Back to the list
Congress: ECR25
Poster Number: C-12773
Type: Poster: EPOS Radiologist (scientific)
Authorblock: K. Bartnik1, T. Bartczak2, M. Krzyziński2, K. Korzeniowski1, K. J. Lamparski1, T. Wróblewski2, K. Mech2, M. M. Januszewicz1, P. Biecek2; 1Warszawa/PL, 2Warsaw/PL
Disclosures:
Krzysztof Bartnik: Grant Recipient: Integra WUM-PW (No. 1W12/ INTEGRA.1.6/N/23)
Tomasz Bartczak: Nothing to disclose
Mateusz Krzyziński: Nothing to disclose
Krzysztof Korzeniowski: Nothing to disclose
Krzysztof Jacek Lamparski: Nothing to disclose
Tadeusz Wróblewski: Nothing to disclose
Katarzyna Mech: Nothing to disclose
Magdalena Maria Januszewicz: Nothing to disclose
Przemysław Biecek: Nothing to disclose
Keywords: Artificial Intelligence, Liver, Oncology, CT, CT-Quantitative, Chemoembolisation, Outcomes analysis, Cancer, Cirrhosis
Purpose Hepatocellular carcinoma (HCC) is the most common primary liver malignancy, presenting a wide spectrum of prognoses and treatment options, including liver transplantation, resection, local-regional therapies, chemotherapy, and immunotherapy. Among these, transarterial chemoembolization (TACE) remains the standard of care for patients with unresectable HCC without extrahepatic spread and with preserved venous flow. Artificial intelligence (AI) is increasingly utilized in HCC research for detection, segmentation, and prognostication, yet the development of robust AI models requires diverse and comprehensive datasets integrating imaging and...
Read more Methods and materials This retrospective dataset includes extensive clinical data, baseline multiphase CT imaging, segmentations, and radiomics features of 233 treatment-naive patients. Inclusion criteria: 1) unresectable HCC; 2) conventional TACE; 3) Child-Pugh Class A/B; 4) baseline contrast-enhanced CT. Exclusion criteria: 1) liver transplantation, resection, or ablation; 2) other neoplasms. Outcome measures included overall survival (OS), progression-free survival (PFS), and TACE response. Segmentation masks for various internal organs were generated using nnU-Net model, while HCC were manually segmented in Slicer 3D by experienced radiologists....
Read more Results The WAT-TACE dataset comprises four key components: clinical data, imaging, segmentations, and radiomic features. The dataset includes clinical variables and outcome measures for 233 patients (median age: 66 years; 185 male, 48 female), with a median overall survival of 27 months and a median progression-free survival of 18 months. Imaging data consists of contrast-enhanced CT scans across multiple phases, including precontrast, late arterial, portal venous, and delayed phases. Segmentation data include 377 HCC lesion masks and 104 organ volumes of...
Read more Conclusion Our dataset significantly expands the available annotated data, complementing a recently published TACE dataset. Notably, it includes patients with multiple HCC lesions, rather than being limited to single tumors. This dataset includes comprehensive clinical data, pre-TACE CT imaging, segmentation masks and critical outcome measures. It is a valuable resource for enhancing AI-based research in radiology, aimed at improving outcomes for HCC patients. 
Read more References Disclosure Statement:While pending for presentation at the ECR Conference, this material has already been published in the Radiology: Artificial Intelligence journal under the title:"WAW-TACE: A Hepatocellular Carcinoma Multiphase CT Dataset with Segmentations, Radiomics Features, and Clinical Data"Authors: Krzysztof Bartnik, Tomasz Bartczak, Mateusz KrzyziÅ„ski, Krzysztof Korzeniowski, Krzysztof Lamparski, Piotr WÄ™grzyn, Eric Lam, Mateusz Bartkowiak, Tadeusz Wróblewski, Katarzyna Mech, Magdalena Januszewicz, PrzemysÅ‚aw BiecekPublished Online: October 23, 2024DOI: https://doi.org/10.1148/ryai.240296We acknowledge that the dataset and findings presented in this poster are based on the...
Read more