Research Presentation Session: Abdominal Viscera & GI Tract
RPS 1501 - Insights in pancreatic imaging with current and novel techniques
March 1, 14:00 - 15:30 CET
7 min
T1 and T2 mapping quantitative MRI in the characterisation of pancreatic cystic lesions: preliminary results
Alessandro Beleù, Treviso / Italy
- Author Block: A. Beleù1, R. Napoli2, E. Grosso1, C. Nistri1, M. Piccino1, M. Massani1, G. Zanus1, G. Morana1; 1Treviso/IT, 2Padua/ITPurpose: MRI characterisation of pancreatic cystic lesions is mainly based on qualitative and morphological criteria. Quantitative imaging techniques, such as T1 and T2 relaxation times mapping, could help in their characterisation of these lesions.Methods or Background: Patients undergoing abdominal MRI were prospectively enrolled. The T1- and T2-values of the content of the pancreatic cyst were sampled by a single radiologist by evaluating mean, minimum, maximum and range of T1 and T
- Dimensional and clinical-laboratory parameters were collected. Relaxation times were compared in different lesion groups using ANOVA and Kruskall-Wallis test. Results or Findings: Of the 350 patients enrolled, 126 were finally included (age 68±11 years, 55% male). 106 branch-duct intraductal papillary mucinous neoplasms (IPMN-BD), two main duct (IPMN-MD), six serous cystadenomas (SCA), two mucinous cystadenomas (MCA), two cystic neuroendocrine tumours (cNET), two pseudocysts, two walled-off necrosis (WON), one lymphoepithelial cyst (LEC) and one cystic chronic pancreatitis were included. Mean T1 of all cystic lesions was 1985±826 ms, while median T2 was 96 [82-126] ms. T1 values were significantly different among all lesions (p=.02). IPMN-BD demonstrated mean (p=.02), minimum (p=.01), and maximum (p=.02) T2 values significantly different from SCA; ROC curve test showed that mean (AUC
- 78; p=.02), minimum (AUC 0.8; p=01) and maximum (AUC 0.78; p=.02) T2 values are sensitive in the differential diagnosis of IPMN-BD vs. SCA. A significant difference in mean T1 was observed (p=.03) compared to IPMN-BD and MCA. IPMN-BD showed significant difference in mean (p=.01), minimum (p=.01), maximum (p=.02) T1 and T2 range (p=.01) compared to WON. SCA mean (p=.01) and minimum T1 (p=.05) were significantly different to MCA. Conclusion: The T1 and T2 relaxation times of pancreatic cystic lesions are different in many different lesion groups.Limitations: The relatively low number of lesions, with different distribution in the groups, was an identified limitation.Funding for this study: No funding was received for this study.Has your study been approved by an ethics committee? Not applicableEthics committee - additional information: No information provided by the submitter.
7 min
Development and validation of an imaging-based model to predict the malignant potential of intraductal papillary mucinous neoplasm of the pancreas: comparison with international consensus guidelines
Junghoan Park, Seoul / Korea, Republic of
- Author Block: J. Park1, J. H. Kim1, J. S. Bae1, H-J. Kang1, S-Y. Choi2; 1Seoul/KR, 2Bucheon/KRPurpose: The study aimed to develop and validate imaging-based models for predicting the malignant risk of intraductal papillary mucinous neoplasm (IPMN).Methods or Background: We retrospectively analysed data from 245 surgically-confirmed IPMN patients who had preoperative CT and MRI scans for model development. Two radiologists assessed cyst size, presence of enhancing mural nodule (EMN), EMN size, main pancreatic duct (MPD) diameter, thickened/enhancing cyst wall, abrupt MPD calibre change with distal pancreatic atrophy, and lymphadenopathy. Multiple logistic regression models predicting malignancy risk were created using either continuous (model C) or dichotomised variables (model D) from the significant imaging features on univariable analysis. Validation included internal (n=55) and external (n=43) datasets with pathologically-confirmed IPMNs. Model performance was assessed using the area under the receiver-operating characteristic curve (AUC) and compared with the Fukuoka guideline-based model (model F).Results or Findings: Model C identified age, EMN size, MPD diameter, and lymphadenopathy as independent predictors on CT, and age, presence of EMN, and EMN size on MRI. In model D, independent predictors were age ≥ 68, size ≥ 31 mm, EMN ≥ 6 mm, MPD ≥ 7 mm, and lymphadenopathy on CT, and age ≥ 68, size ≥ 33 mm, EMN ≥
- 5 mm, MPD ≥ 7.5 mm, and lymphadenopathy on MRI. Model C (AUCs for external datasets, 0.785-0.899) performed slightly better than model D (AUCs, 0.736-0.912) without statistical significance. No significant difference was observed between models C and F (AUCs, 0.827-0.952). Combining model C with the presence of obstructive jaundice improved diagnostic performance (AUCs, 0.883-0.941) without statistical significance. Conclusion: Our imaging-based models effectively predicted the malignant risk of IPMN, and its combination with clinical findings enhanced diagnostic accuracy.Limitations: Identified limitations were the retrospective nature of this study and the limited sample size.Funding for this study: No funding was received for this study.Has your study been approved by an ethics committee? YesEthics committee - additional information: This study was approved by the institutional review board of Seoul National University Hospital, and the requirement for informed consent was waived because of the retrospective nature of this study.
7 min
Common imaging findings and clinical effects of high-grade pancreatic intraepithelial neoplasia in the remnant pancreas in patients with intraductal papillary mucinous neoplasms of the pancreas
Jung Hoon Kim, Seoul / Korea, Republic of
- Author Block: J. H. Kim, M. C. Kim, S. K. Jeon, H-J. Kang; Seoul/KRPurpose: The study aimed to investigate the common computed tomography (CT) findings and clinical effects of HG PanIN in the remnant pancreas in patients with intraductal papillary mucinous neoplasm (IPMN) of the pancreas.Methods or Background: Two hundred and fifty one patients with surgically confirmed IPMNs (118 malignant [invasive carcinoma/high-grade dysplasia] and 133 benign [low-grade dysplasia]) were retrospectively enrolled. The grade of PanIN (233 absent/low-grade and 18 high-grade) was recorded, and all patients underwent serial CT follow-up before and after surgery. Two radiologists analysed the CT findings of high-risk stigmata or worrisome features according to 2017 international consensus guidelines. They also analysed tumour recurrence on serial follow-up CT after surgery. Statistical analyses were performed to identify significant predictors and clinical impact on postoperative outcomes of HG PanIN.Results or Findings: PanIN grade showed a significant association with IPMN grade (p=
- 012). Enhancing mural nodules ≥5 mm, abrupt main pancreatic duct (MPD) changes with distal pancreatic atrophy, increased mural nodule size and MPD diameter were common findings in HG PanIN (P<0.05). In multivariate analysis, abrupt MPD change with distal pancreatic atrophy (odds ratio [OR] 6.59, 95% CI: 2.32-18.72, <0.001) and mural nodule size (OR 1.05, 95% CI: 1.02-1.08, 0.004) were important predictors for HG PanIN. During postoperative follow-up, HG PanIN (OR 4.98, 95% CI: 1.22-20.33, 0.025) was significantly associated with cancer recurrence in the remnant pancreas. Conclusion: Although PanIN is a microscopic, non-invasive precursor of invasive carcinoma, CT can be useful for predicting HG PanIN using common features, such as abrupt MPD changes and mural nodules. In HG PanIN, extra caution is needed to monitor postoperative recurrence during follow-up.Limitations: Identified limitations were the retrospective nature of the study and its limited sample size.Funding for this study: No funding was received for this study.Has your study been approved by an ethics committee? YesEthics committee - additional information: The study was approved by Seoul National University Institutional Review Board with approval code: 2007-183-
7 min
Prediction of tumour cellularity in resectable PDAC from preoperative computed tomography imaging
Friederike Jungmann, Munich / Germany
Author Block: F. Jungmann, G. Kaissis, S. Ziegelmayer, F. N. Harder, K. Steiger, M. Makowski, R. Braren, F. Lohöfer; Munich/DE
Purpose: Pancreatic ductal adenocarcinoma (PDAC) remains a tumour entity with a poor prognosis and a 5-year survival rate below 10%. Until today, non-invasive prediction of individual patient outcomes remains an unresolved task. We previously showed a strong association between magnetic resonance imaging-based tumour cell estimates and patient survival. In this study, we aimed to transfer this finding to more broadly applied computed tomography (CT) imaging for non-invasive risk stratification.
Methods or Background: Discrete cellularity regions of the PDAC resection specimen were analysed by a routine histopathological workup. Regional tumour cellularity and CT-derived Hounsfield Units (HU), as well as iodine concentrations, were regionally matched. One-way ANOVA and pairwise t-tests were performed to assess the relationship between different cellularity levels in conventional, virtual monoenergetic 40 keV (monoE 40 keV) and iodine map reconstructions.
Results or Findings: A statistically significant negative correlation between regional tumour cellularity in histopathology and CT-derived HU from corresponding image regions was identified. Radiological differentiation was best possible in monoE 40 keV CT images. However, HU values differed significantly in conventional reconstructions as well, indicating the possibility of a broad clinical application of this finding.
Conclusion: In this study, we establish a novel method for CT-based prediction of tumour cellularity for in-vivo tumour characterisation in PDAC patients.
Limitations: Identified limitations were: (1) the retrospective nature of the study; (2) the limited cohort size; and (3) that this was a single institution study.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the local institutional review board of the Technical University of Munich.
Purpose: Pancreatic ductal adenocarcinoma (PDAC) remains a tumour entity with a poor prognosis and a 5-year survival rate below 10%. Until today, non-invasive prediction of individual patient outcomes remains an unresolved task. We previously showed a strong association between magnetic resonance imaging-based tumour cell estimates and patient survival. In this study, we aimed to transfer this finding to more broadly applied computed tomography (CT) imaging for non-invasive risk stratification.
Methods or Background: Discrete cellularity regions of the PDAC resection specimen were analysed by a routine histopathological workup. Regional tumour cellularity and CT-derived Hounsfield Units (HU), as well as iodine concentrations, were regionally matched. One-way ANOVA and pairwise t-tests were performed to assess the relationship between different cellularity levels in conventional, virtual monoenergetic 40 keV (monoE 40 keV) and iodine map reconstructions.
Results or Findings: A statistically significant negative correlation between regional tumour cellularity in histopathology and CT-derived HU from corresponding image regions was identified. Radiological differentiation was best possible in monoE 40 keV CT images. However, HU values differed significantly in conventional reconstructions as well, indicating the possibility of a broad clinical application of this finding.
Conclusion: In this study, we establish a novel method for CT-based prediction of tumour cellularity for in-vivo tumour characterisation in PDAC patients.
Limitations: Identified limitations were: (1) the retrospective nature of the study; (2) the limited cohort size; and (3) that this was a single institution study.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the local institutional review board of the Technical University of Munich.
7 min
A nomogram based on contrast-enhanced CT radiomics to preoperatively predict perineural invasion for patients with pancreatic adenocarcinoma
Yan Deng, Chengdu / China
- Author Block: Y. Deng1, H. Yu1, Z. Huang1, B. Song2; 1Chengdu/CN, 2Sanya/CNPurpose: The study aimed to develop and validate a nomogram based on contrast-enhanced CT (CECT) radiomics features and clinical characteristics for preoperative assessment of perineural invasion (PNI) in pancreatic adenocarcinoma (PAC).Methods or Background: The retrospective study recruited 217 PAC patients with histopathology and randomly divided them into a training and testing cohort, at a ratio of 7:
- Radiomics features were extracted from the artery and portal venous phase of CECT. Univariate analysis and least absolute shrinkage and selection operator (LASSO) method were applied using the 10-fold cross-validation for feature selection in the training cohort. The selected features were integrated into the radiomics score (Rad-score). Two experienced radiologists evaluated the status of PNI based on CECT (CTPNI). A nomogram was constructed by Rad-score and characteristics of statistically significant differences. Calibration and classification metrics were used to evaluate the nomogram performance. Results or Findings: The lymph node status determined on CT (CTLN) and CTPNI were statistically different between the PNI positive and negative groups. The AUC for the Rad-score in the training cohort was close to that in the testing cohort (AUC
- 720 and 0.640). The AUC of CTPNI was 0.610 in the training cohort and 0.675 in the testing cohort. The Rad-score, CTLN and CTPNI were used to construct a nomogram, and it achieves favourable discrimination of PNI status with an AUC of 0.846 and 0.741 in the training and testing cohort, respectively. The nomogram achieved best performance both in the training cohort and testing cohort compared with Rad-score and CTPNI (P<0.05). Conclusion: The development of a nomogram incorporating Rad-score, CTLN, and CTPNI holds promise as a valuable non-invasive tool for the preoperative evaluation of PNI status in PAC patients.Limitations: It is a retrospective, single-centre study with limited sample size, and no external validation group.Funding for this study: No funding was received for this study.Has your study been approved by an ethics committee? YesEthics committee - additional information: The retrospective study was approved by the institutional review board of West China Hospital of Sichuan University. Each patient waived informed consent.
7 min
Evaluation of the efficacy of neoadjuvant chemotherapy in pancreatic ductal adenocarcinoma by three-dimensional quantitative analysis of spectral CT
Dan Zhang, Chongqing / China
- Author Block: D. Zhang1, X. Zhang2, Y. Zou1; 1Chongqing/CN, 2Chengdu/CNPurpose: To explore the value of three-dimensional quantitative analysis of spectral CT in the efficacy evaluation of neoadjuvant chemotherapy for pancreatic ductal adenocarcinoma (PDAC).Methods or Background: This retrospective study enrolled 49 patients with PDAC who underwent contrast-enhanced spectral CT before one month and after two months of neoadjuvant chemotherapy from October 2021 to June
- Conventional polyenergetic images (PEIs), iodine-concentration (IC) images and Z-effective (Z-eff) images were generated at portal venous phase (PVP). The whole tumour volume, IC-total and Z-eff-total were measured using a semiautomatic segment software. The changes of the longest diameter and whole tumour volume, IC-mean and IC-total, Z-eff-mean and Z-eff-total were compared using Wilcoxon signed rank test. The cut-off value of changes for curative effect was obtained through a receiver operating characteristic (ROC) curve analysis. Results or Findings: The changes of the longest diameter and whole tumour volume (P＜
- 001), IC-mean and IC-total (P＜0.001), Z-eff-mean and Z-eff-total (P＜0.001) showed a significant difference, respectively. The change of the longest diameter’s cut-off value was 0.193 (sensitivity 57.1%, specificity 93.3%, area under the ROC curve [AUC] 0.725). The change of the whole tumour volume’s cut-off value was 0.222 (sensitivity 85.7%, specificity 73.3%, AUC 0.746). The change of the IC-total’s cut-off value was 0.191 (sensitivity 92.9%, specificity 60%, AUC 0.795). The change of the Z-eff-total’s cut-off value was 0.167 (sensitivity 85.7%, specificity 66.7%, AUC 0.800). Conclusion: Three-dimensional quantitative parameters of spectral CT can effectively evaluate the efficacy of neoadjuvant chemotherapy in PDAC, and the diagnostic efficacy is higher than that of lesion’s size.Limitations: An identified limitation was the small number of patients in the study.Funding for this study: No funding was received for this study.Has your study been approved by an ethics committee? YesEthics committee - additional information: This study was approved by the local institutional review board and informed consent was obtained from all patients.
7 min
Utility of conventional CT and MRI parameters in grading the severity of chronic pancreatitis: can delayed enhancement ratio be the guiding light?
Mohak Narang, New Delhi / India
- Author Block: M. Narang, S. Sharma, K. S. Madhusudhan; New Delhi/INPurpose: The study aimed to evaluate the role of conventional CT and MRI parameters used in routine clinical practice for grading the severity of chronic pancreatitis (CP).Methods or Background: This prospective study conducted from December 2019 to December 2021 included 72 patients (mean age: 30 years; 59 men) with suspected or confirmed CP graded as equivocal, mild, and moderate-marked using composite imaging (Cambridge classification) and endoscopic ultrasound (Rosemont’s criteria) who underwent multiphasic CT and MRI of the abdomen. Routine parameters such as antero-posterior pancreatic thickness (APPT) and delayed enhancement ratio (DER) were evaluated on CT. APPT, signal intensity ratio (SIR) on T1-weighted sequence, pancreatic duct contour (PDC), apparent diffusion coefficient (ADC) and DER were evaluated on MRI. Multiphasic CT of 20 renal donors and abdominal MRI of 20 patients with renal mass served as controls. One way ANOVA with post-hoc Dunn-Bonferroni correction was used for analysing APPT, ADC, SIR and DER. PDC were compared using chi-square test.Results or Findings: The study group included patients with equivocal CP (n=20), mild CP (n=18) and moderate-marked CP (n=34). All individual parameters were significantly different between controls and moderate-marked CP (p<
- 05). Additionally, DER (on CT and MRI) was significantly different between controls and mild CP (p<0.02, cut-off = 1.77 on CT, sensitivity = 88.89% and specificity = 75% and 3.56 on MRI, sensitivity = 83.33% and specificity = 95%). Area under curve (AUC) for DER on CT and MRI for differentiating controls and mild CP were 0.87 and 0.91 respectively. Conclusion: DER on CT and MRI is a simple and accurate imaging marker for grading of CP and may be useful in the diagnosis of mild CP before the appearance of ductal changes.Limitations: The small sample size was an identified limitation of this study.Funding for this study: No funding was received for this study.Has your study been approved by an ethics committee? YesEthics committee - additional information: The study was approved by the Institutional Ethics Committee.
7 min
Intra- and interpatient variability of iodine concentration in pancreatic parenchyma for individualised contrast agent application protocols in dual-layer spectral CT
Mathis Franz Georg Konrad, Heidelberg / Germany
- Author Block: M. F. G. Konrad, L. M. Wünschmann, T. Mokry, H-U. Kauczor, T. F. Weber, W. Stiller; Heidelberg/DEPurpose: Verifying the quantifiability of intra- and interpatient variability of iodine concentration (IC) in pancreatic parenchyma measured using dual-layer spectral CT for distinct individualised contrast agent (CA) application protocols in relation to liver parenchyma.Methods or Background: To compare the variability of IC in pancreatic parenchyma with that known for liver parenchyma, the mean IC of healthy pancreatic parenchyma was measured for 150 oncologic patients, who were retrospectively allocated to three groups (G1-G3, n=50 each) in a previous study. Imaging data were acquired during portal-venous contrast-enhanced thoraco-abdominal spectral CT staging examinations at two subsequent time points. The applied CA volume was either chosen by a radiographer based on body-mass-index resulting in identical (G1) or different (G2) volume for both examinations, or individually calculated (G3) using a predefined function. Intra- and interpatient coefficients of variation (CV) were calculated for the measured IC and attenuation in virtual monoenergetic image data (VMI, 40 keV) without and with normalisation to the reference structures portal vein, aorta, and spleen, respectively.Results or Findings: Intra- and interpatient CV for IC were low in G3 (
- 8% and 21.6%), higher in G2 (23.2% and 23.6%) and mixed in G1 (10.6% and 27.0%). Variability of IC was lower compared to liver parenchyma in G3 (-6.0% and -3.6%). Values with the highest variability (interpatient CV for G1) profit the most from normalisation (mean CV reduction of -5.0%). Normalisation to portal vein and aorta leads to examination-wise differences in IC no longer being significant (p>0.05). For all patient groups all CVs are lower for VMI compared to IC (-7.6% to -0.1%). Conclusion: Variability of IC in pancreatic and liver parenchyma is comparable. Individualised calculation of CA volume leads to lower variability of IC. Normalisation can stabilise high-variability measurements.Limitations: An identified limitation was the retrospective, single-centre nature of this study.Funding for this study: This study received research grant support from Philips Healthcare.Has your study been approved by an ethics committee? YesEthics committee - additional information: This study was approved by the ethics committee of the Medical Faculty of the University of Heidelberg, approval number S-348/2019
7 min
Quantitative image analysis in pancreatic cancer: might contrast agents become redundant?
Jennifer Gotta, Frankfurt a. Main / Germany
- Author Block: J. Gotta, V. Koch, L. D. Grünewald, S. Mahmoudi, S. Martin, T. Vogl; Frankfurt/DEPurpose: The evolution of computed tomography (CT) technology alongside the emerging field of radiomics has opened promising avenues in diagnostic assessment of cancer. This study seeks to explore the potential of a machine learning classifier rooted in dual-energy computed tomography (DECT) radiomics, specifically in its ability to differentiate between malignant and benign pancreatic lesions in non-contrast scans.Methods or Background: In this study, a total of 100 patients who underwent third‑generation DECT between November 2018 and October 2022 were included. Of these, 60 patients had pancreatic cancer, while 40 had normal pancreatic tissue. 107 radiomics features per patient were extracted from non-contrast and arterial-enhanced DECT scans. To develop and validate our models, the dataset was divided into distinct training and test subsets. Stepwise feature reduction was conducted to identify the most important features. Subsequently, two gradient-boosted tree models were trained.Results or Findings: The trained machine learning classifiers achieved a diagnostic accuracy of
- 97 in the arterial-enhanced model and 0.88 in non-contrast scans with areas under the curve of 0.97 (95% CI, 0.9178-1, p<0.001) and 0.96 (95% CI, 0.9013-1, p <0.001), respectively. No significant differences were observed between both models (p=0.52). Meanwhile, both CT-based radiomics models exhibited similar results in distinguishing between pancreatic tumours and normal pancreatic tissue. Conclusion: Our study demonstrates the potential for employing radiomics on non-contrast DECT imaging to differentiate pancreatic cancer from healthy tissue. This innovative method holds promise in enhancing the early detection of pancreatic cancer, reducing radiation exposure and ultimately improving patient prognosis.Limitations: Firstly, this study was conducted at a single centre. Secondly, due to the retrospective study design, inherent limitations may affect the reliability and generalisability of the findings. Finally, clinical variables were not integrated into our radiomics methodology, even though this might have led to a further increase in diagnostic performance.Funding for this study: No funding was received for this study.Has your study been approved by an ethics committee? YesEthics committee - additional information: This study was approved by the ethics committee of the University Hospital Frankfurt.
This session will not be streamed, nor will it be available on-demand!
This session is accredited with 1.5 CME credits.
Nikolaos Kartalis
Stockholm / Sweden
Alessandro Beleù
Treviso / ItalyJunghoan Park
Seoul / Korea, Republic ofJung Hoon Kim
Seoul / Korea, Republic ofFriederike Jungmann
Munich / GermanyYan Deng
Chengdu / ChinaDan Zhang
Chongqing / ChinaMohak Narang
New Delhi / IndiaMathis Franz Georg Konrad
Heidelberg / GermanyJennifer Gotta
Frankfurt / Germany