Synapse - Predicting early hepatic recurrence in patients with pancreatic ductal adenocarcinoma using handcrafted and deep radiomics features

Predicting early hepatic recurrence in patients with pancreatic ductal adenocarcinoma using handcrafted and deep radiomics features Conference Paper

Authors:	Ghahremannezhad, H.; Barekzai, A. B.; Magnin, J.; Zambririnis, C. P.; Horvat, N.; Gonen, M.; Schwartz, L. H.; Do, R. K. G.; Soares, K.; Balachandran, V. P.; Drebin, J. A.; Kingham, P.; D’Angelica, M. I.; Wei, A.; Jarnagin, W. R.; Chakraborty, J.
Title:	Predicting early hepatic recurrence in patients with pancreatic ductal adenocarcinoma using handcrafted and deep radiomics features
Conference Title:	Medical Imaging 2025: Computer-Aided Diagnosis
Abstract:	Pancreatic ductal adenocarcinoma (PDAC) is an aggressive malignancy with a five-year survival rate of only 12%. Although surgical resection represents the most curative option, even for patients with seemingly localized PDAC, 5-year survival after pancreatectomy remains low, primarily due to the likelihood of early recurrence. The liver is the most common site of distant metastasis, often leading to rapid disease progression and poor survival rate.1 Predicting the risk of developing liver metastasis preoperatively may help guide initial treatment towards induction chemotherapy, whereas lower-risk patients may undergo up-front surgical resection. However, current clinical and pathological factors used to predict metastasis are insufficient, highlighting the need for more precise predictive models. Radiomics, which involves extraction of quantitative features from medical imaging, has shown promise in capturing tumor heterogeneity and predicting clinical outcomes in various cancers.2 In this study, handcrafted and deep radiomic features of preoperative CT liver and pancreas were explored to predict early recurrence within liver with a cohort of 223 treatment-naive PDAC patients. Four machine learning models—Logistic Regression, Random Forest, Support Vector Machines, and Naive Bayes classifiers—were employed to design the predictive models after selecting the most discriminating features using the minimum redundancy maximum relevance (mRMR) method.3 With 30% of the data, used as an independent validation cohort, the handcrafted features from the liver and pancreas achieved the best area under the receiver operating characteristic curve (AUC) of 0.73 and 0.77 respectively, whereas deep features from the liver and pancreas achieved an AUC of 0.75 and 0.71 respectively using Random Forest classifier. The findings demonstrated that handcrafted and deep radiomics features from liver and pancreas regions hold promise in predicting early liver recurrence for PDAC. © 2025 SPIE.
Keywords:	survival rate; neurosurgery; distant metastasis; medical imaging; surgical resection; arthroplasty; predictive models; logistic regression; pancreatic ductal adenocarcinoma; ductal adenocarcinomas; machine learning; transplantation (surgical); transplants; deep learning; radiomics; machine-learning; localised; cardiovascular surgery; radiomic
Journal Title	Progress in Biomedical Optics and Imaging - Proceedings of SPIE
Volume:	13407
Conference Dates:	2025 Feb 17-20
Conference Location:	San Diego, CA
ISBN:	1605-7422
Publisher:	SPIE
Date Published:	2025-01-01
Start Page:	134071L
Language:	English
DOI:	10.1117/12.3048789
PROVIDER:	scopus
DOI/URL:	https://www.scopus.com/inward/record.uri?eid=2-s2.0-105004406344&doi=10.1117%2f12.3048789&partnerID=40&md5=fc0a1e8f8ae8d7f9d0e5d9a967c2690a
Notes:	Conference paper (ISBN: 9781510685925) -- Source: Scopus