Research Presentation Session
06:15S. Skornitzke, Heidelberg / DE
Purpose:
Virtual monoenergetic images (VMI) from dual-energy computed tomography (DECT) can improve image quality in contrast-enhanced abdominal CT imaging. A translation to CT perfusion could increase measurement accuracy by improving the delineation of changes in contrast agent concentrations.
Methods and materials:Dual-source DECT scans were acquired in 17 patients with pancreatic carcinomas with a DECT perfusion protocol consisting of 34 acquisitions at 80kVp/Sn140kVp over 51 seconds. VMIs were calculated for energy levels between 40keV and 150keV in 5keV steps (Monoenergetic+, Siemens Healthineers). Perfusion maps of blood flow were calculated using the maximum-slope model for VMIs and polyenergetic 80kVp images. Perfusion maps were quantitatively evaluated in terms of blood flow in healthy tissue and the carcinoma, the contrast between tissue types, noise, and contrast-to-noise ratio (CNR), comparing VMIs to 80kVp images.
Results:The measured blood flow increased with energy level from 114±37ml/100ml/min (healthy tissue) and 46±25ml/100ml/min (carcinoma) for 40keV to 129±59ml/100ml/min (healthy tissue) and 76±50ml/100ml/min (carcinoma) for 150keV, compared to 114±37ml/100ml/min (healthy tissue) and 47±27ml/100ml/min (carcinoma) for 80kVp images. The differences between tissue types were statistically significant for all VMIs and 80kVp images (p<0.05). The measured blood flow was not significantly different between VMIs and 80kVp images for energies below 110keV. Noise in perfusion maps was reduced for low-energy VMIs (40keV:19ml/100ml/min; 80kVp:23ml/100ml/min) and CNR was higher (40keV:3.7±1.9; 80kVp:3.2±1.6) compared to 80kVp images. The differences in noise and CNR were statistically significant for all energy levels except those between 60keV and 75keV.
Conclusion:CT perfusion maps calculated from low-energy VMIs show improved quantitative image quality compared to conventional polyenergetic 80kVp images.
Limitations:Further evaluation of diagnostic image quality and patient radiation exposure seems necessary.
Ethics committee approvalInstitutional review board approval and written informed consent were obtained.
Funding:No funding was received for this work.
06:07E. Salinas-Miranda, Toronto / CA
Purpose:
To assess the prognostic value of body composition biomarkers assessed by CT in pancreatic cancer patients (PC) undergoing FOLFIRINOX chemotherapy.
Methods and materials:We analysed the pre-treatment and first follow-up CT of 64 PC patients enrolled in a trial receiving FOLFIRINOX. Skeletal muscle, intraabdominal fat, subcutaneous fat, and intramuscular fat were measured as normalised cross-sectional areas (cm2/m2) at the L3-level. SliceOmatic software v5.0 was used. We calculated the rate of muscle loss (RML) over a follow-up time of 90 days. The associations between CT body composition measurements and clinical variables with time to progression (TTP) and overall survival (OS) were determined using a Cox proportional hazard analysis. Statistically significant variables in univariate analysis were included for multivariate analysis.
Results:In the multivariate analysis, the rate of muscular loss was significantly associated with OS and TTP, with a hazard ratio (HR) of 2.61 (CI: 1.37-4.95, p=0.003) and 1.92 (CI: 1.17-3.15, p=0.009), respectively. The optimal cut-point of RML was 8.18 cm2/m2 per 90 days for OS (HR:2.21, CI: 1.23-3.98, p<0.007) and 6.02 cm2/m2 per 90 days for TTP (HR:2.95, CI: 1.5-5.46, p<0.001). None of the remaining body composition measurements was associated with OS or TTP in multivariate analysis. RML remained significant in the prognostication for OS while the RECIST status at the first follow-up did not.
Conclusion:In patients with locally advanced and metastatic PC, the high RML between the first and second CT is associated with worse survival and earlier tumour progression.
The RML from baseline to the first follow-up CT in patients receiving FOLFIRINOX may add value to response prediction in PC and help with therapy modulation or nutritional intervention. Further prospective validation is required.
Limitations:A retrospective study with a low sample size.
Ethics committee approvalIRB approved the study.
Funding:Ontario Institute for Cancer Research and the Translational Research Initiative in Pancreatic Cancer and Clinician Investigator Program was granted to Masoom Haider.
05:32Jia Fu, Beijing / CN
Purpose:
To investigate whether DKI parameters are associated with overall survival (OS) in patienta with advanced gastric cancer after neoadjuvant chemotherapy.
Methods and materials:From 2014-2016, 46 consecutive patients with a biopsy-proven gastric cancer who underwent pre-neoadjuvant chemotherapy MRI with DKI (b=0, 200, 500, 800, 1,000, 1,500, and 2,000s/mm2) were enrolled in this retrospective study. The region of interest was manually drawn on the specific slice showing the largest area of the tumour. Three parameters of DKI were calculated automatically: apparent diffusion coefficient (ADC, b=800s/mm2), kurtosis (K), and diffusivity (D) value. Cox proportional hazards analyses were performed to determine the relationship between DKI parameters, neoadjuvant pathologic T-stages, tumour location, histologic subtype, and OS.
Results:The median follow-up was 37.7 months and 11 patients had died at a median time of 16.6 months. There was no significant difference in pre-treatment DKI parameters between the histologic subtypes. The ADC and K values were associated with a negative prognosis (p=0.048, p<0.001). Cox analysis showed that the K value ([HR], 1.011; 95% [CI]: 1.005, 1.018; p<0.001) was the representative predictor of OS. The optimum threshold point was 0.618.
Conclusion:DKI parameters can potentially provide prognostic information complementary to ADC for gastric cancer patients after neoadjuvant chemotherapy.
Limitations:The relatively small population in our study might result in bias in the results. We only explored the correlation between pre-neoadjuvant chemotherapy (pre-NACT) DKI parameters and overall survival, not all prognostic factors such as DKI parameters of post-NACT or PFS.
Ethics committee approvalThis study was approved by our local institute review board.
Funding:No funding was received for this work.
06:31E. Ortiz, Cali / CO
Purpose:
To investigate the significance of T2-weighted-based radiomic data extracted from baseline rectal MRI combined with qualitatively assessed imaging features in predicting the clinical complete response after neoadjuvant CRT for rectal cancer.
Methods and materials:A retrospective study of patients with locally advanced rectal cancer who underwent rectal MRI before neoadjuvant CRT from October 2011–January 2015 followed by surgery was conducted. The reference standard for the pathologic complete response was the surgical histopathologic specimen after CRT. Qualitatively assessed imaging features were extracted from our institutional standardised radiology report. In radiomics analysis, one radiologist manually segmented the primary tumour on T2-weighted images for 102 patients to be used as the training set; two different radiologists independently segmented 66/102 patients to be used as the validation set. Using CERR software, 108 radiomics features were extracted. The Wilcoxon test was used to identify scanner-independent features. Subsequently, a least absolute shrinkage operator analysis was used to extract a radiomics score. A support vector machine model combining the radiomics score and qualitatively assessed clinical descriptors was compared against both clinical and radiomics-only models using the deLong test.
Results:102 patients were included. The radiomics score produced an AUC of 0.75. Comparable results using the validation set were AUC=0.69 and 0.71 for each radiologist, respectively. The SVM model using the qualitative features had an accuracy of 67% (sensitivity 42%, specificity 72%); when adding the radiomics score, the accuracy increased to 74% (sensitivity 58%, specificity 77%).
Conclusion:The prediction of the clinical response to stratify patients before neoadjuvant therapy may be obtained by combining a radiomics-based score and qualitatively assessed features from pre-treatment rectal MRI.
Limitations:The manual segmentation, radiomics analyses based on T2-weighed images only, and imbalanced dataset.
Ethics committee approvalInstitutional IRB approval.
Funding:Funding in part through the NIH/NCI Cancer Center Support Grant P30CA008748 and the Colorectal Cancer Research Center CC50367 at MSKCC.
06:25F. Landolfi, Rome / IT
Purpose:
A retrospective study with a small population sample.
Ethics committee approvaln/a
Funding:No funding was received for this work.
05:52D. Castiglione, Palermo / IT
Purpose:
To test whether a standardised structured radiology report improves the quality of preoperative CT staging of pancreatic ductal adenocarcinoma (PDA) compared to conventional free-text reports.
Methods and materials:In this retrospective, single-centre study, we included 27 patients (mean age, 64±11.1 years) with pathologically proven PDA operated on between 2015-2018 and imaged with pancreatic CT. 4 readers independently reported CT scans with both free-text and a structured report. The differences in reported morphologic and vascular features were assessed through a McNemar Test. The intra-reader and inter-reader agreements were calculated.
Results:216 (108 conventional and 108 structured) reports were analysed. Encasement of the left gastric artery, gastroduodenal artery, and splenic artery was described in up to 14.8% using free-text reports and up to 29.6% using a structured report, resulting in low intra-reader agreement (k=0.033-0.216). Inter-reader agreement improved with a structured report compared to free-text reports for the left gastric artery (ICC=0.844 vs ICC=0.493, respectively), gastroduodenal artery (ICC=0.730 vs ICC=0.449, respectively), portal vein (ICC=0.847 vs ICC=0.638, respectively), portal confluence (ICC=0.848 vs ICC=0.422, respectively), superior mesenteric vein (ICC=0.765 vs ICC=0.695, respectively), and the splenic vein (ICC=0.921 vs ICC=0.841, respectively).
Conclusion:The use of structured reports improves the quality of preoperative CT staging of pancreatic adenocarcinoma by significantly reducing the number of missed morphological and vascular features of PDA. Moreover, structured reporting results in a significant improvement of inter-reader agreement compared to free-text reports.
Limitations:The limited number of patients, retrospective nature, potential to miss unexpected findings not included in the template, potential testing or learning effect, and lack of time constraints for reporting.
Ethics committee approvalIRB approved. Waiver of informed consent obtained.
Funding:No funding was received for this work.
05:37R. Mathy, Heidelberg / DE
Purpose:
Early diagnosis of local pancreatic cancer recurrence remains challenging. We quantitatively evaluated the diagnostic potential of dual-energy (DE) contrast-enhanced CT for distinguishing recurrent carcinoma from the unspecific postoperative soft-tissue formation (PSF).
Methods and materials:After potentially curative pancreatic carcinoma resection, 36 consecutive patients with PSF were examined with DECT, acquiring 34 images (80kVp/140kVp) every 1.5s starting 13s after the bolus injection (80ml contrast agent, 370mg/ml, flow rate 5ml/s). Corresponding time points of arterial, pancreatic, and early venous phase (delay 10s, 19s, and 30s) were calculated from bolus trigger times in prior conventional CTs. Iodine and 120kVp-equivalent (M0.5) images were calculated. Regions of interest were placed in each soft-tissue formation. The diagnosis of local recurrence was confirmed by regular follow-up or by histological study.
Results:The final diagnosis was a local recurrence in 23 patients and unspecific PSF in 13 patients. The early venous average iodine concentration in recurrent carcinomas was significantly higher than in unspecific PSF (1.4mg/ml vs 0.9mg/ml, p=0.02). Arterial and pancreatic phase iodine concentrations in recurrent carcinomas were higher than in unspecific PSF, but not significant (1.2mg/ml vs 1.0mg/ml, p=0.24; 1.3mg/ml vs. 1.2mg/ml, p=0.48, respectively). Early venous density of recurrent carcinoma in 120kVp-equivalent images was significantly higher (70HU vs 48HU, p=0.008). The ROC-curve analysis for early venous iodine concentrations (AUC=0.76) suggests cut-off values >=1.5mg/ml for local recurrence (specificity 0.92, sensitivity 0.57) and <=0.5mg/ml for unspecific PSF (specificity 0.96, sensitivity 0.45).
Conclusion:In difficult cases, measuring iodine concentrations or density in PSF in (early) venous phase DECT can be a valuable additional parameter for differentiating local recurrence from unspecific PSF.
Limitations:The initial purpose of this study was evaluating CT-perfusion, so the protocols were not optimised for evaluating DECT-data.
Ethics committee approvalInstitutional review board approval and written informed consent were obtained.
Funding:Supported by BMBF-grant 031L0163.
04:07V. Tikhonova, Moscow / RU
Purpose:
Attenuation characteristics of the pancreatic ductal adenocarcinoma (PDAC) and contrast material uptake vary depending on the histological grade and pancreatic parenchyma condition. The purpose of this study was to compare dynamic contrast enhancement (DCE) and radiomics features in predicting pancreatic adenocarcinoma grades. The data obtained from an analysis of the radiomics of pancreatic cancer can subsequently be used to predict the prognosis of patients with PDAC and for the inclusion of big data analysis in the diagnosis process.
Methods and materials:62 consecutive patients with histologically-confirmed PDAC who underwent CE-MDCT were enrolled in the study. We compared the mean lesion attenuation and relative tumour enhancement ratio (RTE) of different types of PA and radiomics features using arterial, venous, and delayed-phase MDСT scans.
Results:Tumour attenuation parameters showed significant association with the PDAC grade, especially when calculated using both venous and delayed-phase images. RTEdel showed a significant difference between low and high-grade PDAC (-0.17±0.3 vs -0.06±0.09, p<0.005). RTEart showed a significant difference between intermediate and high-grade PDAC (-0.52±0.23 vs -0.67±0.19, p<0.05). We identified different radiomics features that best correlated with the PDAC grade (MPP, kurtosis, SSF etc). The diagnostic performance of joint CE-MDCT features with radiomics signature increased compared with CE_MDCT features alone (sensitivity 97.3 vs 91.3%, specificity 84.1 vs 80%).
Conclusion:The analysis of radiomics features and DCE features is useful for PDAC grade prediction.
Limitations:The study was single-centred with standardised CE-MDCT protocols. We plan to test the robustness of texture analysis in future studies.
Ethics committee approvaln/a
Funding:No funding was received for this work.