RPS 1101 - Hot Topic: prognostication

Author Block: L. Yuan, J. Zhou; Lanzhou/CN
Purpose: To investigate the predictive value of clinical and pre- and post-treatment CT parameters for the efficacy of bevacizumab in patients with colorectal liver metastasis(CRLM) .
Methods or Background: This retrospective study included 290 patients with CRLM and 455 liver metastases treated with bevacizumab in our hospital . The morphological features of CRLM after four treatment cycles were assessed using CT images. CT values at baseline and after four treatment cycles in different phases (PS, AP, VP, and DP), and density change values (ΔCT values) before and after treatment were calculated. According to RECIST 1.1, CRLM after 12 treatment cycles was categorized into responsive and non-responsive groups. Differences between the two groups were searched for using the T-test, Mann−Whitney U, or chi-square test, and diagnostic performances of the different variables were evaluated using the receiver operating characteristic (ROC) curve.
Results or Findings: RAS mutant-type CRLM was more prevalent in the responsive group than in the wild-type group, with a significant difference (P<0.001). CRLM with a lobulated shape and heterogeneous texture after four treatment cycles showed a poorer treatment response than those with a round shape and homogeneous texture, with a significant difference (P<0.05). Additionally, CT values of AP-pre, VP-pre, DP-pre, VP-post, and ΔAP were higher in the responsive group than in the non-responsive group. The combined use of clinical and contrast-enhanced CT parameters demonstrated better efficacy (AUC >0.7) than single parameters. Parameters such as RAS, texture-post, and AP-pre combined showed higher predictive efficacy.
Conclusion: The combination of clinical parameters and contrast-enhanced CT parameters before and after treatment can effectively predict the early therapeutic response in patients with CRLM.
Limitations: It is single-center, small-sample, and retrospective nature may have affected the robustness of the results.
Funding for this study: This work was supported by grants of the National Natural Science Foundation of China (No. 82371914) and the Cuiying Scientific and Technological Innovation Program of Lanzhou University Second Hospital (CY2021-ZD-01).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the institutional ethical review board (2023A-379).

Author Block: M. V. GÖKAY, K. Kadirli, F. S. Bayramoğlu, H. Mutlu; Istanbul/TR
Purpose: To evaluate the association and prognostic value of CT-derived body composition parameters with gallbladder perforation in patients with acute cholecystitis.
Methods or Background: We retrospectively analyzed 80 surgically treated patients with acute cholecystitis (Jan 2018–Dec 2024). Gallbladder perforation was intraoperatively confirmed in 21 cases (26.2%), while 59 non-perforated patients, matched by age, sex, and comorbidity, served as controls. Preoperative non-contrast abdominal CTs were assessed at the L3 level. Skeletal muscles (psoas, paraspinal, abdominal wall) and visceral/subcutaneous fat were segmented semi-automatically using ImageJ v1.54m. Thresholds: –29 to +150 HU for muscle, –190 to –30 HU for fat. All areas were normalized to vertebral surface area. Measurements were repeated twice by two readers. Comorbidities were grouped as none, 1, 2, or ≥3. Collected variables included demographics, CRP, leukocytes, and time intervals (symptom–CT, symptom–admission, admission–surgery). Univariate tests identified predictors of perforation. Significant variables were entered into multivariate logistic regression (backward LR). ROC assessed diagnostic performance, and VIF tested collinearity. Variables with <20% missingness were handled via listwise deletion, leading to minor sample variations.
Results or Findings: Univariate analysis showed significantly higher IMAT/vertebra ratio in the perforation group (p=0.050). Multivariate analysis identified IMAT/vertebra ratio (OR: 4.63; 95% CI: 1.41–15.15; p=0.011) and leukocyte count (OR: 1.14; 95% CI: 1.02–1.29; p=0.026) as independent predictors. Symptom-to-CT interval showed borderline association (OR: 1.14; 95% CI: 1.00–1.29; p=0.054). Interobserver agreement was excellent (ICC = 0.89).
Conclusion: Our findings show that the IMAT/vertebra ratio is an independent predictor of gallbladder perforation in acute cholecystitis, reflecting inflammatory burden and reduced physiological reserve. Validation in larger multicenter cohorts is needed to confirm its prognostic value and support integration into routine preoperative assessment.
Limitations: Single-center, retrospective design and small sample size.
Funding for this study: None.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by Koşuyolu High Specialization Training and Research Hospital Ethics Committee for Scientific Research (Approval No: 2025/08/1129)

Author Block: A. Galluzzo, L. Scalzone, C. Mugnaini, G. Danti, V. Miele; Firenze/IT
Purpose: To develop two different radiomic models (RMs) based on preoperative portal phase computed
tomography (PP CT) to predict microsatellite instability (MSI) in patients with colorectal cancer
(CRC) before surgery.
Methods or Background: PP CT scans of 115 CC patients were segmented using 3DSlicer (v5.6.1). Model I included images
from three different scanners (GE, Siemens, Philips), while Model II used only one scanner (GE).
For Model I, 80 patients were used for training and 35 for internal validation; for Model II, 46 and
24 patients were used, respectively. Data on sex, age, tumor location, and MSI genomic status were
collected. Significant radiomic features (RFs) were identified using the t-test or Mann–Whitney test
(p<0.05), and the most robust RFs were selected using the LASSO regression method. Both
RMs were internally validated.
Results or Findings: Model I, based on 2 RFs and 1 clinical features (LOCATION) achieved an AUC of 0.76 (95% CI:
0.65–0.87) in the training cohort and 0.74 (95% CI: 0.56–0.92) in the validation cohort. Model II,
based on 3 RFs, achieved an AUC of 0.85 (95% CI: 0.73–0.96) in the training cohort and 0.72 (95%
CI: 0.50–0.94) in the validation cohort.
Conclusion: Both RMs performed well in distinguishing MSI from non-MSI tumors, potentially reducing the need for invasive histology and improving treatment timing. Despite a higher AUC, Model II showed overfitting compared to Model I, which included two RFs and one clinical feature (LOCATION). Developing models on larger, more diverse datasets is preferable to improve generalizability and limit overfitting.
Limitations: This study is limited by its retrospective design, small sample size, lack of external validation, and absence of follow-up. Future studies should use larger, prospective cohorts. Standardized imaging and multicentre trials are key for clinical implementation of radiomics.
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 conducted in accordance with the Declaration
of Helsinki, and approved by the Ethics Committee of Careggi University Hospital (protocol code
13261_OSS).

Author Block: B. Salam, M. Weber, A. M. Sprinkart, S. Nowak, M. Theis, A. Isaak, J. A. Luetkens, J. Vogelhuber, D. Kütting; Bonn/DE
Purpose: The spleen is an essential component of the immune system and closely associated with cardiac function (cardio-splenic axis). In this context, spleen volume (SV) emerges as a potential new prognostic indicator for cardiovascular risk. This study aimed to investigate the prognostic value of SV measured from routine preinterventional CT in patients undergoing MitraClip for treatment of recurrent mitral regurgitation.
Methods or Background: Consecutive patients undergoing MitraClip between February 2011 and November 2022 at the Heart Center Bonn were retrospectively evaluated. SV was determined from pre-interventional CT scans using a dedicated deep learning segmentation model for body composition analysis (TotalSegmentor).
Results or Findings: A total of 214 patients (mean age: 77.6±8.2 years, mean EuroSCORE II: 5.0±3.6%) were investigated. 1-year survivors had a significantly lower SV compared to non-survivors (0.19±0.12 cm3 vs. 0.35±0.35 cm3, P<0.001). According to their SV, patients were dichotomized by the median value and defined to have low (<0.182 cm3) and high SV (≥0.182 cm3), respectively. Following MitraClip, high SV was related to acute kidney injury (53.3% vs. 71.0%, P=0.020) as well as increased 30-day (2.8% vs. 12.1%, P=0.009), 1-year (10.3% vs. 27.1%, P=0.001), 2-year (15.0% vs. 33.6%; P=0.001), and 3-year mortality (15.9% vs. 34.6%; P=0.002). On multivariate Cox regression analysis, SV (Hazard Ratio 8.92 [95% Confidence Interval: 2.67-29.90]; P<0.001), as well as NT-proBNP (HR 1.00 [95% CI: 1.00-1.00]; P=0.015) were identified as independent predictors of 1-year mortality.
Conclusion: Our results indicate CT-derived SV as a promising new imaging biomarker, which provides additional information for risk stratification in MitraClip patients. Future studies should explore the clinical value of SV compared with other established risk stratification tools and the prognostic role of SV for other cardiovascular and oncologic diseases.
Limitations: Single-center retrospective design limits generalizability.
Funding for this study: N/A
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: S. Shi, J. Liu, Y. Luo, S-T. Feng, W. Deng; Guangzhou/CN
Purpose: To investigate the prognostic value of tumor mutation burden (TMB) in pancreatic ductal adenocarcinoma (PDAC) and the predictive utility of dual-layer spectral CT (DLCT) for TMB assessment.
Methods or Background: This retrospective study included patients with histologically confirmed PDAC who underwent DLCT from June 2019 to December 2023. The association between TMB and progression-free survival (PFS) was analyzed using survival analysis and the optimal cutoff was calculated to group patients. The Kaplan-Meier survival curves were used to compare the PFS between both groups. Independent TMB predictors were identified through univariate and LASSO regression. Predictive performance was assessed via ROC and precision-recall (PR) curve analyses.
Results or Findings: Among 75 enrolled patients (mean age 60.4 ± 11.2 years, 41 male), 52 received combination immunotherapy. Median TMB was 2.13 mutations per megabase (mut/Mb; interquartile range: 1.00–4.26). An integer TMB cutoff of 5 mut/Mb (optimized from 5.03 for clinical practicality) stratified patients into distinct prognostic groups, with low-TMB cases demonstrating inferior PFS (median PFS: 5 vs. 7 months, p=0.02); this trend persisted in the immunotherapy subgroup (4 vs. 7 months, p=0.02). Normalized iodine concentration in the pancreatic phase (nICₐ) emerged as the sole independent TMB predictor (area under the curve [AUC] of ROC=0.901, cutoff=0.089; accuracy=0.893, sensitivity=0.818, specificity=0.906), surpassing normalized conventional CT attenuation values in the pancreatic phase (nCTₐ, AUC=0.834), pancreatic tumor infiltration (AUC=0.679), and a combined model (nCTa+PTI, AUC=0.864). PR curves confirmed nICₐ’s superior efficacy. Patients stratified by nICₐ-predicted TMB status exhibited significant PFS differences (7 vs. 5 months, p=0.04).
Conclusion: Low TMB is a negative prognostic biomarker in PDAC, associated with shorter PFS. DLCT-derived nICₐ enables accurate, noninvasive TMB prediction, supporting its potential role in therapeutic stratification.
Limitations: The single-center retrospective design with a limited sample size，a single vendor platform.
Funding for this study: This study has received funding by National Natural Science Foundation of China (82472096) and Natural Science Foundation of Guangdong Province (2024A1515011968) to Yanji Luo.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study design was approved by the ethics review board of our hospital and informed consent was waived owing to the retrospective nature of the study.

Author Block: F. Herr, M. Vornhülz, C. A. Dascalescu, A. E. Sint, J. Ricke, M. Ingrisch, L. K. Shiyam Sundar, J. Mayerle, S. Grosu; Munich/DE
Purpose: To evaluate the prognostic relevance of longitudinal changes in body composition derived from routine CT scans in patients with pancreatic ductal adenocarcinoma (PDAC) using fully automated image analysis.
Methods or Background: In this retrospective single-center study, 55 patients with histologically confirmed PDAC from the prospective METAPAC trial were included. All had at least one venous-phase abdominal CT within 90 days from baseline. Fully automated segmentation was used to quantify volumetric indices for visceral fat (VF), subcutaneous fat (SF), muscle (M), and compute ratios (VF/SF, SF/M). Relative changes from baseline were calculated. Optimal cutoffs were determined using maximally selected rank statistics. Associations with overall survival (OS) were analyzed via Cox regression and evaluated by Harrell’s C-index and AIC.
Results or Findings: A total of 88 CT scans from 55 patients (mean age 67 ± 11 years; 28 male) were analyzed. An increase in subcutaneous fat index (SFI) (HR: 5.17; C-index: 0.60; p = .028) and a decrease in VF/SF ratio (HR: 2.70; C-index: 0.67; p = .014) were significantly associated with reduced OS. Both metrics remained prognostic in multivariable analysis (SFI: HR: 4.67; p = .0398; VF/SF: HR: 2.75; p = .0199; C-index: 0.73). Absolute baseline or follow-up values were not predictive. Kaplan–Meier analysis confirmed shorter OS in patients with SFI > 0.77 (median OS 319 days vs. not reached; p = .015) and VF/SF ≤ 0.97 (median OS 131 vs. 809 days; p = .011).
Conclusion: In PDAC patients, dynamic fat distribution changes, particularly rising SFI and falling VF/SF ratios, were associated with poorer survival. Automated body composition analysis from routine CT may serve as a non-invasive imaging biomarker for early risk stratification.
Limitations: Key limitations of our study include the small sample size and retrospective, single-center design.
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: This single-center retrospective study was approved by the local ethics committee (approval number 19-027).

Author Block: A. Suri¹, M. A. B. C. Rockenbach², B. C. Bizzo², J. Garrett³, D. C. Elton², P. J. Pickhardt³; ¹Los Angeles, CA/US, ²Boston, MA/US, ³Madison, WI/US
Purpose: To investigate whether abdominal CT biomarkers explored in prior literature remain predictive of 5 and 10 year all-cause mortality in a novel, large cohort.
Methods or Background: 180,935 CT patients (with and without contrast, imaging the abdomen, one scan per patient, mean avg 56.6±17.4, 55.7% female) from our multi-hospital institution were selected for usage in this study. Automatically extractable imaging biomarkers (derived from muscle attenuation, bone mineral density, visceral & subcutaneous adipose tissue volume, and abdominal aorta Agatston score at the L3 vertebral body) were calculated on all scans. Date of death of each patient was also gathered. 80% of scans (144,748) were used to train/validate 14 different algorithms to predict 5 or 10 year all-cause mortality. The best algorithm on the training set (based on AUC over 5-fold cross-validation) was evaluated on the hold out test set.
Results or Findings: The best performing algorithm for 5-year mortality prediction was XGBoost which achieved an AUC of 0.84 on the hold out test set. The algorithm did not vary significantly in AUC for age ≤65=0.82 vs >65=0.85. The best performing algorithm for 10-year mortality prediction was Light Gradient Boosting Machines (AUC=0.79) with minimal variation according to age (AUC≤65=0.77 vs AUC>65=0.81). For both 5 and 10 year mortality prediction, the most important values relied on visceral adipose tissue attenuation followed by visceral:subcutaneous adipose tissue volume ratio. The least important feature was abdominal aorta Agatston score.
Conclusion: Automated abdominal CT biomarkers are strong predictors of all-cause mortality even in large, multi-hospital settings. These imaging biomarkers remain predictive regardless of clinical demographics such as age.
Limitations: Follow up data on patients is limited by their interaction with our institution. As a result, information on the death of patients may be incomplete.
Funding for this study: No funding sources to declare.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: A. Buz, K. N. Kılıçarslan; Bolu/TR
Purpose: This study aims to evaluate the feasibility of signal intensity–based apparent extracellular volume derived from dynamic liver MRI as a noninvasive and accessible quantitative method to improve the diagnostic reliability of liver fibrosis assessment.
Methods or Background: A total of 61 adult patients with chronic liver parenchymal disease who underwent contrast-enhanced dynamic liver MRI between July 2021 and August 2025 were included in the study. In the dynamic MRI, signal intensity measurements of the aorta and liver parenchyma from both lobes were obtained in the pre-contrast and late hepatic phases LAVA series. Patients were divided into three groups based on their FIB-4 index to explore the correlation with liver ECV. All data analyses were performed using R version 4.4.1.
Results or Findings: A total of 61 participants were included in the study, comprising 34 males and 27 females. The median age was 65 years (IQR: 12; range: 26–82 years). The median ECV was 0.291 (IQR: 0.104) in the low-risk group (n = 6), 0.316 (IQR: 0.129) in the intermediate-risk group (n = 19), and 0.283 (IQR: 0.101) in the high-risk group (n = 36). Overall, no clear trend of increasing or decreasing ECV with higher FIB-4 value was observed.
Conclusion: Signal intensity–derived liver ECV is a feasible, noninvasive measure, but it did not consistently correlate with FIB-4–based fibrosis risk in this cohort. Incorporating T1 mapping may improve its accuracy, and further studies are needed to determine its predictive value for liver fibrosis.
Limitations: Single center-retrospective study design.
Funding for this study: N/A
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by Bolu Abant İzzet Baysal University Non-Interventional Clinical Research Ethics Committee.