Online Programme
Search for the sessions or presentations
Topic
AI, Machine Learning, Radiomics
[{"name":"SS 5.1 - Radiomics machine-learning model for the prediction of local tumour progression after thermal ablation for colorectal liver metastases","mode":"playlist","public":true,"info":{"id":567,"presentation":"SS 5.1 - Radiomics machine-learning model for the prediction of local tumour progression after thermal ablation for colorectal liver metastases","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"M. Taghavi, Amsterdam, NL","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/567_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/567_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/BCB5DF92.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/567_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/567_1p-thumbnails.vtt"}]},{"name":"SS 5.3 - Radiomic analysis of hepatobiliary-phase primovist MRI is associated with disease-free survival in patients with surgically resectable colorectal liver metastases","mode":"playlist","public":true,"info":{"id":504,"presentation":"SS 5.3 - Radiomic analysis of hepatobiliary-phase primovist MRI is associated with disease-free survival in patients with surgically resectable colorectal liver metastases","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"J. Shur, London, GB","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/504_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/504_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/4120A1FB.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/504_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/504_1p-thumbnails.vtt"}]},{"name":"SS 5.4 - Deep learning for fully automated segmentation of rectal tumours on MRI in a multicenter setting","mode":"playlist","public":true,"info":{"id":849,"presentation":"SS 5.4 - Deep learning for fully automated segmentation of rectal tumours on MRI in a multicenter setting","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"J. Van Griethuysen, Amsterdam, NL","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/849_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/849_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/72376E96.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/849_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/849_1p-thumbnails.vtt"}]},{"name":"SS 5.5 - Assessment of malignant potential in intraductal papillary mucinous neoplasms of the pancreas using MR findings and texture analysis","mode":"playlist","public":true,"info":{"id":423,"presentation":"SS 5.5 - Assessment of malignant potential in intraductal papillary mucinous neoplasms of the pancreas using MR findings and texture analysis","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"S. Jeon, Seoul, KR","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/423_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/423_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/8AD1EB5A.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/423_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/423_1p-thumbnails.vtt"}]},{"name":"SS 5.6 - Reproducibility of radiomics in pelvic MRI: effect of variations between readers, segmentation methodology and software","mode":"playlist","public":true,"info":{"id":559,"presentation":"SS 5.6 - Reproducibility of radiomics in pelvic MRI: effect of variations between readers, segmentation methodology and software","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"N. Schurink, Amsterdam, NL","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/559_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/559_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/90B2F165.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/559_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/559_1p-thumbnails.vtt"}]},{"name":"SS 5.8 - Texture analysis of preoperative CT images of mass-forming cholangiocarcinoma: 2D and 3D texture analysis with disease-free survival","mode":"playlist","public":true,"info":{"id":330,"presentation":"SS 5.8 - Texture analysis of preoperative CT images of mass-forming cholangiocarcinoma: 2D and 3D texture analysis with disease-free survival","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"S. Park, Seoul, KR","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/330_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/330_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/F1F9E383.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/330_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/330_1p-thumbnails.vtt"}]},{"name":"SS 5.9 - Influence of different adaptive statistical iterative reconstruction levels on CT radiomic features","mode":"playlist","public":true,"info":{"id":869,"presentation":"SS 5.9 - Influence of different adaptive statistical iterative reconstruction levels on CT radiomic features","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"F. Pucciarelli, Rome, IT","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/869_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/869_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/5CC14610.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/869_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/869_1p-thumbnails.vtt"}]},{"name":"SS 5.10 - Prediction of splenomegaly in \u003E100,000 structured oncologic radiology reports using natural language processing","mode":"playlist","public":true,"info":{"id":418,"presentation":"SS 5.10 - Prediction of splenomegaly in \u003E100,000 structured oncologic radiology reports using natural language processing","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"S. Sun, New York, US","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/418_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/418_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/EE36CBC9.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/418_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/418_1p-thumbnails.vtt"}]}]
{"type":2,"code":"6E32WM0A"}
[session]
[presentation]
[presenter]
SS 5.1 - Radiomics machine-learning model for the prediction of local tumour progression after thermal ablation for colorectal liver metastases
Presentation Number
SS 5.1
Channel
On-demand channel 4
Purpose
Knowledge about the risk of local tumor progression (LTP) after thermal ablation for colorectal liver metastases (CRLM) patients is critical to optimize percutaneous ablation results and for subsequent follow-up. The aim of this study was to develop and validate a machine-learning radiomics model to predict LTP based on pre-ablation CT in CRLM patients.
Material and methods
93 CRLM patients (143 lesions) treated by means of ablation were included and divided into a training (n=65 patients, n=102 lesions) and validation (n=28 patients, n=41 lesions) set. The validation set was independent of the training set. After manual segmentation and preprocessing, 1,316 radiomics features were extracted for each lesion. Three predictive models were constructed based on (1) radiomics features, (2) clinical features and (3) a combination of clinical and radiomics features to predict two-year LTP-free survival (LTPFS). Bayesian-optimized gradient boosting with wrapper feature selection was trained on the training set and optimized with sequential model-based optimization for prediction models based on radiomics features.
Results
Median follow-up was 24 months (range 6-115). 26 patients had LTP in 32 lesions. The concordance_index in the validation set to predict LTPFS was (0.76; 95% confidence interval [CI]: 0.75-0.77, p=0.01 risk stratification) for the radiomics model, (0.60; 95%CI: 0.58-0.61, p=0.69; risk stratification) for the clinical model and (0.76; 95%CI: 0.75-0.77, p=0.01 risk stratification) for the combined model.
Conclusion
Machine learning-based predictive models incorporating pre-ablation radiomics features showed good prognostic potential and allowed significant stratification for LTPFS after thermal ablation in CRCLM patients.
Collapse
[{"name":"SS 5.1 - Radiomics machine-learning model for the prediction of local tumour progression after thermal ablation for colorectal liver metastases","mode":"playlist","public":true,"info":{"id":567,"presentation":"SS 5.1 - Radiomics machine-learning model for the prediction of local tumour progression after thermal ablation for colorectal liver metastases","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"M. Taghavi, Amsterdam, NL","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/567_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/567_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/BCB5DF92.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/567_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/567_1p-thumbnails.vtt"}]}]
{"status_url":"https:\/\/cslide.ctimeetingtech.com\/play\/8E22WM0aM\/status","status":20,"track_url":"https:\/\/cslide.ctimeetingtech.com\/play\/8E22WM0aM\/track","track":60,"provider":"CTI","provider_live":0,"type":1,"code":"8E22WM0aM"}
[session]
[presentation]
[presenter]
Collapse
SS 5.3 - Radiomic analysis of hepatobiliary-phase primovist MRI is associated with disease-free survival in patients with surgically resectable colorectal liver metastases
Presentation Number
SS 5.3
Channel
On-demand channel 4
Purpose
Colorectal cancer with liver metastases (CRLM) is potentially curable with surgical resection; however, clinical prognostic factors can insufficiently stratify patients. This study aims to assess whether radiomic features from CT and MRI are prognostic and can inform clinical decision-making.
Material and methods
This single-site retrospective study included 102 patients who underwent CRLM resection with pre-operative CT and MRI with gadoxetic-acid (EOB). A lasso-regularized multivariate Cox proportional hazards model was applied to 3 sets of 104 radiomic features derived from the portal-venous CT, unenhanced T1-weighted fat-suppressed (T1FS) and hepatobiliary phase (HBP) data, respectively, to determine association with disease-free survival (DFS). A prognostic index was derived using the significant Cox regression coefficients and their corresponding input features and a threshold was determined to classify patients into high- and low-risk groups, and DFS compared using log-rank tests.
Results
Two radiomic co-variates were significantly associated with DFS; minimum pixel value (MIN) (HR=1.66, p=0.00016) and small area emphasis (HR=0.62, p=0.0013) from the EOB-MRI data. Radiomic T1FS and CT features were not prognostic. The prognostic index stratified high- and low-risk prognostic groups, although this was not significant (HR 0.251, p=0.096). MIN was positively associated with delayed tumour enhancement (r= 0.77, p< 2 x 10-16).
Conclusion
Radiomic HBP primovist MRI features are associated with DFS, but not those derived from CT or T1FS data, and are partly explained by delayed tumour enhancement, likely due to post-treatment tumour fibrosis. This merits further validation for potential clinical implementation to inform patient management.
Collapse
[{"name":"SS 5.3 - Radiomic analysis of hepatobiliary-phase primovist MRI is associated with disease-free survival in patients with surgically resectable colorectal liver metastases","mode":"playlist","public":true,"info":{"id":504,"presentation":"SS 5.3 - Radiomic analysis of hepatobiliary-phase primovist MRI is associated with disease-free survival in patients with surgically resectable colorectal liver metastases","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"J. Shur, London, GB","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/504_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/504_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/4120A1FB.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/504_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/504_1p-thumbnails.vtt"}]}]
{"status_url":"https:\/\/cslide.ctimeetingtech.com\/play\/aE22WM09G\/status","status":20,"track_url":"https:\/\/cslide.ctimeetingtech.com\/play\/aE22WM09G\/track","track":60,"provider":"CTI","provider_live":0,"type":1,"code":"aE22WM09G"}
[session]
[presentation]
[presenter]
Collapse
SS 5.4 - Deep learning for fully automated segmentation of rectal tumours on MRI in a multicenter setting
Presentation Number
SS 5.4
Channel
On-demand channel 4
Purpose
To explore the performance of deep learning to perform fully automated segmentation of rectal tumors on baseline MRI using a multicenter dataset (with variations in acquisition protocols and scan quality reflecting daily practice).
Material and methods
Baseline staging MRIs of 355 patients (from 6 centers) were analyzed. Overall scan quality was assessed using a 5-point Likert score (0=poor) to 4=excellent). Data were randomly split into train:validation:test cohorts (ratio of 5:1:4). An expert-radiologist manually delineated all rectal tumours to serve as training input. Using the T2-weighted, diffusion-weighted imaging (DWI)-b1000 and apparent diffusion coefficient (ADC)-maps from the training+validation cohorts, an attention-gated u-net was trained, with optimal hyperparameters (including learning rate, optimizer algorithm and batch-size) determined via grid search. A second expert-radiologist independently re-segmented all patients in the test cohort to calculate inter-reader agreement using the dice similarity coefficient (DSC). Agreement (DSC) between the trained network and the expert-segmentations was compared to the expert inter-reader agreement (serving as the standard of reference).
Results
Average DSC between expert-readers was 0.75 (±0.18). Average DSC for the network-generated segmentations was 0.64 (±0.22) and 0.59 (±0.22) compared to expert-reader 1 and 2, respectively. When excluding poor quality (score ≤ 2) scans, DSC between the network and expert-readers increased to 0.68 (±0.17) and 0.63 (±0.17).
Conclusion
Despite large variations in scan protocol/image quality, deep learning networks achieved a promising overall performance to automatically segment rectal tumours on MRI. Better results were achieved after exclusion of poor-quality scans, with agreement levels up to 0.63-0.68 (versus 0.75 between expert-radiologists).
Collapse
[{"name":"SS 5.4 - Deep learning for fully automated segmentation of rectal tumours on MRI in a multicenter setting","mode":"playlist","public":true,"info":{"id":849,"presentation":"SS 5.4 - Deep learning for fully automated segmentation of rectal tumours on MRI in a multicenter setting","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"J. Van Griethuysen, Amsterdam, NL","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/849_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/849_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/72376E96.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/849_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/849_1p-thumbnails.vtt"}]}]
{"status_url":"https:\/\/cslide.ctimeetingtech.com\/play\/aE22WM0fD\/status","status":20,"track_url":"https:\/\/cslide.ctimeetingtech.com\/play\/aE22WM0fD\/track","track":60,"provider":"CTI","provider_live":0,"type":1,"code":"aE22WM0fD"}
[session]
[presentation]
[presenter]
Collapse
SS 5.5 - Assessment of malignant potential in intraductal papillary mucinous neoplasms of the pancreas using MR findings and texture analysis
Presentation Number
SS 5.5
Channel
On-demand channel 4
Purpose
To investigate the usefulness of MR findings and texture analysis for predicting the malignant potential of pancreatic intraductal papillary neoplasms (IPMNs).
Material and methods
248 patients with surgically confirmed IPMNs (106 high grade (HG; invasive carcinoma and high-grade dysplasia) and 142 low grade (LG; low/intermediate-grade dysplasia)) and who underwent preoperative MRI with MRCP were included. MR findings suggestive of high-risk stigmata or worrisome features based on the international consensus Fukuoka guidelines 2017 were analyzed. Quantitative features were extracted using texture analysis of T2-weighted MRCP. Multivariate analysis was used to identify independent predictors for HG IPMNs. Diagnostic performance was also analyzed using receiver operating curve analysis.
Results
Among MR findings, enhancing mural nodules ≥5mm, main pancreatic ductal (MPD) dilatation ≥10mm, and abrupt change of MPD with upstream parenchymal atrophy were significant predictors for HG IPMNs (all Ps <0.05). Among texture variables, the significant predictors for HG IPMNs were lower sphericity (P=0.004) and lower compactness (P<0.001). At multivariate analysis, enhancing mural nodule ≥5mm (odds ratios (ORs), 7.97; 95% confidence interval (CI), 4.10-15.52; P<0.001), MPD dilatation ≥10mm (OR, 2.59; 95% CI, 1.16-5.79; P=0.021) and lower compactness on texture analysis (OR, 0.81; 95 % CI, 0.67-0.98; P=0.032) were significant factors for predicting HG IPMNs. Addition of texture variable to MR findings showed better diagnostic performance for predicting HG IPMNs than using MR findings only (AUC, 0.83 vs. 0.79, P=0.008).
Conclusion
MRCP-derived texture features are useful for predicting malignant potential of IPMNs and addition of texture analysis to MRI features may improve diagnostic performance for predicting HG IPMNs.
Collapse
[{"name":"SS 5.5 - Assessment of malignant potential in intraductal papillary mucinous neoplasms of the pancreas using MR findings and texture analysis","mode":"playlist","public":true,"info":{"id":423,"presentation":"SS 5.5 - Assessment of malignant potential in intraductal papillary mucinous neoplasms of the pancreas using MR findings and texture analysis","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"S. Jeon, Seoul, KR","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/423_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/423_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/8AD1EB5A.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/423_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/423_1p-thumbnails.vtt"}]}]
{"status_url":"https:\/\/cslide.ctimeetingtech.com\/play\/9E22WM08i\/status","status":20,"track_url":"https:\/\/cslide.ctimeetingtech.com\/play\/9E22WM08i\/track","track":60,"provider":"CTI","provider_live":0,"type":1,"code":"9E22WM08i"}
[session]
[presentation]
[presenter]
Collapse
SS 5.6 - Reproducibility of radiomics in pelvic MRI: effect of variations between readers, segmentation methodology and software
Presentation Number
SS 5.6
Channel
On-demand channel 4
Purpose
Although several studies have investigated the reproducibility of radiomics data derived from CT and PET/CT, data on the reproducibility of MR-based radiomics are scarce. This study assesses the reproducibility of radiomic features from pelvic MRI data and investigates the effects of variations between readers, segmentation methodology and software packages.
Material and methods
25 pelvic MRIs (T2W-MRI of anal cancer) were retrospectively analyzed and segmented by two readers to include the [1] whole-tumour volume, and [2] largest single axial tumour slice. Pixel intensities were normalized to mean=300/SD=100, and images were resampled isotropically (2x2x2mm3). Radiomic features were extracted using 2 open-source packages (PyRadiomics-v2.2.0, CaPTk-v1.7.3), using comparable settings without image filtration. A fixed histogram bin of 5 was used. Only features defined in both packages were extracted (first-order, shape, GLCM, GLRLM, GLSZM and NGTDM features, 51 total). For each feature, the intra-class correlation coefficient (ICC) was calculated between the [1] two readers, [2] two segmentation methods (whole-volume vs. single-slice) and [3] two software packages.
Results
Inter-reader reproducibility was moderate (20/51 features; 0.5<ICC<=0.75) to good (15/51; 0.75<ICC<=0.9). Between segmentation methods, most features (in particular GLRLM, GLSZM, NGTDM) showed poor reproducibility (31/45; ICC<0.5), though first-order features showed good (7/15; 0.75<ICC<=0.9) to excellent (2/15; ICC>0.9) reproducibility. Between software packages, most first-order, shape, GLCM and GLRLM features showed excellent reproducibility (23/30; ICC>0.9). The remaining higher order features (GLSZM, NGTDM) were all poorly reproducible (21/21; ICC<0.5).
Conclusion
Variations in software and segmentation methodology negatively affected measurement reproducibility in MRI-based radiomics, especially higher order features. Inter-reader reproducibility was moderate-to-good.
Collapse
[{"name":"SS 5.6 - Reproducibility of radiomics in pelvic MRI: effect of variations between readers, segmentation methodology and software","mode":"playlist","public":true,"info":{"id":559,"presentation":"SS 5.6 - Reproducibility of radiomics in pelvic MRI: effect of variations between readers, segmentation methodology and software","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"N. Schurink, Amsterdam, NL","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/559_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/559_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/90B2F165.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/559_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/559_1p-thumbnails.vtt"}]}]
{"status_url":"https:\/\/cslide.ctimeetingtech.com\/play\/aE22WM0aD\/status","status":20,"track_url":"https:\/\/cslide.ctimeetingtech.com\/play\/aE22WM0aD\/track","track":60,"provider":"CTI","provider_live":0,"type":1,"code":"aE22WM0aD"}
[session]
[presentation]
[presenter]
Collapse
SS 5.8 - Texture analysis of preoperative CT images of mass-forming cholangiocarcinoma: 2D and 3D texture analysis with disease-free survival
Presentation Number
SS 5.8
Channel
On-demand channel 4
Purpose
To determine whether CT texture analysis (CTTA) has a value in the prediction of disease-free survival (DFS) in patients with mass-forming type intrahepatic cholangiocarcinoma (mICC) undergoing surgical resection.
Material and methods
The late arterial-phase CT scans of 89 patients with mICC who underwent surgical treatment were retrospectively analyzed. CTTA was performed using a software (Radiomics, Syngo.via Frontier, Siemens Healthineers, Forchheim, Germany) that employed a first-order and second-order texture analysis by drawing a region of interest of 1) the largest cross-sectional area of the tumor (2D) and 2) whole tumor volume (3D). Patients were followed up until disease progression. Cox proportional hazard models were used to determine the relationship between texture features and DFS.
Results
Univariate analysis of 2D texture identified that first-order mean (p=.001), energy (p=.037), kurtosis (p=.001), and shape-flatness (p=.006) were significant univariate markers of DFS. Univariate analysis of 3D texture yielded mean (p<.001) as a significant factor. Among clinicopathologic parameters, size (p <.001), extrahepatic involvement (p=.006), multiplicity (p=.016), lymph node involvement (p=.000), and CEA (p=.003) were significant univariate markers. A Cox regression model including all significant univariate markers identified no significant texture factors on 2D analysis but first-order mean (p=.006) on 3D analysis. Size and lymph node (LN) involvement were significant factors on 2D and 3D analyses and CEA was a significant factor in 2D analysis on multivariate analysis.
Conclusion
The mean of the 3D texture parameters is independently associated with poorer DFS in patients with mICC, while other texture parameters did not show correlation with DFS.
Collapse
[{"name":"SS 5.8 - Texture analysis of preoperative CT images of mass-forming cholangiocarcinoma: 2D and 3D texture analysis with disease-free survival","mode":"playlist","public":true,"info":{"id":330,"presentation":"SS 5.8 - Texture analysis of preoperative CT images of mass-forming cholangiocarcinoma: 2D and 3D texture analysis with disease-free survival","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"S. Park, Seoul, KR","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/330_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/330_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/F1F9E383.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/330_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/330_1p-thumbnails.vtt"}]}]
{"status_url":"https:\/\/cslide.ctimeetingtech.com\/play\/6E22WM06G\/status","status":20,"track_url":"https:\/\/cslide.ctimeetingtech.com\/play\/6E22WM06G\/track","track":60,"provider":"CTI","provider_live":0,"type":1,"code":"6E22WM06G"}
[session]
[presentation]
[presenter]
Collapse
SS 5.9 - Influence of different adaptive statistical iterative reconstruction levels on CT radiomic features
Presentation Number
SS 5.9
Channel
On-demand channel 4
Purpose
To evaluate the influence of different levels of adaptive statistical iterative reconstruction (ASIR) on CT radiomic features.
Material and methods
38 patients who underwent unenhanced CT scans of the abdomen with the same scanner (Revolution Evo, GE Healthcare, USA) were analyzed. Subsequently, raw data of filtered backprojection (FBP) were reconstructed with 10 levels of ASIR (from 10 to 100). Two radiologists analyzed texture features of liver and kidney tissues using two different regions of interest (ROIs) that were cloned for all eleven different iteration level datasets. Data were elaborated with TexRad Medical Imaging Software. Six different radiomic features (mean, sd, entropy, mpp, skewness, kurtosis) were extrapolated and compared between FBP and all ASIR levels.
Results
Texture analysis of the liver revealed significant differences between FBP and all ASIR reconstructions for mean (all p<0.002), sd (all p<0.0001), entropy (all p<0.0001) and mpp (all p<0.0001), while no significant differences were observed for skewness and kurtosis between FBP and all ASIR reconstructions (all p>0.45 and all p>0.58, respectively). Similar results were obtained for kidney analysis with no significant differences for skewness and kurtosis (all p>0.053 and all p>0.176, respectively) and significant changes for mean (all p<0.0001), sd (all p<0.0001), entropy (all p<0.0036) and mpp (all p<0.0001).
Conclusion
No influence of iterative reconstruction algorithm was reported for skewness and kurtosis compared to FBP in liver and kidney analysis whereas mean, sd, entropy and mpp were significantly affected by ASIR. Skewness and kurtosis may be reliable quantitative parameters.
Collapse
[{"name":"SS 5.9 - Influence of different adaptive statistical iterative reconstruction levels on CT radiomic features","mode":"playlist","public":true,"info":{"id":869,"presentation":"SS 5.9 - Influence of different adaptive statistical iterative reconstruction levels on CT radiomic features","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"F. Pucciarelli, Rome, IT","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/869_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/869_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/5CC14610.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/869_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/869_1p-thumbnails.vtt"}]}]
{"status_url":"https:\/\/cslide.ctimeetingtech.com\/play\/aE22WM0fZ\/status","status":20,"track_url":"https:\/\/cslide.ctimeetingtech.com\/play\/aE22WM0fZ\/track","track":60,"provider":"CTI","provider_live":0,"type":1,"code":"aE22WM0fZ"}
[session]
[presentation]
[presenter]
Collapse
SS 5.10 - Prediction of splenomegaly in >100,000 structured oncologic radiology reports using natural language processing
Presentation Number
SS 5.10
Channel
On-demand channel 4
Purpose
To develop and assess the accuracy of a natural language processing (NLP) model to identify splenomegaly from structured CT radiology reports at a tertiary cancer center.
Material and methods
In an IRB-approved, retrospective study, all CT chest/abdomen/pelvis reports (July 2009 to April 2019) adhering to departmental structured template were included. The SPLEEN subsection was extracted and those with default ‘unremarkable’ text were excluded from training. For patients with colorectal cancers (CRC), hepatobiliary cancers (HB), leukemia, Hodgkin’s lymphoma (HL) and non-HL (NHL), 1920 of 105,042 reports were annotated as positive or negative/uncertain for splenomegaly. Model training was performed on 1536 and model accuracy was tested on 384 reports. The prediction model was then applied to the remaining reports to calculate frequencies of splenomegaly.
Results
In the annotated reports, splenomegaly was present in 42.2%. After training, the splenomegaly classifier achieved 94% overall accuracy, 94.6% precision (positive predictive value) and 94% recall (sensitivity). When the model was applied to all unannotated reports, the predicted frequency of splenomegaly for CRC patients was 8.7% (5275/60462), HB 17.7% (2210/12506), leukemia 31.5% (1684/5340), HL 6.1% (390/6386) and NHL 9.2% (1866/20348).
Conclusion
NLP can predict splenomegaly from structured radiology reports after training from a limited sample of annotated text. At our institution, the frequency of splenomegaly in CRC patients was similar to HL and NHL patients, and lower than both patients with HB cancers and leukemia. Validation with splenic volumetry is ongoing.
Collapse
[{"name":"SS 5.10 - Prediction of splenomegaly in \u003E100,000 structured oncologic radiology reports using natural language processing","mode":"playlist","public":true,"info":{"id":418,"presentation":"SS 5.10 - Prediction of splenomegaly in \u003E100,000 structured oncologic radiology reports using natural language processing","session":"SS 5 - Machine learning and radiomics: current applications in GI imaging","presenter":"S. Sun, New York, US","photo":""},"playlist":[{"name":"","source":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/418_1p.mp4","source_path":"s3:\/\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/418_1p.mp4","cover":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/EE36CBC9.jpg","data":[],"tracks":[],"chapters":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/418_1p.vtt","thumbnails":"https:\/\/s3.eu-central-1.amazonaws.com\/eu-cslide-prod-recordings\/esgar2020\/34\/v\/418_1p-thumbnails.vtt"}]}]
{"status_url":"https:\/\/cslide.ctimeetingtech.com\/play\/9E22WM08d\/status","status":20,"track_url":"https:\/\/cslide.ctimeetingtech.com\/play\/9E22WM08d\/track","track":60,"provider":"CTI","provider_live":0,"type":1,"code":"9E22WM08d"}
[session]
[presentation]
[presenter]
Collapse