RPS 2401 - Acute abdominal diseases and imaging of the bowel

Author Block: B. J. Choi, W. Chang, J. H. Hwang, J. Cho, Y. J. Lee, Y. H. Kim, S. H. Park, J. Y. Choi; Seongnam-si/KR
Purpose: To demonstrate that 0.5-mSv ultra-low dose CT using deep learning–based denoising algorithm (DLA) has non-inferiority in diagnosing acute appendicitis, compared to conventional 2.0-mSv low dose CT.
Methods or Background: We used 2.0-mSv CT images of 30 patients with suspected appendicitis from the prior prospective study. The original 2.0-mSv CT were reconstructed using iterative model reconstruction (IMR). We simulated 0.5-mSv CT images from the original 2.0-mSv CT. Then we applied IMR and DLA, resulting in three CT image groups per patient (IMR 2.0-mSv, IMR 0.5-mSv, and DLA 0.5-mSv groups). Six radiologists (three abdominal and three non-abdominal radiologists) rated the likelihood of appendicitis on a five-point Likert scale. Primary end point was comparison of the pooled area under the receiver operating characteristic curve (AUC) between DLA 0.5-mSv CT and IMR 2.0-mSv CT, with a non-inferiority margin of 0.06. Secondary end points included comparison of AUC between DLA 0.5-mSv CT and IMR 0.5-mSv CT, diagnostic sensitivity/specificity.
Results or Findings: The AUC of DLA 0.5-mSv CT was non-inferior to that of IMR 2.0-mSv CT [AUC difference: 0.003 (95% CI: -0.011, 0.017)]. The AUC of DLA 0.5-mSv CT was slightly higher among the non-abdominal radiologists compared to IMR 0.5-mSv CT [AUC difference: 0.034 (95% CI: -0.130, 0.198)]. Diagnostic sensitivity/specificity were 100% (9/9)/95% (20/21) for all readers with both DLA 0.5-mSv CT and IMR 2.0-mSv CT. However, the sensitivities of two non-abdominal radiologists with IMR 0.5-mSv CT were mildly compromised [78% (7/9) and 89% (8/9), respectively].
Conclusion: 0.5-mSv ultra-low dose CT using DLA was non-inferior to conventional 2.0-mSv low dose CT using IMR in diagnosing acute appendicitis.
Limitations: The limitations of the study are 1) 0.5-mSv CT images were simulations, not real data and 2) small sample size, collected from a single tertiary hospital.
Funding for this study: Funding was provided by the National Research Foundation of Korea(NRF) grant funded by the Korea government(MSIT) (NRF-2022R1F1A1072570).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The institutional review board approved this study, and the requirement for informed consent was waived (B-2401-879-112).

Author Block: T. Yoshikawa, H. Nagata, T. Ueda, M. Nomura, D. Takenaka, Y. Ozawa, Y. Ohno; Toyoake/JP
Purpose: Upright CT is recently and clinically set in our institution, and the purpose of this study was to directly compare capabilities for diagnosis and subtype classification of inguinal herniation between upright CT (uCT) and conventional supine CT (sCT).
Methods or Background: 258 consecutive patients who suspected inguinal hernia underwent sCT and uCT within a week, surgical treatment or follow-up examination. From this cohort, 120 inguinal hernias and computationally selected 120 out of 396 non-inguinal hernia were visually assessed by two board certified general and abdominal radiologists by 5-point scales as well as subtypes of hernia. Inter-observer agreements for probability of hernia and subtype were assessed by kappa statistics with χ2 test. Then, ROC analysis was performed to compare diagnostic performance between two CTs. Then, sensitivity (SE), specificity (SP) and accuracy (AC) were compared each other by McNemar’s test. Moreover, subtype classification accuracy (SAC) was also compared between uCT and sCT by McNemar’s test.
Results or Findings: Inter-observer agreement for probability of hernia were determined as significant and almost perfect on uCT (κ=0.84, p<0.0001) and substantial on sCT (κ=0.77, p<0.0001), and that for subtype classification were also significant and almost perfect on both CTs (uCT: κ=0.83, p<0.0001; sCT: κ=0.81, p<0.0001). Area under the curve (AUC) of uCT (AUC=0.99) were significantly larger than that of sCT (AUC=0.97, p<0.05). SE and AC of uCT (SE=92.5%, AC=96.3%) were significantly higher than those of sCT (SE: 80.8%, p<0.0001; AC: 90.4%, p<0.0001). SAC of uCT (87.5%) was significantly higher than that of sCT (73.3%, p<0.0001).
Conclusion: Upright CT has better diagnostic performance for inguinal hernia and subtype classification than conventional supine CT in routine clinical practice.
Limitations: Lack of clinical outcome evaluation
Funding for this study: Research grant from Canon Medical Systems Corporation
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Fujita Health University Hospital

Author Block: M. Oberparleiter, C. Breit, J. Vosshenrich, P. Hehenkamp, A. C. Seifert, A. Kobe, C. J. Zech, M. Obmann; Basel/CH
Purpose: Current guidelines suggest replacing unenhanced images with DECT-derived virtual non-contrast images (VNC) in suspected upper GI bleeding based on only two clinical studies.

Our study compares diagnostic accuracy, reader confidence, and reading time of a conventional triphasic versus a dual-energy CT protocol in patients with upper and lower GI bleeding.
Methods or Background: This retrospective study included 52 patients with active GI bleeding (22 upper, 30 lower) and 52 controls who underwent non-contrast, arterial, and portal-venous phase abdominal CT. For each case, a triphasic conventional CT dataset and a DECT dataset with VNC, iodine images, and arterial and portal venous phase images were created. Two residents and two fellowship-trained abdominal radiologists evaluated all cases for active GI bleeding. Radiation dose and reading time were recorded. Diagnostic confidence was rated on a 5-point Likert scale. Inter-reader agreement was assessed using Fleiss' kappa. Sensitivity and specificity were compared using McNemar's test, reading time, and reader confidence with the Wilcoxon signed-rank test.
Results or Findings: Inter-reader agreement was substantial (𝜅=0.80). Sensitivity and specificity for detecting GI bleeding using conventional CT did not differ from DECT (91% and 95%, vs. 93% and 96%, p=0.30 and p=0.77, respectively). Subgroup analysis of lower GI bleeding showed a sensitivity of 88% in conventional CT versus 93% in DECT (p=0.18). Diagnostic confidence increased from 4(IQR, 4-5) to 5(IQR, 4-5) when using DECT (p<0.01). Mean reading time per case was 102 s for both datasets (p=0.62). Total DLP without true unenhanced images was 21% lower.
Conclusion: DECT-derived VNC and iodine images can replace true non-contrast images when searching for GI-bleeding. Guidelines should be extended to include lower-GI-bleeding.
Limitations: Sample size was moderate, the study had a single-center design, and only dual-source and split-beam DECT scanners were used.
Funding for this study: This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The need for informed consent was waived due to the retrospective nature of this study.

Author Block: A. W. Marka, M. Graf, S. Ziegelmayer, M. R. Makowski, A. Sauter, T. Huber; Munich/DE
Purpose: To investigate the relationship between inflammatory laboratory markers and quantitative CT-derived imaging features in patients with acute pyelonephritis (APN).
Methods or Background: In this single-center retrospective study, we evaluated patients with clinical symptoms of APN at our institution from December 2018 to April 2024. Inclusion criteria comprised APN symptoms, elevated inflammatory markers (CRP and/or WBC), and CT morphologic signs of APN. Exclusion criteria included concomitant acute pathology and poor image quality. A total of 102 patients (mean age 60.1±18.8) were initially identified; 14 were excluded due to acute pathology and 5 due to poor image quality, leaving 83 for final analysis. CT scans followed a standardized protocol, and two radiologists blinded to clinical and lab data conducted image analysis. Inflammatory markers were collected on the scan day. Statistical analyses included Spearman correlation, Mann-Whitney-U tests, and linear regression.
Results or Findings: Spearman correlation analysis revealed strong positive correlations between CRP levels and both total volume (r=0.76, p<0.001) and percentage (r=0.71, p<0.001) of renal perfusion deficit. Multivariate linear regression showed total perfusion deficit volume explained 54.5% of CRP variability (p<0.001). WBC count also correlated significantly with total volume (r=0.379, p<0.001) and percentage (r=0.374, p<0.001) of perfusion deficit. Procalcitonin levels moderately correlated with fat stranding area (r=0.482, p=0.0014) but not other CT features. Locoregional lymphadenopathy was significantly associated with elevated CRP and WBC counts, but not procalcitonin levels.
Conclusion: Quantitative CT-derived features, particularly renal perfusion deficits, are significantly associated with inflammatory markers in APN. These findings suggest CT imaging can serve as a surrogate for inflammation severity, potentially guiding clinical management. Further research is needed to explore the clinical implications of these associations.
Limitations: -Small cohort
-Scarcity of procalcitonin levels
-No correlation with duration of a patient's hospitalization
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Data collection, processing, and analysis were approved by the institutional review board (protocol number 180/17S), and informed consent was waived.

Author Block: J. Xu, L. Zhu, W. Liu, Y. Lu, J. Liu, C. Ma, Y. Zhang, X. Wang, F. Feng; Beijing/CN
Purpose: To evaluate the image quality and diagnostic performance of single-shot fast spin-echo T2 weighted imaging with deep learning reconstruction (SSFSE-DL) in volunteers and patients with acute abdomen, in comparison to SSFSE without deep learning reconstruction (SSFSE-nonDL) and conventional PROPELLER sequences.
Methods or Background: Thirty-five healthy volunteers, as well as 35 patients with acute abdominal pain from emergency room were prospectively enrolled. Abdominopelvic MRI at 3T was performed using three T2-weighted imaging sequences: SSFSE-DL (acquisition time: 34s), SSFSE-nonDL, and conventional PROPELLER (acquisition time: 2-3min), in random order. Two blinded radiologists independently evaluated the overall image quality, noise, motion artifacts and clarity of major abdominopelvic organs. Diagnostic confidence for the presence or absence of common abdominopelvic diseases was rated on a 1-5 Likert scale. Signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR) as well as image noise for the liver, pancreas and spleen were also quantified. Intra- and inter- observer agreement were assessed, and comparisons of image quality and diagnostic performance between the three sequences were made.
Results or Findings: Intra- and inter- observer agreement for the qualitative analysis and diagnostic performance were good to excellent (0.776-0.967). SSFSE-DL yielded significant higher SNR and CNR, and lower noise than SSFSE-nonDL and PROPELLER in both volunteers and patients (all P<0.05). SSFSE-DL obtained significantly higher image quality and lower noise than SSFSE-nonDL and PROPELLER (both P<0.05). SSFSE-DL and SSFSE-nonDL had significantly lower motion artifacts and better clarity of major abdominopelvic organs than PROPELLER (both P<0.05). The AUC for detecting common abdominopelvic diseases in SSFSE-DL (0.977-1) and SSFSE-nonDL (0.887-1) were significantly higher than PROPELLER (0.585-0.953).
Conclusion: SSFSE-DL achieved superior image quality and diagnosis performance for volunteers and patients with acute abdomen.
Limitations: The number of patients with positive diagnosis for each specific disease was relatively small.
Funding for this study: This study was funded by the National Natural Science Foundation of China (82371950)
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This prospective single-center study was approved by the local institutional review board, and informed consent was obtained from all participants prior to inclusion in the study.

Author Block: X. Zhang¹, S. Zhong², G. Zhang², X. Zhou¹; ¹Chongqing/CN, ²Shanghai/CN
Purpose: To characterize the diagnostic performance of low-dose (LD) abdominal CT combined with artificial intelligence iterative reconstruction (AIIR) for assessing acute pancreatitis based on CT severity index (CTSI).
Methods or Background: A total of 30 patients with acute pancreatitis who underwent follow-up CT examination were prospectively enrolled. All patients underwent standard-dose (SD) CT followed by LD-CT in the same breath hold, where an immediate LD-scan was added in the portal venous phase. The SD-protocol was 120 kVp, ref. 141 mAs, and hybrid iterative reconstruction (HIR), whereas the LD-protocol was 120kVp, ref. 50mAs and AIIR reconstruction. To obtain the CTSI, SD- and LD-CT images at portal venous phase were independently scored by two radiologists for assessing pancreatic inflammation, necrosis, and extrapancreatic complications. Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of healthy and inflamed parenchyma were measured and calculated.
Results or Findings: Compared to SD-CT, LD-CT examination achieved a 65.18% reduction in effective radiation dose for the portal venous phase (6.06 ± 1.28 mSv vs. 2.11 ± 0.45 mSv, p < 0.05). Based on CTSI scoring, LD-AIIR was found comparable to SD-HIR in evaluating the severity of acute pancreatitis (5.02 ± 1.42 vs. 5.11 ± 1.78, p = 0.86). Inter-observer agreement for assessing the severity of acute pancreatitis was excellent (k = 0.89). LD-AIIR showed superior conspicuity compared to SD-HIR for both the inflamed (SNR: 1.96 ± 0.83 vs. 1.74 ± 0.98; CNR: 4.17 ± 1.58 vs. 2.11 ± 0.87; both p < 0.05) and the healthy parenchyma (SNR: 8.03 ± 1.71 vs. 4.91 ± 1.04; CNR: 3.06 ± 1.84 vs. 1.78 ± 1.03; both p < 0.05).
Conclusion: AIIR allows for significant radiation dose reduction without compromising image quality or diagnostic performance for the evaluation of acute pancreatitis.
Limitations: n/a
Funding for this study: n/a
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethics committee of Chongqing University Jiangjin Hospital

Author Block: M. Zerunian¹, S. Nardacci¹, N. Petrucciani¹, I. Toniolo², D. De Santis¹, D. Caruso¹, C. G. Fontanella², G. Silecchia¹, A. Laghi¹; ¹Rome/IT, ²Padova/IT
Purpose: Total small bowel length(TSBL) is crucial to achieve successful metabolic/bariatric surgery. A non-invasive measurement of the TSBL will impact on surgical strategy to avoid short-bowel syndrome after surgery. Cross-sectional imaging(CSI) might play an important role by measuring TSBL non-invasively. We aimed to set up a reliable AI-based automatic method using preoperative CSI to measure the TSBL in candidate to bariatric/metabolic surgery.
Methods or Background: This multicentre prospective TRIAL included patients eligible for bariatric surgery(BMI >35 kg/m2 and at least one obesity-related comorbidity,BMI>40 kg/m2) underwent the same day MRI-and CT-enterography before bariatric surgery. TSLB assessed right before the surgery at the operation table and, TSLB <250 cm considered the cut-off as risk for developing short-bowel syndrome. TSLB obtained on MRI and CT by manual segmentation(Slicer3D). A Convolutional Neural Network with U-NET developed, consisting in a contracting path followed by an expansive path, RELU activation and a softmax activation function to reduce the feature map. The Adam optimization algorithm with a constant learning rate was used for the learning process. Training stopped after 100 epochs. DICE coefficient were calculated to quantify the accuracy of the prediction compared to annotated-by-radiologist images.
Results or Findings: Fifty patients enrolled (27 female,age range 26-54 years old,mean BMI 41.47).
Patients underwent surgery showed a TSBL mean measured intraoperatively of 652.8587.37cm. CSI showed good concordance with TSBL mean of 589.0882.95cm(k= 0.70).DICE coefficient of the training set showed a DICE score ranging between 35%-45%, confirmed in the validation set. All methods correctly categorized the patients according to the cut-off considered as risk factor to develop short-bowel syndrome.
Conclusion: Automatic AI-based segmentation of small bowel on non-invasive CSI might be a useful tool to assess obese patients to personalize the treatment and reduce complications.
Limitations: Small sample size
Funding for this study: Italian Ministry of University and Research (MUR) Research Projects of Significant National Interest – PRIN (ID: MUR 2022MPAE29_003)
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Multicenter interventional study

Author Block: A. Stanzione, V. Arpaia, A. E. Antonini, R. Liuzzi, L. Sommella, L. Mannacio, A. Brunetti, L. Camera; Naples/IT
Purpose: To evaluate the effects of a fixed injection duration (FID) on vascular enhancement in a Single-Pass (SP) abdominal CT performed in patients with nontraumatic acute abdomen (ANTA).
Methods or Background: 100 patients (58M, 42F; aged 52±20 yrs ) with ANTA underwent a SP contrast-enhanced CT (Somatom Drive, Siemens) performed using a Single Source at either 80 kVp (Group A; BMI 19±3), 100 Kvp (Group B; BMI 25±4) or 120 Kvp (Group C; BMI 30±2.5) . In all groups a non-ionic iodinated contrast media (370 mgI/ml) was administered as follows: Group A (0.37 grI/Kg); Group B (0.52 grI/Kg); Group C (0.63 grI/Kg). All patients underwent a SP protocol with a FID (50 sec) and a tailored scan delay (SD). In all patients Signal- (SNR) and Contrast-to-Noise Ratios (CNR) were calculated for the abdominal aorta (AA) and the main portal vein (MPV) using the psoas muscles as reference tissue. Statistical analysis was performed with ANOVA (p < 0.05).
Results or Findings: No significant differences were observed in the demographics of either Group A (20M/12F; 47±20 yrs) or B (36M/22F; 54±19 yrs) whereas a female preponderance was observed in Group C (2M/8F; 50±18 yrs). Despite significant differences (p < 0.001) were observed in both the volumes (53±11 vs 103±17 vs 141±19 ml) as well as the injection rates (1.1±0.2 vs 2.1±0.4 vs 2.8±0.4 ml/sec) of the contrast media for Group A, B and C, respectively, SNR and CNR were not significantly different for both AA and MPV.
Conclusion: SP performed with a FID results in a consistent vascular enhancement.
Limitations: Unbalanced sample size
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Local IRB approval