST 1.1, follow-up assessments showed heterogeneous results between metabolic, morphologic and biochemical measures.

mint Lesion was utilized to evaluate diagnostic CT scans according to RECIST 1.1 to provide a morphologic assessment of treatment response.

DOI: 10.1016/j.ejrad.2026.112674

This study evaluates the accuracy of the nn-UNet TotalSegmentator AI model for automatic liver segmentation in abdominal CT scans, specifically comparing healthy livers against atypical cases involving pathologies such as polycystic liver disease (PLD), cirrhosis with ascites, metastases, and hepatomegaly. While the model performed excellently on healthy livers with high Dice scores, clinical evaluations revealed significant limitations in complex pathological cases, particularly PLD.

mint Lesion was utilized as the data management platform to pseudonymize and transfer imaging data from the local PACS to the research infrastructure. It also functioned as a structured reporting platform that served as the primary stage in the workflow before data was moved to specialized tools for deep learning analysis.

DOI: 10.1016/j.ejrad.2026.112659

This study compares the performance of TGSE-BLADE DWI and RESOLVE DWI for identifying perioperative ischemic changes in patients following the resection of intracranial tumors, such as glioblastoma, pilocytic astrocytoma, and metastases. The findings demonstrate that TGSE-BLADE DWI significantly reduces susceptibility artifacts and geometric distortions caused by intracranial air, resulting in superior image quality and higher diagnostic confidence compared to RESOLVE DWI.

mint Lesion was employed by neuroradiologists to perform bidimensional measurements of the resection defect. These measurements allowed the researchers to quantitatively assess the extent of geometric distortion across different MRI sequences compared to T1-weighted reference images.

DOI: link.springer.com/article/10.1007/s00330-025-11806-8

This study established 1.5-T MRI T2-relaxation reference values for knee cartilage across 929 volunteers, demonstrating that T2-values are significantly higher in superficial layers and increase with age and female sex. The research also found that pathological cartilage morphology, defined by the modified Noyes Score, is associated with elevated T2-values compared to structurally normal cartilage.

mint Lesion was utilized for cartilage T2-mapping analysis and morphological knee evaluation, including the assessment of the modified Noyes Grading. It also facilitated manual cartilage segmentation by readers to delineate regions of interest across the knee joint.

2025

DOI: 10.1186/s40644-025-00959-w

This study evaluates the prognostic value of maximum tumor dissemination (Dmax) in patients with relapsed or refractory large B-cell lymphoma (LBCL) and mantle-cell lymphoma (MCL) treated with CD19-specific CAR T-cell therapy. The research found that high baseline Dmax is a significant predictor of shorter progression-free survival (PFS) when assessed via Lugano criteria, though it showed no significant association with overall survival. The authors suggest that while Dmax is a useful imaging biomarker, future studies should explore its combination with radiomics and artificial intelligence to further improve risk stratification.

mint Lesion was the dedicated trial reporting software used to perform all structured imaging analyses for Lugano assessment at baseline and follow-up timepoints for patients with relapsed or refractory large B-cell lymphoma (LBCL) and mantle-cell lymphoma (MCL).

DOI: 10.1158/2767-9764.CRC-25-0507

This phase Ia/Ib multicenter study evaluated the safety, pharmacokinetics, and preliminary antitumor activity of BPM31510IV (a lipid nanodispersion of oxidized Coenzyme Q10) as monotherapy or in combination with chemotherapy in patients with advanced solid tumors. The trial found the treatment to be well-tolerated with evidence of a metabolic shift from glycolysis to oxidative phosphorylation.

mint Lesion was used for tumor response assessment in line with RECIST 1.1 and to automatically generate quantitative metabolic data from 18FDG-PET/CT scans, including standardized uptake values (SUVmax, SUVpeak, SUVmean), metabolic tumor volume (MTV), and total glycemic index (TGI). These metrics provided a whole-body assessment of metabolic tumor burden to measure longitudinal changes in response to treatment.

DOI: 10.21203/rs.3.rs-7334701/v1

This phase II trial (POLAR) evaluated the efficacy of maintenance pembrolizumab and olaparib in patients with metastatic pancreatic cancer (mPC) who achieved disease control on platinum-based chemotherapy. While the primary endpoint was not met in the core homologous recombination deficiency (HRD) cohort, the combination demonstrated clinical activity with a 6-month progression-free survival rate of 64% and an overall response rate of 35%.

mint Lesion was utilized for the independent radiology review to conduct standardized treatment response assessments according to RECIST 1.1 guidelines. The software ensured reproducible data capture for the trial’s primary and secondary efficacy endpoints.

DOI: 10.1007/s11307-025-02053-w 

This study evaluates the efficacy and safety of Lu-177-DOTATATE peptide receptor radionuclide therapy (PRRT) in seven patients with advanced radioiodine-refractory differentiated thyroid carcinoma (DTC). The results demonstrate that PRRT is a beneficial and well-tolerated treatment option, achieving renewed tumor control even in a "rechallenge" scenario where treatment was reinitiated after an interruption of over one year. No major side effects (CTCAE Grade 3–5) were observed, indicating its potential for patients with limited alternative therapies.

mint Lesion was used for the standardized evaluation of CT datasets to define and measure target/non-target lesions according to RECIST 1.1.

DOI: 10.3390/jcm14176111

This retrospective, multireader multicase study evaluated the diagnostic accuracy of artificial intelligence-assisted biparametric MRI (AI-bpMRI) compared to standard bpMRI and multiparametric MRI (mpMRI) for detecting prostate cancer (PCa). The study found that AI-bpMRI is non-inferior to mpMRI for detecting clinically significant prostate cancer (Gleason score ≥ 3+4) and superior to standard bpMRI for all PCa cases.

In this study, the Prostate.Carcinoma.ai AI algorithm was integrated into the mint Lesion platform to provide a standardized, computer-assisted reading environment for the PI-RADS 2.1 assessment. The software was used to automate the segmentation of the prostate gland and suspicious lesions, automatically calculate volumes and PSA density, and facilitate the generation of structured reports by mapping findings directly onto the prostate sector map.

DOI: 10.1186/s12880-025-02006-3

This study shows that 3D radiomics combined with a multiclass machine learning model (LightGBM) can non-invasively classify healthy lymph nodes and differentiate between major Non-Hodgkin lymphoma (NHL) subtypes—specifically DLBCL, FL, CLL, and MCL—using routine contrast-enhanced CT. The findings establish a high-accuracy "precision imaging" approach for subtype-level classification, which could streamline biopsy guidance and enhance therapeutic monitoring.

The researchers used mint Lesion to perform semi-automatic 3D segmentation of 1,762 individual lymph nodes and to conduct the subsequent texture analysis. The software was utilized to extract 78 radiomic features per lesion following the Image Biomarker Standardisation Initiative (IBSI) guidelines to ensure standardized quantification of tumor heterogeneity.

DOI: 10.1177/17588359251379665

This study demonstrates that an integrated diagnostic approach combining CT-based radiomics, [¹⁸F]FDG-PET parameters (such as metabolic tumor volume), and clinical staging (UICC 8th edition) can accurately predict early disease progression and survival in patients with small-cell lung cancer (SCLC). Using machine learning and the LASSO algorithm, the researchers developed a multiparametric model that achieved a high predictive capacity (AUC 0.9) for treatment response and was significantly associated with both progression-free survival (PFS) and overall survival (OS).

mint Lesion was used for the standardized objective assessment of treatment response following RECIST 1.1 guidelines.

DOI: 10.3389/fonc.2025.1591742

This study evaluates the use of CT-based machine learning radiomics to preoperatively differentiate between organ-confined (T2) and extravesical (T3) muscle-invasive bladder cancer. The researchers found that integrating clinical biomarkers and optimizing the timing of imaging relative to transurethral resection (TURB) significantly improved the model's predictive performance.

mint Lesion was used to perform semi-automated, three-dimensional (3D) tumor segmentation on contrast-enhanced CT scans to extract full-volume radiomic features. It provided the platform for manual slice-by-slice refinement of tumor boundaries by experienced radiologists to ensure high-quality data for the machine learning algorithms.

DOI: 10.3390/jpm15090400

This publication evaluates a semi-automated method for quantifying imaging features of Hepatocellular Carcinoma (HCC) to enhance the reliability of the LI-RADS (Liver Imaging Reporting and Data System) classification. The study demonstrates that arithmetic assessment of key features, such as Arterial-Phase Hyperenhancement (APHE) and non-peripheral washout, provides high agreement with traditional visual assessment and can help resolve ambiguous cases.

mint Lesion was used for semi-automatic 3D segmentation of liver lesions and background liver to extract volumetric and density data (Hounsfield Units). It facilitated structured reporting by automatically calculating LI-RADS classifications based on extracted values and providing a text module for easy clinical integration.

DOI: 10.1212/WNL.0000000000208956

This study examines the relationship between post-surgical tumor volume and the persistence of B7-H3 targeted CAR-T cells in patients with recurrent glioblastoma multiforme (GBM). The results demonstrated that CAR-T cell persistence was higher in patients with larger post-surgical tumor volumes, suggesting that this therapy may be a promising regimen for first-recurrence GBM.

mint Lesion was used to generate tumor response assessment reports. These reports provided critical volumetric measurements of the tumor lesions, specifically categorizing them into target enhancing, non-target enhancing, and non-target non-enhancing lesions.

DOI: 10.1007/s00259-025-07677-2

This study evaluates the clinical utility of a high-performing AI segmentation model for automated TNM staging in 306 treatment-naïve patients with non-small cell lung cancer (NSCLC) using [18F]FDG PET/CT. While the model demonstrated high lesion detection sensitivity, it achieved only 67.6% concordance with expert-derived UICC staging (9th edition), primarily due to frequent upstaging caused by false-positive predictions.

To establish the ground truth for the study, two hybrid imaging experts used mint Lesion to review all cases. All lesions were annotated and staged according to the 9th edition of the TNM classification system.

DOI: 10.1080/02656736.2025.2525422

This retrospective study uses mathematical modeling to characterize the volumetric involution of microwave liver ablation zones in primary and metastatic tumors, such as colorectal liver metastases. The research demonstrates that this involution is best described by mono-exponential decay, with zones typically stabilizing at approximately one-third of their baseline volume within one year. Significant predictors of these dynamics include initial tumor diameter, initial ablation zone volume, and the tumor-to-ablation zone (T:AZ) volume ratio.

Researchers utilized mint Lesion to perform longitudinal semi-automatic segmentations of the ablation zones using the "interpolated VOI" tool, which employs a boundary-based interpolation algorithm. The software enabled the precise calculation of three-dimensional volumes by summing segmented voxels rather than assuming an ellipsoid or other geometric model.

DOI: 10.1016/j.ejrad.2025.112146

This study developed an XGBoost machine learning model to non-invasively differentiate grade 4 renal cell carcinoma (RCC) from lower-grade tumors using radiomic features extracted from pre-treatment CT images. The model, which achieved a high AUC of 0.92 in the testing set, utilized the ISUP-WHO 2016 grading system as a reference standard and adhered to CLEAR and METRICS reporting guidelines.

Researchers utilized mint Lesion for image assessment and segmentation, defining a region of interest (ROI) at the largest axial diameter of the tumor. The software extracted 72 texture parameters in accordance with the Image Biomarker Standardisation Initiative (IBSI) to ensure international standardization and reproducibility.

DOI: 10.1016/j.acra.2024.11.054

This multicenter study within the RACOON network found that low pectoralis muscle density (LPMD) is a strong independent predictor of 30-day mortality in male patients and individuals under 60 hospitalized with COVID-19. While muscle quality was predictive, pectoralis muscle area did not predict mortality, and no significant associations were found for female patients or those over age 60. Assessment of patients also included a visual lung damage CT score to classify the extent of ground glass opacities and consolidations.

mint Lesion was utilized as the dedicated reporting and viewing software for radiologists to manually perform measurements of the pectoralis major and minor muscles. Specifically, it was used to draw polygonal regions of interest (ROI) on axial CT images to calculate pectoralis muscle area and density.

DOI: 10.1186/s12880-025-01727-9

This study investigates a machine learning-based radiomics algorithm to non-invasively differentiate upper urinary tract urothelial carcinoma (UTUC) from renal cell carcinoma (RCC) using preoperative venous-phase CT datasets. Utilizing a LASSO regression model, the researchers analyzed 59 standardized radiomic features to achieve high diagnostic accuracy in distinguishing these tumor entities and identifying high-grade versus low-grade UTUC.

mint Lesion was utilized to perform manual tumor segmentation in the axial plane according to International Image Biomarker Standardization Initiative (IBSI) standards. Additionally, the software's algorithm was used to extract 59 radiomic features, encompassing first-order statistics and second-order texture descriptors, from the entire delineated tumor volume.

DOI: 10.1055/a-2452-6978

This publication discusses the integration of AI-based systems into the reporting of multiparametric MRI (mpMRI) for prostate cancer to enhance diagnostic accuracy and workflow efficiency. It highlights the importance of standardized assessment guidelines, including PI-RADS version 2.1 and the S3-Guideline Prostate Cancer, to ensure high-quality and reproducible reporting.

mint Lesion was utilized as a software for generating structured reports, serving as the platform where AI-identified lesions are pre-segmented within T2-weighted sequences for the radiologist to review.

DOI: 10.1002/pros.24909

This sub-group analysis of OSPREY cohort B patients with recurrent or metastatic prostate cancer found that the diagnostic performance of 18F-piflufolastat PET/CT was unaffected by concomitant hormone therapy (HT) or castration status. The study demonstrated high median sensitivity (95.3%–96.4%) and similar positive predictive values across patient groups, regardless of whether they were receiving HT.

mint Lesion was used by independent readers to perform SUV measurements (SUVmax and SUVpeak) for identified lesions in various locations, such as bone, lymph nodes, and soft tissue. Readers utilized the software to place a volume of interest (VOI) on each lesion to determine the maximum and peak standardized uptake values.

DOI: 10.1159/000545965

This retrospective multicenter study compared the diagnostic performance of contrast-enhanced CT and gadoxetic acid-enhanced MRI for diagnosing hepatocellular carcinoma (HCC) according to LI-RADS, APASL, and KLCA-NCC guidelines. The findings revealed that while MRI offered higher sensitivity under APASL and KLCA-NCC criteria, CT was more sensitive under LI-RADS due to the specific timing requirements for washout.

The study utilized mint Lesion to develop a review system and templates for structured image analysis. Radiologists used this platform to complete 46 questionnaires for assessing MRI features and 31 questionnaires for CT imaging features, specifically targeting major and ancillary features of focal liver lesions. As these questionnaires were completed, mint Lesion automatically assigned diagnostic classifications for each lesion based on LI-RADS, KLCA-NCC, and APASL guidelines. While the diagnostic results were automated, reviewers had the option for manual interaction, as they were permitted to manually adjust LI-RADS categories in specific cases where tie-breaking rules applied.

DOI: 10.1093/jamia/ocae259

This study establishes a real-world federated learning (FL) infrastructure within the German Radiological Cooperative Network (RACOON) to train deep learning segmentation models for lung pathologies, specifically consolidation, ground-glass opacity, and pleural effusion, on CT scans. The authors provide a comprehensive guide for overcoming practical and legal hurdles in radiology FL and demonstrate through benchmarking that collaborative FL approaches outperform local training and ensembling in both personalization and generalization scenarios.

Each participating hospital is equipped with a server hosting mint Lesion to facilitate structured radiological reporting. It serves as a key component of the initiative's clinical IT ecosystem alongside other tools used for imaging data annotation and processing.

DOI: 10.1016/j.clinimag.2024.110303

This German multicenter study, conducted within the RACOON project, identifies the presence of pleural effusion (PE) as a significant independent predictor of increased 30-day mortality, ICU admission, and the need for mechanical ventilation in COVID-19 patients. Utilizing a visual lung damage CT scoring system to assess disease severity, the researchers determined that the mere detection of PE on CT scans is a critical prognostic marker, regardless of the effusion's volume or density.

Radiologists used mint Lesion as a dedicated reporting and viewing software to manually measure the presence, width, and density of pleural effusions. The software supported a standardized measurement scheme where readers, blinded to clinical outcomes, performed calculations and placed regions of interest.

DOI: 10.1016/j.ejrad.2024.111825.

This study attempted to replicate a previously successful machine learning model that uses [18F]FDG PET/CT radiomics to predict PD-L1 expression in non-small cell lung cancer patients. While the original results were not mirrored in this cohort, the findings highlight the critical importance of rigorous validation and the ongoing challenges of standardizing radiomics for clinical use.

DOI: 10.1016/j.ejro.2025.100636

This prospective study utilized machine learning algorithms to evaluate the diagnostic performance of radiomics extracted from CT and MRI for identifying acute pancreatitis (AP), as defined by the revised Atlanta classification. The results demonstrated that MRI radiomics (specifically T2-weighted imaging) outperformed CT models, while a multi-modality approach combining both CECT and MRI achieved the highest diagnostic accuracy.

The researchers used mint Lesion to manually perform three-dimensional segmentation of the entire pancreatic parenchyma across CECT and multiple MRI sequences. This segmentation defined the volumes of interest from which 107 radiomics features—characterizing shape, intensity distribution, and texture—were extracted for analysis.

DOI: 10.1007/s00259-024-06944-y

This study evaluates the feasibility of using the novel somatostatin receptor (SSTR)-directed radiotracer [18F]SiTATE for PET/CT imaging in patients with differentiated (DTC) and medullary thyroid carcinoma (MTC). The researchers found that [18F]SiTATE PET/CT effectively identifies metastatic lesions and correlates significantly with calcitonin tumor markers in MTC, offering a logistical alternative to 68Ga-based tracers for assessing peptide receptor radionuclide therapy (PRRT) eligibility. While PET results were compared using PERCIST 1.0, the study also noted that reporting could be standardized in the future using the SSTR-RADS 1.0 framework.

mint Lesion was employed to evaluate CT datasets by manually measuring target and non-target lesions. These assessments were conducted according to RECIST 1.1 guidelines to define the metastatic status at baseline and follow-up.

DOI: 10.1007/s00270-024-03948-x

This clinical investigation demonstrates that bolus-tracked biphasic contrast-enhanced CT (CECT) significantly improves the conspicuity and semi-automatic segmentation quality of microwave liver ablation zones, particularly for colorectal cancer metastases. The results indicate that imaging quality declines by 3–4% for each minute that passes after ablation, making early post-procedural imaging critical for accurate assessment.

The study utilized mint Lesion's “interpolated VOI” tool to perform semi-automatic 3D segmentation of ablation zones, where an algorithm identifies object boundaries based on a user-defined rough extent. The software's performance was evaluated using a five-point Likert scale to measure the automated algorithm's quality and the necessity of manual intervention.

2024

DOI: 10.1016/j.ejrad.2024.111806

DOI: 10.1007/s00345-024-05143-6

DOI: 10.1016/j.clinimag.2024.110303

DOI: 10.1007/s00330-024-11060-4

DOI: 10.1186/s13244-024-01787-5

DOI:  10.1007/s00277-023-05507-9

DOI: 10.1038/s41467-024-48416-9

DOI: 10.3390/cancers16101827

DOI: 10.1186/s42466-024-00310-x

DOI: 10.1148/radiol.233114

DOI: 10.1016/j.ejro.2024.100576

DOI: 10.1212/WNL.0000000000205643

DOI: org/10.1016/j.ejrad.2024.111447

DOI: 10.2967/jnumed.123.266817

DOI: 10.3390/cancers16050993

DOI: 10.3390/diagnostics14020198

2023

DOI: 10.3390/diagnostics13040744 

DOI: 10.3389/fendo.2023.1244342

DOI: 10.1016/j.acra.2023.12.003

DOI: 10.1097/CMR.0000000000000952

DOI: 10.3390/cancers15235630

DOI. 10.1007/s42399-023-01618-6

DOI: 10.3390/cancers15204942

DOI: 10.3390/cancers15184650

DOI: 10.1148/rycan.230001

DOI: 10.3390/cancers15153757

DOI: 10.1148/rycan.220022

DOI: 10.1186/s40170-023-00306-2

DOI: 10.1007/s00247-023-05622-9

DOI: 10.3390/onco3020006

DOI: 10.3389/fsurg.2023.1106177

DOI: 10.2967/jnumed.122.265077

DOI: 10.1016/j.ejrad.2023.110827

2022

DOI: 10.3233/SHTI220804

DOI: 10.3389/fonc.2022.993611

DOI: 10.1007/s00259-022-06075-2

DOI: 10.3390/healthcare10101899

DOI: 10.3390/healthcare10081471

DOI: 10.1002/cnr2.1687

DOI: 10.1016/j.ejrad.2022.110454

DOI: 10.3390/biomedicines10050946

DOI: 10.1007/s00432-022-03997-0

DOI: 10.3390/cancers14020393

DOI: 10.3390/diagnostics12030585

DOI: 10.1093/neuonc/noac037

2021

DOI: 10.1007/s00117-021-00924-1

DOI: 10.1097/COC.0000000000000862

DOI: 10.1136/jitc-2021-003261

DOI: 10.1016/j.xcrm.2021.100444

DOI: 10.1016/j.jvir.2021.07.008

DOI: 10.2214/AJR.20.22823

DOI: 10.1016/j.acra.2021.04.015

DOI: 10.1016/j.mri.2021.06.003

DOI: 10.1038/s41746-021-00439-y

DOI: 10.1016/j.ejrad.2021.109603

DOI: 10.1186/s40644-020-00374-3

DOI: 10.3390/cancers13020317

2020

DOI: 10.1007/s00259-020-05165-3

DOI: 10.1007/s00330-020-07511-3

DOI: 10.1177/0846537120968782

DOI: 10.1007/s11060-020-03641-2

DOI: 10.1097/RLI.0000000000000724

DOI: 10.2217/imt-2020-0155

DOI: 10.1007/s00330-020-07298-3

DOI: 10.1016/j.ejrad.2020.109242

DOI: 10.1007/s00432-020-03370-z

DOI: 10.3174/ajnr.A6555

DOI: 10.1101/2020.04.02.20048793

2019

DOI: 10.1177/0284185119887588

DOI: 10.1155/2019/1517208

DOI: 10.1097/RCT.0000000000000930

DOI:  10.18632/oncotarget.27239

DOI: 10.1007/s00761-019-00685-6

DOI: 10.1016/j.ejrad.2019.108688

2018

DOI: 10.1016/j.ccell.2018.06.004

DOI: 10.1007/s00330-017-5014-6. Epub 2017 Sep 7

DOI: 10.2967/jnumed.117.199737

DOI: 10.2967/jnumed.117.199760

DOI: 10.1007/s00270-018-2004-2

DOI: 10.1371/journal.pone.0195270

DOI: 10.1093/nop/npx008

2017

DOI: 10.1007/s00117-017-0289-9

DOI: 10.1055/s-0043-116220

DOI: 10.1016/j.ijrobp.2017.06.061

DOI: 10.1159/000478004

DOI: 10.1007/s00432-017-2488-1

2016

DOI 10.1055/s-0036-1581767

DOI: 10.1007/s12312-016-0131-8

DOI: 10.1016/S1470-2045(16)30463-6

2014

DOI: 10.1055/s-0034-1366741

2013

DOI: 10.1016/j.jvir.2014.01.025

DOI: 10.3748/wjg.v19.i46.8502

DOI: 10.1055/s-0033-1344352

DOI: 10.1016/j.acra.2012.09.029