Machine learning and artificial intelligence in liquid biopsy-based early detection of pancreatic cancer: a scoping review

www.nature.com/bjcreports REVIEW OPEN Machine learning and artificial intelligence in liquid biopsybased early detection of pancreatic cancer: a scoping review Joy Ku1, Meenakshi Singhal1, Margaret Burnette2 and Samar A. Hegazy1 ✉ © The Author(s) 2026 1234567890();,: Pancreatic ductal adenocarcinoma (PDAC) presents as a cancer with an especially poor prognosis, largely due to the challenges surrounding its early diagnosis. Liquid biopsy has emerged as a promising, noninvasive method for screening across a variety of cancers. This approach is limited, however, by the extensive heterogeneity of biological samples, a challenge that teams are looking to address using artificial intelligence (AI) and machine learning (ML) strategies. By harnessing the ability of ML algorithms to extract the most salient features from complex datasets, researchers can identify biomarkers with high predictive value for PDAC. This review explores the current landscape of AI-powered liquid biopsy for early PDAC diagnosis, focusing on specific techniques and their respective degrees of success. Following PRISMA-ScR guidelines, 85 studies were extracted from PubMed and Scopus with a final 18 studies included. The majority of papers utilized blood (n = 15) as the source of liquid biopsy, with the remainder analyzing urine, bile, or cyst fluid. Random forests (n = 9) and support vector machines (n = 7) were the most frequently implemented ML models, while two papers focused on deep learning methods. Limitations include the lack of standardized reporting for model performance metrics and small cohort sizes with non-granular labels. BJC Reports; https://doi.org/10.1038/s44276-026-00232-y INTRODUCTION Pancreatic cancer, typically referring to pancreatic ductal adenocarcinoma (PDAC), is among the most lethal common cancers due to delayed diagnosis and subsequent poor prognosis and survival rates. One of the key contributing factors to this is the nonspecific nature of the disease’s presenting symptoms, which include fatigue, back or abdominal pain, weight loss, and jaundice [1, 2]. Challenges in screening for PDAC are further exacerbated by early metastatic dissemination of PDAC through possible cellular reprogramming mechanisms, as recently highlighted by Opsahl et al. and their investigations in organ-specific premetastatic niche formation [3]. Unfortunately, the incidence of PDAC continues to rise in the United States across all ethnicities in both men and women, with minority women experiencing the most disproportionate increases [4, 5]. PDAC is also projected to become the second-leading cause of cancer-related mortality by 2030 [6]. A large proportion of cases are metastatic or locally advanced at diagnosis, precluding surgical intervention for these patients. Only 15–20% of patients are eligible for resection at presentation, and even with optimal therapy, the 5-year survival rate remains approximately 10% in the United States [7]. Coupled with limited effective treatment options, there is also a lack of available screening strategies. As such, teams across the world are actively investigating potential biomarkers for early detection of PDAC, leveraging artificial intelligence (AI) or machine learning (ML) tools to streamline the process of identifying those with high diagnostic and prognostic value. This scoping review focuses on the most recent developments in this area, first highlighting the main PDAC biomarkers and sources of liquid biopsies being investigated. Then, specific current AI techniques and associated performance metrics are discussed, including current limitations and future potential. BACKGROUND INFORMATION Pancreatic cancer PDAC patients are more likely to present with advanced disease at initial diagnosis, with roughly 28% exhibiting regional spread to the lymph nodes and 51% with distant metastasis as reported by the National Cancer Institute’s Surveillance, Epidemiology, and End Results Program [8]. The lack of timely detection can be partially attributed to the pancreas’s retroperitoneal location, which leads to a clinically silent malignancy until the neoplasm grows to a significant size that obstructs surrounding ducts. PDAC is also considered genetically aggressive due to its high frequency of driver mutations like KRAS (>90%), and inactivation of key tumor suppressor genes such as CDKN2A, TP53, and SMAD4 [9, 10]. These mutations result in the progression of pancreatic precursor lesions to malignancy associated with rapid local invasion and distant metastatic spread. With rates of PDAC incidence on the rise, there is a growing need for strategic screening to identify possible cases at earlier timepoints. Current guidelines recommend that individuals who may have a strong family history of pancreatic cancer or genetic syndromes that increase their risk of developing PDAC, along with patients who have pancreatic cysts (intraductal papillary mucinous neoplasms or mucinous cystic neoplasms) be screened [11, 12]. While pancreatic cysts do not necessarily indicate subsequent Carle Illinois College of Medicine, University of Illinois Urbana-Champaign, Urbana, IL, USA. 2University of Illinois Urbana-Champaign, Urbana, IL, USA. ✉email: 1 Received: 19 November 2025 Revised: 10 April 2026 Accepted: 4 May 2026 J. Ku et al. 2 malignant transformation, the ability to define an at-risk population who should undergo further testing is the first step to shifting the timeline in the detection of PDAC. Current screening measures include performing an endoscopic ultrasound (EUS) or magnetic resonance cholangiopancreatography (MRCP) [13]. However, these can be costly, semi-invasive, and increase the risk of potential harm to patients [14]. Liquid biopsy Liquid biopsy, the use of readily available biological fluid samples, is an increasingly employed diagnostic approach for cancer biomarker analysis. Liquid biopsies are either non- or minimally invasive and can potentially lower equipment expenses to allow for more routine and accessible screenings. Utilizing liquid biopsybased testing for PDAC surveillance can mitigate some of the challenges faced by imaging-based screening methods. Carbohydrate antigen 19-9 (CA19-9) is the only currently FDA-approved blood biomarker for PDAC [15]. Diagnostic efficacy of CA19-9 alone remains suboptimal, with its pooled sensitivity, specificity, and AUC at 72%, 86%, and 0.8474, respectively [16]. As such, there is a call for the use of multianalyte panels in conjunction with CA19-9 in order to improve the diagnostic potential of liquid biopsies. Analytes available from liquid biopsies vary widely (Fig. 1), with extracellular vesicles (EVs), microRNAs (miRNAs), circulating tumor DNA (ctDNA), and circulating tumor cells (CTCs), being at the forefront of developments within the field of early cancer detection [17]. EVs are nanoparticles with a lipid bilayer and are secreted by all cell types into surrounding fluids. They display tissue-specific surface protein ma (...truncated)