RPS 1716 - Innovations in pancreatic cancer imaging

Author Block: E. V. Kondratyev, A. Zharikova, I. Gruzdev, A. Ustalov, S. A. Shmeleva, V. Egorov, E. P. Yasakova, P. V. Markov, D. V. Kalinin; Moscow/RU
Purpose: To develop and compare diagnostic models, including a combinеd model, so as to predict the patologic responce to neoadjuvant chemotherapy (NAC) for pancreatic ductal adenocarcinoma (PDAC).
Methods or Background: 59 patients with histologically confirmed PDAC and preoperative computed tomography (CT) were included in the study.
Patients were divided into two groups depending on the grade of histological response of the tumour based on Tumour Regression Score (TRS) criteria. The first group had a favourable response (TRS 0, 1 ,2), the second - unfavourable response (TRS 3).
A radiologist with 6 years of experience, segmented the region of interest (lesion) for radiomics structure analysis in the arterial and venous CT phases before and after NAC. The extracted texture features were divided into 3 groups (pre-NAC, post-NAC, combined model) and analysed using machine learning techniques.
Results or Findings: The AdaBoost ensemble model (pre-NAC) - ROC AUC (0,831), PR-AUC (0,874) и F1 Score (80%), accuracy (77%), precision (88%), specificity (85,7%) and the Optimized Random Forest (post-NAC) - ROC AUC (0,870), PR-AUC (0,941), F1 Score (85,7%), accuracy (83,3%), precision (90%), recall/sensitivity (81,8%) are the best models for recognising tumours with an unfavourable response, if the high accuracy is priority.
Gradient Boosting is the best fitting model both pre- and post-NAC, when focusing on ROC AUC(0.896) and PR-AUC (0.95).
Comparing the results of the pre- and post-NAC models, the latter were more efficient.
Conclusion: Machine learning models, specifically Optimized Random Forest and Gradient Boosting, trained on texture features from post-NAC CT scans demonstrated high accuracy in detecting non-responders with an unfavorable prognosis.
Limitations: A relatively small sample size and the absence of an external validation group, which complicates the wider application of our model.
Funding for this study: It was not required.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Extract from the minutes № 003-2025 of the meeting of the Scientific Research Ethics Committee of the Federal State Budgetary Institution “A.V. Vishnevsky National Medical Research Center of Surgery” of the Ministry of Health of the Russian Federation dated March 21, 2025.

Author Block: S. Ming, W. Wei; Hefei/CN
Purpose: To investigate the value of dual-energy computed tomography (DECT)-derived extracellular volume fraction (ECV) for predicting tumor collagen ratio and assessing the response to neoadjuvant chemotherapy (NAC) in pancreatic ductal adenocarcinoma (PDAC).
Methods or Background: This retrospective study enrolled 176 patients with pathologically confirmed PDAC (64 resected, 112 unresectable) who underwent DECT. The ECV fraction based on iodine concentration (ECV_IC) was calculated. Histological collagen ratio was measured from surgical specimens using Masson’s trichrome staining and digital image analysis. For unresectable patients, chemotherapy response was evaluated according to RECIST 1.1. Univariate and multivariate analyses were performed to identify predictors of collagen ratio and treatment response. Receiver operating characteristic (ROC) curves were used to evaluate predictive performance.
Results or Findings: ECV_IC showed a strong positive correlation with the histological collagen ratio (r = 0.618, p < 0.001) and was an independent predictor of high collagen ratio (HR = 1.420, p = 0.009). In the unresectable cohort, ECV_IC was also an independent predictor of poor response to NAT (HR = 1.259, p = 0.023). ROC analysis demonstrated that ECV_IC could discriminate between responders and non-responders with area under the curve (AUC) of 0.83 (sensitivity 81.0%, specificity 70.4%) at a cutoff value of 31.8%, significantly outperforming venous-phase iodine concentration (IC_VP; AUC = 0.73, P = 0.026).
Conclusion: DECT-derived ECV_IC is a promising non-invasive biomarker for predicting tumor collagen ratio and assessing response to NAC in PDAC, with potential to guide personalized treatment strategies.
Limitations: First, its retrospective and single-center nature introduces potential selection bias. Second, the chemotherapy regimens were not completely standardized, which might have influenced the response assessment. Third, the evaluation of treatment response was based on RECIST 1.1 criteria rather than pathological confirmation, which remains the gold standard.
Funding for this study: This study warted by the following projects:the National Natural Science Foundation of China (NSFC, No.82271991);Joint Fund for Medical Artificial Intelligence of USTC (No.MAI2023C006);Level A Funding Project for Reserve Candidates of Academic and Technical Leaders in Anhui Province (No.2022H279).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This retrospective study was approved by the ethics committee of First Affiliated Hospital of University of Science and Technology of China (Anhui Provincial Hospital), Approval No: 2024-RE-438.

Author Block: L. Lofino, A. Ammirabile, R. Levi, C. Bonifacio, A. Laghi; Rozzano/IT
Purpose: To compare quantitative and qualitative image quality parameters and radiation dose between photon-counting CT (PCCT) and energy-integrating detector CT (EIDCT) for the detection of pancreatic cancer.
Methods or Background: In this IRB-approved prospective study, 32 patients pancreatic cancer underwent multiphase CT (16 PCCT, 16 EIDCT). Patients were matched by age and BMI. Arterial and 5-minute delayed venous phase imaging were retrieved and analyzed. Contrast-to-noise ratio (CNR) was calculated as follows: (pancreatic parenchyma HU - pancreatic cancer HU) / noise, where noise was considered as the standard deviation of subcutaneous fat attenuation. Three radiologists with 7, 10 and 22 years of experience independently evaluated tumor conspicuity and overall image quality on a 5-point scale. Dose-length product (DLP) was retrieved for all examinations.
Statistical analysis included t-tests, ANOVA and inter-reader agreement.
Results or Findings: PCCT demonstrated significantly higher tumor CNR than EIDCT in both arterial (5.3 ± 0.8 vs 4.0 ± 1.1, P<0.01) and delayed venous phases (4.9 ± 1.0 vs 2.3 ± 0.6, P<0.01). Image quality scores were consistently higher with PCCT for overall image quality (arterial: 4.5 ± 0.4 vs 3.3 ± 1.2; venous: 4.2 ± 0.8 vs 3.0 ± 1.1) and tumor conspicuity (all P<0.03). DLP was significantly lower for PCCT compared with EIDCT (1,050 ± 120 vs 2,350 ± 750 mGy·cm, P<0.01). Inter-reader agreement was strong both for PCCT and for EIDCT. Of note, in 2 patients imaged with both modalities, PCCT showed a clearer tumor delination which allowed for downstaging.
Conclusion: In our population, PCCT has shown higher objective and subjective image quality with reduced radiation dose compared to EIDCT. Enhanced delineation of pancreatic cancer and surrounding structures may improve staging and increase the cohort of potentially operable patients.
Limitations: Small sample size, single center.
Funding for this study: This study was funded by ANTHEM foundation.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethics committee reviewed and approved the study protocol.

Author Block: N. G. Nacul Mora¹, A. Andreou², B. Strücker², S. Katou², H. Morgül², F. Becker², A. Pascher², M. Köhler¹, G. H. Pöhler¹; ¹Muenster/DE, ²Münster/DE
Purpose: To compare NCCN radiological vessel contact grading and binary surgical infiltration assessment in pancreatic cancer with surgical and histological finding as reference.
Methods or Background: Single-center, retrospective cohort study including 103 patients (mean age 68 years, male 61%) with CT-based diagnosis of pancreatic carcinoma (2012 - 2023) undergoing primary resection. CT-tumor-vessel contact was blindly classified following NCCN criteria by two abdominal radiologists. By discrepancies, the statement of the radiologist with longer experience was chosen. Radiological infiltration was defined as any-contact / no-contact and compared to surgical and histological findings using McNemar-test, Spearman-correlation, sensitivity (how well CT detects actual infiltration) and specificity (how well CT rules out infiltration when there’s none). Subgroup analysis of CT interval to operation ≤4 weeks (n=83) vs. >4 weeks (n=20) was compared using ROC-Analysis.
Results or Findings: Vessel infiltration was radiologically assessed higher (radiological 51%, surgical 39%, p= 0.041): arterial infiltration was comparable (radiological 7%, surgical 5%, p= 0.625), but venous infiltration was radiologically higher (radiological 44%, surgical 34%, p= 0.019), with 83.5% overall agreement but systematic radiological overestimation of venous involvement.
Radiological artery infiltration sensitivity was 99% and specificity 98%, venous infiltration sensitivity was 88% and specificity 81%.
CT ≤4 weeks subgroup showed enhanced correlations (arterial AUC: 0.731, venous AUC: 0.888).
69% of patients had R0 resection status, 28% R1 status. The histological finding of R0 patients did not show arterial invasion. There was a significant difference in venous infiltration between surgical and histological invasion (p < 0.001).
Conclusion: The interdisciplinary study reveals the complexity of vascular status in NCCN-based grading, resulting in radiological overestimation of venous infiltration, which may lead to over-recommendation for neoadjuvant therapy in interdisciplinary tumor boards. A CT-surgery interval ≤4 weeks should be prioritized.
Limitations: Small cohort, single-centre Study
Funding for this study: No funding.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethik Komission Westfalen Lippe (2025-506-f-S)

Author Block: S. Han¹, D. H. Lee²; ¹SEOUL/KR, ²Seoul/KR
Purpose: To retrospectively evaluate the clinical significance of PET based systemic surveillance for distant metastasis in borderline resectable pancreatic cancer (BRPC) patients who are undergoing FOLFIRINOX based neoadjuvant chemotherapy (NAC).
Methods or Background: 161 patients who underwent FOLFIRINOX based NAC for BRPC from two institutes during January 2013 to December 2019 were retrospectively reviewed. All of the patients underwent an initial PET scan to confirm local disease. Image review including pre- and post-NAC images was done by 2 radiologists in consensus based on the NCCN 2020 guideline. Clinical information were obtained from patients records. Risk factor analysis for metastasis development was performed.
Results or Findings: Among the 161 patients who underwent NAC, 22.4% (36/161) converted to palliative setting due to local progression (14.3%, 23/161) or distant metastasis (8.1%, 13/161). The remaining 84.6% (125/161) proceeded to preoperative evaluation with (n=70) or without (n=55) PET evaluation. Among the patients with preoperative PET scan, distant metastasis was detected in 7.1% (5/70) of the patients. Among them, one liver metastasis (20%, 1/5) in one patient was only found on PET scan. Overall, 11.8% (19/161) patients developed metastasis during NAC. Among them, 2 cases of metastasis were only detected in the PET scan (2/19, 10.5%). Regarding the risk factors, the increment of CA 19-9 level during treatment (OR 4.43 [1.62-12.05], p=0.004) and present of major vein invasion before NAC (OR 5.70 [1.27-25.62], p=0.02) was associated with development of metastasis during NAC.
Conclusion: Considerable proportion of patients undergoing NAC for BRPC develop distant metastasis during treatment, especially for patients with increased CA 19-9 during the NAC or patients with initial major vein invasion. Systemic PET can aid in detection before undergoing curative resection.
Limitations: This is a retrospective study with heterogenous patient group.
Funding for this study: N/A
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Approved by the IRB of the appropriate institutions.

Author Block: S. Poli¹, A. Wenning², P. Lombardo², V. C. Obermann², A. T. Huber¹; ¹Lucerne/CH, ²Bern/CH
Purpose: Investigate the association between baseline tumor stiffness measured by magnetic resonance elastography (MRE) and survival in pancreatic ductal adenocarcinoma (PDAC). We hypothesized that patients with an MRE “ring sign” (high stiffness in the periphery but low in the center, indicating necrosis) or homogeneously high-stiffness have worse survival than those with homogeneous lower stiffness.
Methods or Background: 55 patients with newly diagnosed PDAC (Jan-21 to April-23, ethics approval Bern, registry-ID:NCT03469726) underwent baseline fasting (≥6 hours) multiparametric-MRI and 3D-MRE (3T Magnetom Prisma, Siemens) for cancer staging and stiffness assessment, and exclusion of liver metastasis. 3D-MRE was performed with a pneumatic driver (Resoundant, 40Hz, 40A), and a spin-echo 3D-MRE sequence (10 slices, 3.5 mm, end-expiration), providing pancreatic coverage. Kaplan-Meier curves and Cox-proportional hazard models were used to compare mortality.
Results or Findings: During follow-up, unresectable PDAC showed poorest survival, confirming its adverse prognosis (0% 3-years survival). Patients with homogeneous stiffness <13.3 kPa demonstrated significantly better outcomes (39% 3-year survival). In contrast, high tumor stiffness (hazard ratio, HR[95%CI] = 2.3[1.2,4.3]; p=<.05) and the presence of the MRE ring sign (HR[95%CI] = 2.1 [1.1,4.0]; p=<.05) were independently associated with increased 3-year mortality (26% and 0% survival, respectively). Resected patients with a ring sign had a prognosis comparable to that of patients with unresectable disease. As tumor necrosis has been histologically linked to aggressive PDAC, our findings support MRE-derived stiffness as a promising, noninvasive survival imaging biomarker.
Conclusion: A single baseline measurement of tumor stiffness by MRE is predictive of patient survival in newly diagnosed PDAC. MRE may serve as a valuable imaging biomarker for risk stratification in PDAC and could be incorporated into standard clinical MRI.
Limitations: The limitation of the study is the relatively small, single-center patient cohort, which may limit generalizability.
Funding for this study: SNF-project:REPORT-IT(#10003604)
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This was a local, single-center sub-study of the DIA-PANC multi-center study (registry ID NCT03469726). The registry was approved by the Bern cantonal ethics committee, and the study was carried out in accordance with the principles of the Declaration of Helsinki.

Author Block: Q. Wang, L. Zhu, J. Liu, M. Dai; Beijing/CN
Purpose: To compare the stiffness and fluidity in the central part and peripheral part of pancreatic cancer using multi-frequency magnetic resonance elastography (MF-MRE), to correlate the MF-MRE parameters with clinicopathological factors that indicates tumor aggressiveness, and to identify risk factors for patient survival.
Methods or Background: MF-MRE was performed in 97 pancreatic cancer patients before treatment. High-resolution shear wave speed (SWS) and loss angle (φ) maps were generated, representing tissue stiffness and fluidity. SWS and φ were measured in the central part, the peripheral part and covering the entire pancreatic lesion, respectively. Pearson’s and Spearman’s correlation analysis were performed to evaluate the relationship between MRE parameters and tumor stage; grade; vascular and perineural invasion, regional lymphadenopathy and distant metastasis. Kaplan-Meier and Cox proportional hazards models were used to identify prognostic factors in patients with and without R0 resection (n=48 and 30, respectively).
Results or Findings: Fluidity in the peripheral part of the tumor was higher compared to the central part (1.23±0.25 vs. 0.96±0.20 rad, p＜0.001), whereas stiffness of the tumor didn’t show regional difference (p=0.064). In all patients, tumors with higher stiffness had more frequent regional lymphadenopathy (p=0.038), and tumors with higher fluidity in the peripheral part had more frequent vascular invasion, regional lymphadenopathy and distant metastasis (p=0.037, 0.007, and 0.027, respectively). Fluidity in the peripheral part of the tumor was positively correlated with tumor stage (rho=0.26; p=0.010). For patients with R0 resection, patients with higher fluidity in the peripheral part of the tumor (φ＞1.17 rad) had shorter disease-free survival (12.1 vs. 19.4 months, p=0.022).
Conclusion: MF-MRE may help to predict pancreatic cancer aggressiveness and patient survival.
Limitations: Not applicable.
Funding for this study: Funding was provided by National Natural Science Foundation of China (grant number: 82371950).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by Peking Union Medical College Hospital Institutional Review Board (I-24PJ2450).