Diagnostic Performance of Artificial Intelligence -based Computed Tomography Techniques in Detecting Pancreatic Cancer: A Systematic Review

Diagnostic Performance of Artificial Intelligence -based Computed Tomography Techniques in Detecting Pancreatic Cancer: A Systematic Review | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Diagnostic Performance of Artificial Intelligence -based Computed Tomography Techniques in Detecting Pancreatic Cancer: A Systematic Review Mustafa S. Alhasan, Ayman S. Alhasan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7356052/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 17 You are reading this latest preprint version Abstract Introduction: Pancreatic cancer is among the most lethal malignancies, often diagnosed at an advanced stage due to its subtle clinical presentation. Artificial intelligence (AI) techniques applied to computed tomography (CT) have shown potential in improving diagnostic accuracy across various imaging tasks, including cancer detection and lesion classification. This systematic review and meta-analysis aimed to evaluate the diagnostic performance of AI-based CT models specifically for the detection of pancreatic cancer, while accounting for the clinical task type and algorithmic heterogeneity among studies. Methods A comprehensive literature search was conducted across PubMed/MEDLINE, the Cochrane Library, Web of Science, Embase, Scopus, CINAHL, and Google Scholar to identify relevant studies published up to April 2025. Eligible studies were screened according to predefined criteria, and diagnostic performance data including true positives, false positives, true negatives, and false negatives were extracted or calculated as needed from the selected publications. Results Ten studies comprising a total of 33,174 patients met the inclusion criteria. The pooled diagnostic sensitivity and specificity were 0.92 with a 95 percent confidence interval of 0.92 to 0.93, and 0.98 with a 95 percent confidence interval of 0.98 to 0.98, respectively. The area under the summary receiver operating characteristic curve was 0.959, and the diagnostic odds ratio was 179.57 with a 95 percent confidence interval of 57.98 to 556.16, indicating high diagnostic accuracy for pancreatic cancer detection. Significant heterogeneity was observed among the studies included. However, subgroup analyses did not reveal any statistically significant differences. No evidence of publication bias was detected. Conclusions The findings of this study suggest that artificial intelligence assisted computed tomography may serve as a valuable tool in supporting the diagnosis of pancreatic cancer. However, further validation using larger and more diverse datasets is necessary to confirm its clinical utility and generalizability. Artificial intelligence Pancreatic cancer Computed tomography Diagnostic accuracy Systematic review Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction Pancreatic cancer (PC) is among the most aggressive and lethal malignancies of the digestive system [ 1 ]. It carries a poor prognosis, with limited therapeutic options, and a five-year relative survival rate of only 12 percent [ 2 ]. Globally, the incidence of pancreatic cancer has increased in recent decades. Contributing factors include aging, alcohol consumption, tobacco use, physical inactivity, obesity, diabetes, chronic pancreatitis, genetic predisposition, and prolonged exposure to environmental pollutants. Unhealthy lifestyle choices and poor dietary habits have also been implicated [ 3 , 4 ]. Surgery remains the primary treatment modality for pancreatic cancer. However, several factors, including the absence of specific clinical symptoms and reliable molecular markers, often result in late-stage diagnosis, which limits the effectiveness of surgical intervention. Therefore, early detection and accurate staging are essential for improving treatment outcomes. Various imaging techniques have been explored for the diagnosis of pancreatic cancer, including endoscopic ultrasonography. Endoscopic ultrasonography (EUS), computed tomography (CT), and magnetic resonance imaging (MRI) are among the primary imaging modalities used in the evaluation of pancreatic cancer [ 6 ]. CT is the most widely employed technique for both detection and staging. Reported diagnostic accuracy for pancreatic cancer using CT and MRI ranges from 56 percent to 88 percent, while EUS offers superior spatial resolution [ 7 ]. The absence of specific symptoms or clinical indicators in patients with pancreatic tumors poses a major challenge to early diagnosis. When symptoms do occur, they are often nonspecific and typically emerge several months after the onset of pancreatic ductal adenocarcinoma (PDAC). Common presentations include jaundice, abdominal pain, and unexplained weight loss [ 8 ]. However, the majority of patients remain asymptomatic at the time of diagnosis [ 7 ]. In this challenging clinical context, artificial intelligence techniques may support and accelerate the early diagnosis of pancreatic cancer [ 9 ]. In recent years, numerous studies have been published on this topic, most of which employed a variety of algorithms and were conducted in experimental offline settings. Among these, convolutional neural networks (CNNs) are particularly prominent due to their strong capabilities in automating image analysis and processing large datasets. Artificial intelligence has shown potential as a supportive tool for radiologists in detecting both pre-neoplastic and neoplastic pancreatic lesions [ 10 ]. However, evidence regarding the diagnostic performance of artificial intelligence based on CT imaging for pancreatic cancer remains limited. Therefore, the aim of this manuscript was to systematically review published studies evaluating the diagnostic performance of artificial intelligence algorithms applied to computed tomography images for the detection of pancreatic cancer. Methods This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [ 11 ]. Resources and search techniques The following databases were searched from their inception through April 2024: PubMed/MEDLINE, the Cochrane Library, Web of Science, Embase, Scopus, CINAHL, and Google Scholar. The complete search strategy is provided in Supplementary File 1. Selection criteria After removing duplicates, articles were screened based on titles and abstracts. Studies investigating the diagnosis of pancreatic cancer using artificial intelligence applied to computed tomography were considered for inclusion. Full texts of potentially eligible studies were then reviewed to confirm eligibility. Inclusion criteria were as follows: (1) observational studies reporting the diagnostic performance of artificial intelligence applied exclusively to CT imaging for pancreatic cancer diagnosis; (2) studies involving patients with pancreatic cancer confirmed by CT; (3) publications reporting sensitivity and specificity outcomes; and (4) original research articles. Exclusion criteria were as follows: (1) studies published in languages other than English; (2) lack of full-text availability in electronic form; (3) insufficient outcome data; and (4) publication types such as comments, letters, editorials, protocols, guidelines, review papers, non-peer-reviewed articles, and conference abstracts. Data extraction Based on the established inclusion and exclusion criteria, two independent authors extracted data from the eligible studies and recorded the information using a standardized data sheet. The following variables were collected: (1) article title, (2) country of origin, (3) study design, (4) sample size, (5) number of cases, (6) number of controls, (7) patient age, (8) gender distribution, (9) tumor size in centimeters, (10) neuroimaging modality, and (11) type of algorithm used. Risk of bias assessment The methodological quality of the included studies was independently assessed by two authors using the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool [ 12 ]. This tool evaluates four domains related to risk of bias and applicability: patient selection, index test, reference standard, and flow and timing. Each item was rated as “yes,” “no,” or “unclear.” A response of “yes” indicated a low risk of bias, while “no” or “unclear” responses reflected potential bias. The results of the assessment were visualized using Review Manager (RevMan) version 5.4 (Cochrane Collaboration, Oxford, United Kingdom). Outcome measures Sensitivity, specificity, accuracy, and diagnostic odds ratio (DOR) were evaluated to determine the diagnostic performance of artificial intelligence methods in detecting pancreatic cancer. Statistical analysis To analyze pooled sensitivity, specificity, diagnostic odds ratio (DOR), and area under the curve (AUC) with 95 percent confidence intervals, this meta-analysis was performed using Meta-Disc version 1.4 and Comprehensive Meta-Analysis version 3 (Biostat Inc., USA). A two-sided p-value less than 0.05 was considered statistically significant. Diagnostic performance was evaluated using the summary receiver operating characteristic (SROC) curve, derived from the sensitivity and specificity values reported in individual studies. Heterogeneity across studies was assessed using the Cochrane chi-squared test, with a p-value less than 0.05 indicating statistical heterogeneity. The I-squared statistic was also calculated, with values greater than 50 percent indicating substantial heterogeneity. To explore the presence of a threshold effect, the Spearman correlation coefficient was computed. A strong positive correlation would suggest the presence of a threshold effect. Sensitivity analysis was conducted to assess the robustness of the findings. Publication bias was evaluated using Egger’s test, and further assessed through visual inspection of funnel plot symmetry. Results Identification of studies The database search yielded 462 records for initial screening. Of these, 149 abstracts were deemed potentially eligible and were retrieved for full-text review. Following a detailed assessment, 10 articles met the eligibility criteria and were included in this systematic review and meta-analysis. The study selection process is illustrated in the PRISMA flowchart ( Fig. 1 ) . Characteristics of included studies The included articles were published between 2020 and 2023 and originated from six countries: China (n = 4), Taiwan (n = 3), the United States (n = 2), and the Republic of Korea (n = 1). Among the ten studies included in this systematic review and meta-analysis, six were retrospective in design, two were case-control studies, and two did not report their study design. Sample sizes ranged from 412 to 20,530 participants. The majority of studies (seven out of ten) utilized contrast-enhanced computed tomography. Convolutional neural networks were the most frequently used algorithm, employed in six of the ten studies. A summary of study characteristics is provided in Table 1 . Table 1 Characteristics of included studies Article Country Study design Sample size Cases Controls Age Gender M/F Tumor size (cm) Neuroimaging modalities Algorithm Cao et al. 2023 [ 13 ] China Retrospective 20,530 ND ND ND ND ND non-contrast CT CNN Chang et al. 2023 [ 14 ] Taiwan Retrospective 1279 546 patients with pancreatic adenocarcinoma 733 controls with normal pancreas Case: 64.6 ± 11.8 Control: 54.2 ± 16.2 Case:297/249 Control:374/359 -4cm: 25.8% CE-CT XGBoost Chen et al. 2023 [ 15 ] Taiwan Retrospective 1229 546 patients with pancreatic cancer 683 Case: 65 ± 12 Control: 54 ± 16 Case:297/249 Control:374/309 2.9 (2.1–4.3) CE-CT CNN Korfiatis et al. 2023 [ 16 ] USA Case-control 3014 1105 patients with treatment-naïve biopsy proven PDA 1909 controls Case: 66 Control: 56 Case:627/478 Control:872/1037 4.9 (1.7) CT 3D-CNN Liu et al. 2020 [ 17 ] Taiwan Retrospective 690 370 patients with pancreatic cancer 320 controls 64·8 ± 12·0 260/211 3·0 (2·2–4·6) CE-CT CNN Ma et al. 2020 [ 18 ] China Retrospective 412 222 with pancreatic cancer 190 diagnosed with normal pancreas Case: 63.8 (8.7) Control: 61.0 (12.3) Case: 124/98 Control: 98/92 3.5 (2.7–4.3) CT CNN Mukherjee et al. 2022 [ 19 ] USA Case-control 420 155 265 Case: 69 Control: 67 Case: ratio 1.4 Control: ratio 1.1 ND CE-CT KNN, SVM, RM, XGBoost Park et al. 2023 [ 20 ] Republic of Korea Retrospective 2044 2044 patients with pancreatic lesions ND 60.33 1181/863 ND CE-CT 3D nnU-Net Si et al. 2021 [ 21 ] China ND 666 549 patients with abnormal CT 117 controls Training dataset: 63.3 Testing dataset: 61.8 Training dataset: 211/108 Testing dataset: 216/131 ND CE-CT fully end-to-end deep-learning Zhang et al. 2020 [ 22 ] China ND 2890 2890 CT images of patients diagnosed with pancreatic cancer ND ND ND ND CE-CT Deep CNN CT: computed tomography; CE-CT: contrast-enhanced computed tomography; PDAC: pancreatic ductal adenocarcinoma; CNN: convolutional neural network; ND: not defined; XGBoost: extreme gradient boosting; RM: random forest; SVM: support vector machine; KNN: k-nearest neighbor. Risk of bias assessment The methodological quality of the ten included studies was evaluated using the QUADAS-2 tool. In the domain of patient selection, two studies were identified as having a high risk of bias, while three studies had an unclear risk. A notable concentration of bias was observed in the index test domain, where five studies preset diagnostic thresholds without justification. The domains of reference standard and flow and timing were not associated with significant risk of bias. Concerns regarding applicability were also present. Specifically, two studies raised concerns in the patient selection domain, and four studies showed applicability issues in the index test domain. A detailed summary of the risk of bias and applicability assessments is provided in Fig. 2 . Data analysis Analysis of the forest plots revealed significant heterogeneity across studies for sensitivity, specificity, and diagnostic odds ratio (DOR). Specifically, sensitivity showed a Chi-squared value of 108.66 with a p-value of 0.000 and an I-squared of 87.1 percent ( Fig. 3 ) . Specificity demonstrated a Chi-squared value of 1677 with a p-value of 0.000 and an I-squared of 99.2 percent ( Fig. 4 ) . The DOR analysis yielded a Chi-squared value of 946.42 with a p-value of 0.000 and an I-squared of 98.5 percent ( Fig. 5 ) . Due to this substantial heterogeneity, pooled estimates were calculated using a random effects model. The pooled sensitivity of artificial intelligence-based CT methods for diagnosing pancreatic cancer was 0.92 with a 95 percent confidence interval of 0.92 to 0.93 ( Fig. 3 ) . The pooled specificity was 0.98 with a 95 percent confidence interval of 0.98 to 0.98 ( Fig. 4 ) , and the DOR was 179.57 with a 95 percent confidence interval of 57.98 to 556.16 ( Fig. 5 ) . The summary receiver operating characteristic (SROC) curve demonstrated an area under the curve (AUC) of 0.959, indicating excellent diagnostic accuracy ( Fig. 6 ) . Investigation for heterogeneity Heterogeneity among the included studies was evident in the forest plots. As heterogeneity cannot be entirely eliminated in meta-analyses, its extent and potential sources were further explored. In diagnostic accuracy studies, a common source of heterogeneity is the threshold effect. This was assessed using the Spearman correlation coefficient and the Moses model, weighted by inverse variance. The Spearman coefficient was − 0.356 with a p-value of 0.193, indicating that the threshold effect did not significantly contribute to differences in diagnostic accuracy across studies. To identify other potential sources of heterogeneity, a meta-regression analysis was conducted based on algorithm type, study design, patient ethnicity, sample size, tumor size, and imaging modality. The analysis showed no significant influence of algorithm type, study design, ethnicity, tumor size, or neuroimaging modality (p > 0.05). However, sample size was identified as a significant source of heterogeneity, with a p-value of 0.038 ( Table 2 ) . Table 2 Meta-regression analysis of potential sources of heterogeneity. Source of heterogeneity Coefficient p value Ethnicity -0.087 0.897 Study design -0.688 0.132 Algorithm -0.619 0.485 Sample size -1.846 0.038 Tumor size -0.654 0.214 Neuroimaging modality -0.109 0.775 Publication bias Publication bias was assessed using both funnel plot analysis and Egger’s linear regression test across the eight included studies. The funnel plot of the pooled diagnostic odds ratio for artificial intelligence-based CT methods appeared symmetrical, suggesting minimal risk of bias. Consistently, Egger’s test did not indicate significant evidence of publication bias, with a p-value of 0.383 (Fig. 7 ). Subgroup analysis The pooled diagnostic accuracy of artificial intelligence based CT methods for detecting pancreatic cancer was further analyzed through subgroup comparisons based on patient ethnicity (Asian and American), study design (retrospective and case-control), and algorithm type (convolutional neural networks and other models) (Table 3 ). Table 3 Subgroup analysis. Moderator Subgroups Diagnostic performance Sensitivity (95% CI) Specificity (95% CI) DOR (95% CI) AUC P value Ethnicity Asia 0.924(0.920- 0.983(0.981- 270.94 (60.120- 0.961 0.142 0.927) 0.984) 1221.0) America 0.887(0.863- 0.911(0.891- 81.424 (56.894- 0.959 0.908) 0.929) 116.53) Study Retrospective 0.925(0.921- 0.984(0.982- 489.44 (91.482- 0.972 0.321 design 0.928) 0.985) 2618.6) Case-control 0.887(0.863- 0.911(0.891- 81.424 (56.894- 0.961 0.908) 0.929) 116.53) ND 0.859(0.820–0.893) 0.817(0.753–0.870) 27.940(7.559–103.28) ND Algorithm CNN 0.924(0.921- 0.927) 0.991(0.990- 0.992) 557.84 (107.62- 2891.5) 0.971 0.057 Others algorithm 0.895(0.879- 0.911) 0.840(0.823- 0.856) 49.152 (30.032- 80.445) 0.941 ND: not defined. In the subgroup analysis by ethnicity, eight studies were conducted in Asian populations and two in American populations. The diagnostic accuracy of artificial intelligence based CT methods was slightly higher in studies involving Asian patients, with an area under the curve (AUC) of 0.961, compared to an AUC of 0.959 in studies involving American patients. However, this difference was not statistically significant (p = 0.142). Regarding study design, six studies were retrospective and two were case-control. Diagnostic accuracy was slightly higher in retrospective studies (AUC = 0.972) compared to case-control studies (AUC = 0.961), although the difference was not significant (p = 0.321). With respect to algorithm type, six studies used convolutional neural networks (CNN), while four employed other models including XGBoost, k-nearest neighbors (KNN), support vector machine (SVM), random forest (RM), 3D nnU-Net, and fully end-to-end artificial intelligence frameworks. CNN-based models showed a slightly higher diagnostic accuracy (AUC = 0.971) compared to the other algorithms (AUC = 0.941), but this difference also did not reach statistical significance (p = 0.057). Discussion There is growing interest in the application of artificial intelligence for cancer detection and diagnosis, particularly in enhancing the capabilities of medical imaging. Our systematic review and meta-analysis indicate that artificial intelligence methods, when applied to computed tomography, can detect pancreatic cancer with high sensitivity and specificity. The scientific community increasingly recognizes the value of data-driven approaches such as artificial intelligence to support clinicians in identifying abnormalities, while minimizing false positive and false negative findings. Integrating artificial intelligence with clinical and imaging expertise offers promising opportunities for advancing research in pancreatic cancer screening and diagnosis. According to the existing literature, there is no consistent evidence of significant differences in diagnostic accuracy among various artificial intelligence approaches used for pancreatic cancer detection [ 23 ]. This meta-analysis highlights the potential role of artificial intelligence based computed tomography in improving the diagnostic accuracy of pancreatic cancer detection. Our findings demonstrated a pooled sensitivity of 92 percent and a pooled specificity of 98 percent, with minimal variance across studies. The area under the curve (AUC) was 0.959, reflecting strong overall diagnostic performance. Despite these encouraging results, the implementation of artificial intelligence based CT in routine clinical practice remains premature. Further research is necessary to validate its effectiveness in diverse and real-world clinical settings. Artificial intelligence has emerged as one of the most transformative innovations in medical imaging over the past decade. Its influence spans the entire imaging workflow, including image acquisition, registration, processing, and interpretation [ 24 ]. However, most artificial intelligence models are trained and tested on datasets derived from similar patient populations, which may limit their generalizability. This lack of diversity in training data raises concerns about performance in broader clinical environments. Additionally, such limitations may extend to technical factors, including variations in imaging devices and protocols, as well as differences in patient demographics [ 25 ]. Most of the included studies utilized convolutional neural networks (CNNs). These models are highly effective in processing medical images and are capable of performing both generative and descriptive tasks by extracting global and local features through convolutional kernels. CNNs are a specialized form of artificial neural network designed for pixel-level data interpretation and have become standard in medical image analysis. In this review, six studies comprising a total of 28,765 patients employed CNN-based models. The pooled sensitivity and specificity for this subgroup were 0.924 (95 percent confidence interval: 0.921 to 0.927) and 0.991 (95 percent confidence interval: 0.990 to 0.992), respectively. The area under the curve was 0.971, indicating that CNN-based methods performed slightly better than the overall pooled estimates. A major strength of this meta-analysis lies in the generally high methodological quality of the included studies, as assessed using the QUADAS-2 tool. While some risk of bias was identified, particularly in the index test domain, the overall results remained robust. Although unpublished studies were not included, the funnel plot analysis did not indicate the presence of publication bias. Another strength is the comprehensive search strategy, which involved seven major databases to maximize the retrieval of relevant literature. While the findings of this study provide valuable insights into the application of artificial intelligence in detecting pancreatic cancer using computed tomography, several limitations should be noted. First, only ten original research articles met the inclusion criteria, reflecting a limited body of evidence. In many cases, insufficient data were available from the selected studies, which restricted deeper analysis. The small number of studies and their predominantly retrospective designs introduce a potential risk of selection bias. Second, the included studies reported a variety of diagnostic performance measures. A key limitation is that many did not clearly specify the diagnostic threshold or justify its selection, despite the necessity of reporting true positive, true negative, false positive, and false negative values at a defined cut-off. Third, substantial heterogeneity was observed in this meta-analysis, a common issue in diagnostic accuracy reviews [ 26 ]. This heterogeneity likely stemmed from differences in study populations, algorithm types, and interpretation of outcomes. While many studies used convolutional neural networks, others employed models such as XGBoost, k-nearest neighbors, support vector machine, random forest, 3D nnU-Net, and fully end-to-end architectures. Although meta-regression revealed that algorithm type did not significantly contribute to heterogeneity, variation in sample size appeared to be a notable source. Finally, broader challenges in the current literature include the absence of large, annotated datasets, limited interpretability of artificial intelligence models, lack of methodological standardization across studies, and a shortage of prospective research. Artificial intelligence has the potential to complement human expertise and reduce the risk of error in clinical practice by delivering consistent and rapid performance. In this context, artificial intelligence based computed tomography can support both experienced clinicians and trainees by serving as an educational and diagnostic aid. However, to accurately determine its clinical effectiveness, future research should include larger patient cohorts drawn from multiple institutions and diverse populations. Conclusion Our systematic review and meta-analysis identified a marked increase in publications over the past four years focusing on the use of artificial intelligence models for pancreatic cancer diagnosis in computed tomography images. These studies have demonstrated promising diagnostic performance, with reported sensitivity, specificity, and accuracy exceeding 90 percent. However, most of these models remain in the developmental or testing phase. Our findings suggest that artificial intelligence techniques may serve as valuable tools to support radiologists in clinical practice. Before routine implementation can be considered, prospective clinical trials and multicenter studies are needed to validate performance. Additionally, integrating explainable artificial intelligence approaches will be essential to enhance transparency, and standardizing model evaluation and reporting practices across studies will be critical for broader clinical adoption. Declarations Funding: No funding was received for this study. Data Availability: All data analyzed during this study are included in this published article and its supplementary files. Further details are available upon request from the corresponding author. Ethics, Consent to Participate, and Consent to Publish declarations: Not applicable. Competing Interests: The authors declare no competing interests. Clinical trial number: not applicable. Author Contributions: Mustafa S. Alhasan (Conceptualization, Methodology, Literature Search, Data Extraction, Statistical Analysis, Manuscript Drafting, Critical Revision) Ayman S. Alhasan (Study Design, Data Validation, Risk of Bias Assessment, Manuscript Editing, Final Approval of the Manuscript) References Rawla P, Sunkara T, Gaduputi V. Epidemiology of Pancreatic Cancer: Global Trends, Etiology and Risk Factors. World J Oncol. 2019;10:10–27. Bakasa W, Viriri S. Pancreatic Cancer Survival Prediction: A Survey of the State-of-the-Art. Comput Math Methods Med. 2021;2021:1–17. Gordon-Dseagu VL, Devesa SS, Goggins M, Stolzenberg-Solomon R. Pancreatic cancer incidence trends: evidence from the Surveillance, Epidemiology and End Results (SEER) population-based data. Int J Epidemiol. 2018;47:427–39. Maisonneuve P, Lowenfels AB. Epidemiology of Pancreatic Cancer: An Update. Dig Dis. 2010;28:645–56. Keane MG, Afghani E. A Review of the Diagnosis and Management of Premalignant Pancreatic Cystic Lesions. JCM. 2021;10:1284. Hameed BS, Krishnan UM. Artificial Intelligence-Driven Diagnosis of Pancreatic Cancer. Cancers. 2022;14:5382. Singhi AD, Koay EJ, Chari ST, Maitra A. Early Detection of Pancreatic Cancer: Opportunities and Challenges. Gastroenterology. 2019;156:2024–40. Qureshi TA, Gaddam S, Wachsman AM, et al. Predicting pancreatic ductal adenocarcinoma using artificial intelligence analysis of pre-diagnostic computed tomography images. CBM. 2022;33:211–7. Gorris M, Hoogenboom SA, Wallace MB, Van Hooft JE. Artificial intelligence for the management of pancreatic diseases. Dig Endoscopy. 2021;33:231–41. Cai J, Lu L, Zhang Z, Xing F, Yang L, Yin Q. Pancreas Segmentation in MRI Using Graph-Based Decision Fusion on Convolutional Neural Networks. In: Ourselin S, Joskowicz L, Sabuncu MR, Unal G, Wells W, editors. Medical Image Computing and Computer-Assisted Intervention – MICCAI 2016. Cham: Springer International Publishing; 2016. pp. 442–50. Page MJ, McKenzie JE, Bossuyt PM et al. (2021) The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ n71. Whiting PF. QUADAS-2: A Revised Tool for the Quality Assessment of Diagnostic Accuracy Studies. Ann Intern Med. 2011;155:529. Cao K, Xia Y, Yao J, et al. Large-scale pancreatic cancer detection via non-contrast CT and deep learning. Nat Med. 2023;29:3033–43. Chang D, Chen P-T, Wang P, et al. Detection of pancreatic cancer with two- and three-dimensional radiomic analysis in a nationwide population-based real-world dataset. BMC Cancer. 2023;23:58. Chen P-T, Wu T, Wang P, Chang D, Liu K-L, Wu M-S, Roth HR, Lee P-C, Liao W-C, Wang W. Pancreatic Cancer Detection on CT Scans with Deep Learning: A Nationwide Population-based Study. Radiology. 2023;306:172–82. Korfiatis P, Suman G, Patnam NG, et al. Automated Artificial Intelligence Model Trained on a Large Data Set Can Detect Pancreas Cancer on Diagnostic Computed Tomography Scans As Well As Visually Occult Preinvasive Cancer on Prediagnostic Computed Tomography Scans. Gastroenterology. 2023;165:1533–e15464. Liu K-L, Wu T, Chen P-T, Tsai YM, Roth H, Wu M-S, Liao W-C, Wang W. Deep learning to distinguish pancreatic cancer tissue from non-cancerous pancreatic tissue: a retrospective study with cross-racial external validation. Lancet Digit Health. 2020;2:e303–13. Ma H, Liu Z-X, Zhang J-J, Wu F-T, Xu C-F, Shen Z, Yu C-H, Li Y-M. Construction of a convolutional neural network classifier developed by computed tomography images for pancreatic cancer diagnosis. WJG. 2020;26:5156–68. Mukherjee S, Patra A, Khasawneh H, et al. Radiomics-based Machine-learning Models Can Detect Pancreatic Cancer on Prediagnostic Computed Tomography Scans at a Substantial Lead Time Before Clinical Diagnosis. Gastroenterology. 2022;163:1435–e14463. Park HJ, Shin K, You M-W, Kyung S-G, Kim SY, Park SH, Byun JH, Kim N, Kim HJ. Deep Learning–based Detection of Solid and Cystic Pancreatic Neoplasms at Contrast-enhanced CT. Radiology. 2023;306:140–9. Si K, Xue Y, Yu X, Zhu X, Li Q, Gong W, Liang T, Duan S. Fully end-to-end deep-learning-based diagnosis of pancreatic tumors. Theranostics. 2021;11:1982–90. Zhang Z, Li S, Wang Z, Lu Y. (2020) A Novel and Efficient Tumor Detection Framework for Pancreatic Cancer via CT Images. In: 2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC). IEEE, Montreal, QC, Canada, pp 1160–1164. Goyal H, Mann R, Gandhi Z, Perisetti A, Zhang Z, Sharma N, Saligram S, Inamdar S, Tharian B. Application of artificial intelligence in pancreaticobiliary diseases. Clin Med Insights Gastroenterol. 2021;14:263177452199305. Lee J-G, Jun S, Cho Y-W, Lee H, Kim GB, Seo JB, Kim N. Deep Learning in Medical Imaging: General Overview. Korean J Radiol. 2017;18:570. Sarker IH. Machine Learning: Algorithms, Real-World Applications and Research Directions. SN COMPUT SCI. 2021;2:160. Lijmer JG, Bossuyt PMM, Heisterkamp SH. Exploring sources of heterogeneity in systematic reviews of diagnostic tests. Stat Med. 2002;21:1525–37. Additional Declarations No competing interests reported. Supplementary Files SupplementaryTable1Searchstrategy.doc Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 25 Sep, 2025 Reviews received at journal 24 Sep, 2025 Reviewers agreed at journal 22 Sep, 2025 Reviews received at journal 21 Sep, 2025 Reviewers agreed at journal 21 Sep, 2025 Reviewers agreed at journal 20 Sep, 2025 Reviews received at journal 15 Sep, 2025 Reviewers agreed at journal 14 Sep, 2025 Reviewers agreed at journal 09 Sep, 2025 Reviewers agreed at journal 09 Sep, 2025 Reviewers agreed at journal 09 Sep, 2025 Reviewers agreed at journal 09 Sep, 2025 Reviewers invited by journal 09 Sep, 2025 Editor invited by journal 21 Aug, 2025 Editor assigned by journal 18 Aug, 2025 Submission checks completed at journal 18 Aug, 2025 First submitted to journal 12 Aug, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7356052","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":515098380,"identity":"3a58c6de-c9d5-469f-bac7-566fb0adfadd","order_by":0,"name":"Mustafa S. Alhasan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA8ElEQVRIiWNgGAWjYFACHjApx8DA2MbA2ABiJxCnxZiHZC2JPQwMbMRp0e0/e/Djj5o76fvZD7c9+LnjMAM/e44B448a3FrMbuQlS/Mce5bbw5PYbth75jCDZM8bA2aeY/i08BhIM7Adzu1hSGyT4G07zGBwI8eAGehI3FrOnzH++ePf4XQe/odtkn+BWuxvgBz2D4+WAzlmIMMTeCQS26TBtkjkGDDwtuFzWI6ZNW/fYcOeGw/bjWXb0nkkzjwrOMzbh99hN398OyzP3p/+7OHbNms5/vbkjQ9/fMOtBQOAo+kACRpGwSgYBaNgFGABAJIlU/ccnPJfAAAAAElFTkSuQmCC","orcid":"","institution":"Taibah University","correspondingAuthor":true,"prefix":"","firstName":"Mustafa","middleName":"S.","lastName":"Alhasan","suffix":""},{"id":515098383,"identity":"0a127a7d-fdf5-454e-9cfd-91abb35ab97e","order_by":1,"name":"Ayman S. Alhasan","email":"","orcid":"","institution":"Taibah University","correspondingAuthor":false,"prefix":"","firstName":"Ayman","middleName":"S.","lastName":"Alhasan","suffix":""}],"badges":[],"createdAt":"2025-08-12 13:08:12","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7356052/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7356052/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":91511157,"identity":"132e5411-918b-46b7-9159-c4834ece88e8","added_by":"auto","created_at":"2025-09-17 08:47:20","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":515619,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figure161.png","url":"https://assets-eu.researchsquare.com/files/rs-7356052/v1/bfe09bf45c1cba71875fc726.png"},{"id":91509491,"identity":"1805e7b9-14b9-4c13-a619-3fb4857c3261","added_by":"auto","created_at":"2025-09-17 08:39:24","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":543014,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figure162.png","url":"https://assets-eu.researchsquare.com/files/rs-7356052/v1/ecfc9d1429792818d19bd5da.png"},{"id":91509480,"identity":"e8a2e580-d130-4414-aec0-f6e1c3d34607","added_by":"auto","created_at":"2025-09-17 08:39:22","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":992792,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figure163.png","url":"https://assets-eu.researchsquare.com/files/rs-7356052/v1/8a7526a95f7a73bb939108a2.png"},{"id":91509486,"identity":"071d359c-7d70-4f22-8e09-bf5162feab0c","added_by":"auto","created_at":"2025-09-17 08:39:23","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":726011,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figure164.png","url":"https://assets-eu.researchsquare.com/files/rs-7356052/v1/02e5608eea2901ae0ba47fc9.png"},{"id":91509504,"identity":"2099ae01-28d9-47df-8d65-5dbe6934ba90","added_by":"auto","created_at":"2025-09-17 08:39:26","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":621750,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figure165.png","url":"https://assets-eu.researchsquare.com/files/rs-7356052/v1/652ea2cffc0c09e7f613973e.png"},{"id":91509506,"identity":"874ab68b-79eb-4f2a-a0a7-20ed002aa897","added_by":"auto","created_at":"2025-09-17 08:39:28","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":399868,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figure166.png","url":"https://assets-eu.researchsquare.com/files/rs-7356052/v1/58d1c888a9d1fa4ca260087e.png"},{"id":91511162,"identity":"4c8c0c5d-5d2d-450c-997f-dda47cef591d","added_by":"auto","created_at":"2025-09-17 08:47:24","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":38235,"visible":true,"origin":"","legend":"\u003cp\u003eSee image above for figure legend\u003c/p\u003e","description":"","filename":"Figure167.png","url":"https://assets-eu.researchsquare.com/files/rs-7356052/v1/cd01f31d8fb7fac8b26811ef.png"},{"id":91511213,"identity":"49e3ce7b-a2ac-45b3-9279-c1c1d103cf01","added_by":"auto","created_at":"2025-09-17 08:47:36","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4147478,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7356052/v1/6bcf77f6-5970-4228-9354-1ea8df398655.pdf"},{"id":91511163,"identity":"40cfa970-1047-446e-acd9-86b5874f8bf8","added_by":"auto","created_at":"2025-09-17 08:47:25","extension":"doc","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":37376,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryTable1Searchstrategy.doc","url":"https://assets-eu.researchsquare.com/files/rs-7356052/v1/2b01a4d6564bba4c0b202fea.doc"}],"financialInterests":"No competing interests reported.","formattedTitle":"Diagnostic Performance of Artificial Intelligence -based Computed Tomography Techniques in Detecting Pancreatic Cancer: A Systematic Review","fulltext":[{"header":"Introduction","content":"\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003ePancreatic cancer (PC) is among the most aggressive and lethal malignancies of the digestive system [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. It carries a poor prognosis, with limited therapeutic options, and a five-year relative survival rate of only 12 percent [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Globally, the incidence of pancreatic cancer has increased in recent decades. Contributing factors include aging, alcohol consumption, tobacco use, physical inactivity, obesity, diabetes, chronic pancreatitis, genetic predisposition, and prolonged exposure to environmental pollutants. Unhealthy lifestyle choices and poor dietary habits have also been implicated [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eSurgery remains the primary treatment modality for pancreatic cancer. However, several factors, including the absence of specific clinical symptoms and reliable molecular markers, often result in late-stage diagnosis, which limits the effectiveness of surgical intervention. Therefore, early detection and accurate staging are essential for improving treatment outcomes. Various imaging techniques have been explored for the diagnosis of pancreatic cancer, including endoscopic ultrasonography.\u003c/p\u003e\u003cp\u003eEndoscopic ultrasonography (EUS), computed tomography (CT), and magnetic resonance imaging (MRI) are among the primary imaging modalities used in the evaluation of pancreatic cancer [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. CT is the most widely employed technique for both detection and staging. Reported diagnostic accuracy for pancreatic cancer using CT and MRI ranges from 56 percent to 88 percent, while EUS offers superior spatial resolution [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eThe absence of specific symptoms or clinical indicators in patients with pancreatic tumors poses a major challenge to early diagnosis. When symptoms do occur, they are often nonspecific and typically emerge several months after the onset of pancreatic ductal adenocarcinoma (PDAC). Common presentations include jaundice, abdominal pain, and unexplained weight loss [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. However, the majority of patients remain asymptomatic at the time of diagnosis [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eIn this challenging clinical context, artificial intelligence techniques may support and accelerate the early diagnosis of pancreatic cancer [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. In recent years, numerous studies have been published on this topic, most of which employed a variety of algorithms and were conducted in experimental offline settings. Among these, convolutional neural networks (CNNs) are particularly prominent due to their strong capabilities in automating image analysis and processing large datasets. Artificial intelligence has shown potential as a supportive tool for radiologists in detecting both pre-neoplastic and neoplastic pancreatic lesions [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. However, evidence regarding the diagnostic performance of artificial intelligence based on CT imaging for pancreatic cancer remains limited.\u003c/p\u003e\u003cp\u003eTherefore, the aim of this manuscript was to systematically review published studies evaluating the diagnostic performance of artificial intelligence algorithms applied to computed tomography images for the detection of pancreatic cancer.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eThis systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eResources and search techniques\u003c/h2\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eThe following databases were searched from their inception through April 2024: PubMed/MEDLINE, the Cochrane Library, Web of Science, Embase, Scopus, CINAHL, and Google Scholar. The complete search strategy is provided in Supplementary File 1.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eSelection criteria\u003c/h3\u003e\n\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eAfter removing duplicates, articles were screened based on titles and abstracts. Studies investigating the diagnosis of pancreatic cancer using artificial intelligence applied to computed tomography were considered for inclusion. Full texts of potentially eligible studies were then reviewed to confirm eligibility.\u003c/p\u003e\u003cp\u003eInclusion criteria were as follows:\u003c/p\u003e\u003cp\u003e(1) observational studies reporting the diagnostic performance of artificial intelligence applied exclusively to CT imaging for pancreatic cancer diagnosis;\u003c/p\u003e\u003cp\u003e(2) studies involving patients with pancreatic cancer confirmed by CT;\u003c/p\u003e\u003cp\u003e(3) publications reporting sensitivity and specificity outcomes; and\u003c/p\u003e\u003cp\u003e(4) original research articles.\u003c/p\u003e\u003cp\u003eExclusion criteria were as follows:\u003c/p\u003e\u003cp\u003e(1) studies published in languages other than English;\u003c/p\u003e\u003cp\u003e(2) lack of full-text availability in electronic form;\u003c/p\u003e\u003cp\u003e(3) insufficient outcome data; and\u003c/p\u003e\u003cp\u003e(4) publication types such as comments, letters, editorials, protocols, guidelines, review papers, non-peer-reviewed articles, and conference abstracts.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\n\u003ch3\u003eData extraction\u003c/h3\u003e\n\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eBased on the established inclusion and exclusion criteria, two independent authors extracted data from the eligible studies and recorded the information using a standardized data sheet. The following variables were collected: (1) article title, (2) country of origin, (3) study design, (4) sample size, (5) number of cases, (6) number of controls, (7) patient age, (8) gender distribution, (9) tumor size in centimeters, (10) neuroimaging modality, and (11) type of algorithm used.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\n\u003ch3\u003eRisk of bias assessment\u003c/h3\u003e\n\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eThe methodological quality of the included studies was independently assessed by two authors using the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. This tool evaluates four domains related to risk of bias and applicability: patient selection, index test, reference standard, and flow and timing. Each item was rated as \u0026ldquo;yes,\u0026rdquo; \u0026ldquo;no,\u0026rdquo; or \u0026ldquo;unclear.\u0026rdquo; A response of \u0026ldquo;yes\u0026rdquo; indicated a low risk of bias, while \u0026ldquo;no\u0026rdquo; or \u0026ldquo;unclear\u0026rdquo; responses reflected potential bias. The results of the assessment were visualized using Review Manager (RevMan) version 5.4 (Cochrane Collaboration, Oxford, United Kingdom).\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\n\u003ch3\u003eOutcome measures\u003c/h3\u003e\n\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eSensitivity, specificity, accuracy, and diagnostic odds ratio (DOR) were evaluated to determine the diagnostic performance of artificial intelligence methods in detecting pancreatic cancer.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eStatistical analysis\u003c/h2\u003e\u003cp\u003eTo analyze pooled sensitivity, specificity, diagnostic odds ratio (DOR), and area under the curve (AUC) with 95 percent confidence intervals, this meta-analysis was performed using Meta-Disc version 1.4 and Comprehensive Meta-Analysis version 3 (Biostat Inc., USA). A two-sided p-value less than 0.05 was considered statistically significant. Diagnostic performance was evaluated using the summary receiver operating characteristic (SROC) curve, derived from the sensitivity and specificity values reported in individual studies.\u003c/p\u003e\u003cp\u003eHeterogeneity across studies was assessed using the Cochrane chi-squared test, with a p-value less than 0.05 indicating statistical heterogeneity. The I-squared statistic was also calculated, with values greater than 50 percent indicating substantial heterogeneity. To explore the presence of a threshold effect, the Spearman correlation coefficient was computed. A strong positive correlation would suggest the presence of a threshold effect.\u003c/p\u003e\u003cp\u003eSensitivity analysis was conducted to assess the robustness of the findings. Publication bias was evaluated using Egger\u0026rsquo;s test, and further assessed through visual inspection of funnel plot symmetry.\u003c/p\u003e\u003c/div\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003eIdentification of studies\u003c/h2\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eThe database search yielded 462 records for initial screening. Of these, 149 abstracts were deemed potentially eligible and were retrieved for full-text review. Following a detailed assessment, 10 articles met the eligibility criteria and were included in this systematic review and meta-analysis. The study selection process is illustrated in the PRISMA flowchart \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e\u003cb\u003e)\u003c/b\u003e.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003eCharacteristics of included studies\u003c/h2\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eThe included articles were published between 2020 and 2023 and originated from six countries: China (n\u0026thinsp;=\u0026thinsp;4), Taiwan (n\u0026thinsp;=\u0026thinsp;3), the United States (n\u0026thinsp;=\u0026thinsp;2), and the Republic of Korea (n\u0026thinsp;=\u0026thinsp;1). Among the ten studies included in this systematic review and meta-analysis, six were retrospective in design, two were case-control studies, and two did not report their study design. Sample sizes ranged from 412 to 20,530 participants. The majority of studies (seven out of ten) utilized contrast-enhanced computed tomography. Convolutional neural networks were the most frequently used algorithm, employed in six of the ten studies. A summary of study characteristics is provided in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eCharacteristics of included studies\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"11\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c10\" colnum=\"10\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c11\" colnum=\"11\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eArticle\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCountry\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStudy design\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSample size\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eCases\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eControls\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eAge\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eGender M/F\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c9\"\u003e\u003cp\u003eTumor size (cm)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c10\"\u003e\u003cp\u003eNeuroimaging modalities\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c11\"\u003e\u003cp\u003eAlgorithm\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCao et al. 2023 [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eChina\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eRetrospective\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e20,530\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003enon-contrast CT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c11\"\u003e\u003cp\u003eCNN\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eChang et al. 2023 [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTaiwan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eRetrospective\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1279\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e546 patients with pancreatic adenocarcinoma\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e733 controls with normal pancreas\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eCase: 64.6\u0026thinsp;\u0026plusmn;\u0026thinsp;11.8\u003c/p\u003e\u003cp\u003eControl: 54.2\u0026thinsp;\u0026plusmn;\u0026thinsp;16.2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eCase:297/249\u003c/p\u003e\u003cp\u003eControl:374/359\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e-\u0026lt; 2cm: 24.7%\u003c/p\u003e\u003cp\u003e-2\u0026ndash;4cm: 49.5%\u003c/p\u003e\u003cp\u003e- \u0026gt;4cm: 25.8%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eCE-CT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c11\"\u003e\u003cp\u003eXGBoost\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eChen et al. 2023 [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTaiwan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eRetrospective\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e1229\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e546 patients with pancreatic cancer\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e683\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eCase: 65\u0026thinsp;\u0026plusmn;\u0026thinsp;12\u003c/p\u003e\u003cp\u003eControl: 54\u0026thinsp;\u0026plusmn;\u0026thinsp;16\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eCase:297/249\u003c/p\u003e\u003cp\u003eControl:374/309\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e2.9 (2.1\u0026ndash;4.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eCE-CT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c11\"\u003e\u003cp\u003eCNN\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eKorfiatis et al. 2023 [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eUSA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCase-control\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3014\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1105 patients with treatment-na\u0026iuml;ve biopsy proven\u003c/p\u003e\u003cp\u003ePDA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e1909 controls\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eCase: 66\u003c/p\u003e\u003cp\u003eControl: 56\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eCase:627/478\u003c/p\u003e\u003cp\u003eControl:872/1037\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e4.9 (1.7)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eCT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c11\"\u003e\u003cp\u003e3D-CNN\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLiu et al. 2020 [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTaiwan\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eRetrospective\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e690\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e370 patients with pancreatic\u003c/p\u003e\u003cp\u003ecancer\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e320 controls\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e64\u0026middot;8\u0026thinsp;\u0026plusmn;\u0026thinsp;12\u0026middot;0\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e260/211\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e3\u0026middot;0 (2\u0026middot;2\u0026ndash;4\u0026middot;6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eCE-CT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c11\"\u003e\u003cp\u003eCNN\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMa et al. 2020 [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eChina\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eRetrospective\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e412\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e222 with pancreatic cancer\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e190 diagnosed with normal pancreas\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eCase: 63.8 (8.7)\u003c/p\u003e\u003cp\u003eControl: 61.0 (12.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eCase: 124/98\u003c/p\u003e\u003cp\u003eControl: 98/92\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003e3.5 (2.7\u0026ndash;4.3)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eCT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c11\"\u003e\u003cp\u003eCNN\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMukherjee et al. 2022 [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eUSA\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eCase-control\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e420\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e155\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e265\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eCase: 69\u003c/p\u003e\u003cp\u003eControl: 67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eCase: ratio 1.4\u003c/p\u003e\u003cp\u003eControl: ratio 1.1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eCE-CT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c11\"\u003e\u003cp\u003eKNN, SVM, RM,\u003c/p\u003e\u003cp\u003eXGBoost\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePark et al. 2023 [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRepublic of Korea\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eRetrospective\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2044\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2044 patients with pancreatic\u003c/p\u003e\u003cp\u003elesions\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003e60.33\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003e1181/863\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eCE-CT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c11\"\u003e\u003cp\u003e3D nnU-Net\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSi et al. 2021 [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eChina\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e666\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e549 patients with abnormal CT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e117 controls\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eTraining dataset: 63.3\u003c/p\u003e\u003cp\u003eTesting dataset: 61.8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eTraining dataset: 211/108\u003c/p\u003e\u003cp\u003eTesting dataset: 216/131\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eCE-CT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c11\"\u003e\u003cp\u003efully end-to-end deep-learning\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eZhang et al. 2020 [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eChina\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e2890\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2890 CT images of patients diagnosed\u003c/p\u003e\u003cp\u003ewith pancreatic cancer\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c8\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c9\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c10\"\u003e\u003cp\u003eCE-CT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c11\"\u003e\u003cp\u003eDeep CNN\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"11\"\u003eCT: computed tomography; CE-CT: contrast-enhanced computed tomography; PDAC: pancreatic ductal adenocarcinoma; CNN: convolutional neural network; ND: not defined; XGBoost: extreme gradient boosting; RM: random forest; SVM: support vector machine; KNN: k-nearest neighbor.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eRisk of bias assessment\u003c/h2\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eThe methodological quality of the ten included studies was evaluated using the QUADAS-2 tool. In the domain of patient selection, two studies were identified as having a high risk of bias, while three studies had an unclear risk. A notable concentration of bias was observed in the index test domain, where five studies preset diagnostic thresholds without justification. The domains of reference standard and flow and timing were not associated with significant risk of bias.\u003c/p\u003e\u003cp\u003eConcerns regarding applicability were also present. Specifically, two studies raised concerns in the patient selection domain, and four studies showed applicability issues in the index test domain. A detailed summary of the risk of bias and applicability assessments is provided in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003eData analysis\u003c/h2\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eAnalysis of the forest plots revealed significant heterogeneity across studies for sensitivity, specificity, and diagnostic odds ratio (DOR). Specifically, sensitivity showed a Chi-squared value of 108.66 with a p-value of 0.000 and an I-squared of 87.1 percent \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e\u003cb\u003e)\u003c/b\u003e. Specificity demonstrated a Chi-squared value of 1677 with a p-value of 0.000 and an I-squared of 99.2 percent \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e\u003cb\u003e)\u003c/b\u003e. The DOR analysis yielded a Chi-squared value of 946.42 with a p-value of 0.000 and an I-squared of 98.5 percent \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e\u003cb\u003e)\u003c/b\u003e. Due to this substantial heterogeneity, pooled estimates were calculated using a random effects model.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThe pooled sensitivity of artificial intelligence-based CT methods for diagnosing pancreatic cancer was 0.92 with a 95 percent confidence interval of 0.92 to 0.93 \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e\u003cb\u003e)\u003c/b\u003e. The pooled specificity was 0.98 with a 95 percent confidence interval of 0.98 to 0.98 \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e\u003cb\u003e)\u003c/b\u003e, and the DOR was 179.57 with a 95 percent confidence interval of 57.98 to 556.16 \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e\u003cb\u003e)\u003c/b\u003e. The summary receiver operating characteristic (SROC) curve demonstrated an area under the curve (AUC) of 0.959, indicating excellent diagnostic accuracy \u003cb\u003e(\u003c/b\u003eFig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e\u003cb\u003e)\u003c/b\u003e.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003eInvestigation for heterogeneity\u003c/h2\u003e\u003cp\u003e\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eHeterogeneity among the included studies was evident in the forest plots. As heterogeneity cannot be entirely eliminated in meta-analyses, its extent and potential sources were further explored. In diagnostic accuracy studies, a common source of heterogeneity is the threshold effect. This was assessed using the Spearman correlation coefficient and the Moses model, weighted by inverse variance. The Spearman coefficient was \u0026minus;\u0026thinsp;0.356 with a p-value of 0.193, indicating that the threshold effect did not significantly contribute to differences in diagnostic accuracy across studies.\u003c/p\u003e\u003cp\u003eTo identify other potential sources of heterogeneity, a meta-regression analysis was conducted based on algorithm type, study design, patient ethnicity, sample size, tumor size, and imaging modality. The analysis showed no significant influence of algorithm type, study design, ethnicity, tumor size, or neuroimaging modality (p\u0026thinsp;\u0026gt;\u0026thinsp;0.05). However, sample size was identified as a significant source of heterogeneity, with a p-value of 0.038 \u003cb\u003e(\u003c/b\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e\u003cb\u003e)\u003c/b\u003e.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eMeta-regression analysis of potential sources of heterogeneity.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"3\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSource of heterogeneity\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCoefficient\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003ep value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEthnicity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.087\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.897\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStudy design\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.688\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.132\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAlgorithm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.619\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.485\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSample size\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-1.846\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.038\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTumor size\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.654\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.214\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNeuroimaging modality\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e-0.109\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.775\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003ePublication bias\u003c/h2\u003e\u003cp\u003ePublication bias was assessed using both funnel plot analysis and Egger\u0026rsquo;s linear regression test across the eight included studies. The funnel plot of the pooled diagnostic odds ratio for artificial intelligence-based CT methods appeared symmetrical, suggesting minimal risk of bias. Consistently, Egger\u0026rsquo;s test did not indicate significant evidence of publication bias, with a p-value of 0.383 (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003eSubgroup analysis\u003c/h2\u003e\u003cp\u003eThe pooled diagnostic accuracy of artificial intelligence based CT methods for detecting pancreatic cancer was further analyzed through subgroup comparisons based on patient ethnicity (Asian and American), study design (retrospective and case-control), and algorithm type (convolutional neural networks and other models) (Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eSubgroup analysis.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"7\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eModerator\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eSubgroups\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colspan=\"5\" nameend=\"c7\" namest=\"c3\"\u003e\u003cp\u003eDiagnostic performance\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eSensitivity\u003c/p\u003e\u003cp\u003e(95% CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eSpecificity\u003c/p\u003e\u003cp\u003e(95% CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eDOR (95% CI)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eAUC\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003eP value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEthnicity\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAsia\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.924(0.920-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.983(0.981-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e270.94 (60.120-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.961\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.142\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.927)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.984)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e1221.0)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\" morerows=\"1\" rowspan=\"2\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAmerica\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.887(0.863-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.911(0.891-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e81.424 (56.894-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.959\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.908)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.929)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e116.53)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStudy\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eRetrospective\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.925(0.921-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.984(0.982-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e489.44 (91.482-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.972\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e0.321\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003edesign\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.928)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.985)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e2618.6)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c7\" morerows=\"3\" rowspan=\"4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCase-control\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.887(0.863-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.911(0.891-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e81.424 (56.894-\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.961\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.908)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.929)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e116.53)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.859(0.820\u0026ndash;0.893)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.817(0.753\u0026ndash;0.870)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e27.940(7.559\u0026ndash;103.28)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003eND\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eAlgorithm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCNN\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.924(0.921-\u003c/p\u003e\u003cp\u003e0.927)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.991(0.990-\u003c/p\u003e\u003cp\u003e0.992)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e557.84 (107.62-\u003c/p\u003e\u003cp\u003e2891.5)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.971\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003e0.057\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eOthers\u003c/p\u003e\u003cp\u003ealgorithm\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e0.895(0.879-\u003c/p\u003e\u003cp\u003e0.911)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.840(0.823-\u003c/p\u003e\u003cp\u003e0.856)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c5\"\u003e\u003cp\u003e49.152 (30.032-\u003c/p\u003e\u003cp\u003e80.445)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c6\"\u003e\u003cp\u003e0.941\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"7\"\u003eND: not defined.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eIn the subgroup analysis by ethnicity, eight studies were conducted in Asian populations and two in American populations. The diagnostic accuracy of artificial intelligence based CT methods was slightly higher in studies involving Asian patients, with an area under the curve (AUC) of 0.961, compared to an AUC of 0.959 in studies involving American patients. However, this difference was not statistically significant (p\u0026thinsp;=\u0026thinsp;0.142).\u003c/p\u003e\u003cp\u003eRegarding study design, six studies were retrospective and two were case-control. Diagnostic accuracy was slightly higher in retrospective studies (AUC\u0026thinsp;=\u0026thinsp;0.972) compared to case-control studies (AUC\u0026thinsp;=\u0026thinsp;0.961), although the difference was not significant (p\u0026thinsp;=\u0026thinsp;0.321).\u003c/p\u003e\u003cp\u003eWith respect to algorithm type, six studies used convolutional neural networks (CNN), while four employed other models including XGBoost, k-nearest neighbors (KNN), support vector machine (SVM), random forest (RM), 3D nnU-Net, and fully end-to-end artificial intelligence frameworks. CNN-based models showed a slightly higher diagnostic accuracy (AUC\u0026thinsp;=\u0026thinsp;0.971) compared to the other algorithms (AUC\u0026thinsp;=\u0026thinsp;0.941), but this difference also did not reach statistical significance (p\u0026thinsp;=\u0026thinsp;0.057).\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003eThere is growing interest in the application of artificial intelligence for cancer detection and diagnosis, particularly in enhancing the capabilities of medical imaging. Our systematic review and meta-analysis indicate that artificial intelligence methods, when applied to computed tomography, can detect pancreatic cancer with high sensitivity and specificity. The scientific community increasingly recognizes the value of data-driven approaches such as artificial intelligence to support clinicians in identifying abnormalities, while minimizing false positive and false negative findings. Integrating artificial intelligence with clinical and imaging expertise offers promising opportunities for advancing research in pancreatic cancer screening and diagnosis. According to the existing literature, there is no consistent evidence of significant differences in diagnostic accuracy among various artificial intelligence approaches used for pancreatic cancer detection [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eThis meta-analysis highlights the potential role of artificial intelligence based computed tomography in improving the diagnostic accuracy of pancreatic cancer detection. Our findings demonstrated a pooled sensitivity of 92 percent and a pooled specificity of 98 percent, with minimal variance across studies. The area under the curve (AUC) was 0.959, reflecting strong overall diagnostic performance. Despite these encouraging results, the implementation of artificial intelligence based CT in routine clinical practice remains premature. Further research is necessary to validate its effectiveness in diverse and real-world clinical settings.\u003c/p\u003e\u003cp\u003eArtificial intelligence has emerged as one of the most transformative innovations in medical imaging over the past decade. Its influence spans the entire imaging workflow, including image acquisition, registration, processing, and interpretation [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. However, most artificial intelligence models are trained and tested on datasets derived from similar patient populations, which may limit their generalizability. This lack of diversity in training data raises concerns about performance in broader clinical environments. Additionally, such limitations may extend to technical factors, including variations in imaging devices and protocols, as well as differences in patient demographics [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eMost of the included studies utilized convolutional neural networks (CNNs). These models are highly effective in processing medical images and are capable of performing both generative and descriptive tasks by extracting global and local features through convolutional kernels. CNNs are a specialized form of artificial neural network designed for pixel-level data interpretation and have become standard in medical image analysis. In this review, six studies comprising a total of 28,765 patients employed CNN-based models. The pooled sensitivity and specificity for this subgroup were 0.924 (95 percent confidence interval: 0.921 to 0.927) and 0.991 (95 percent confidence interval: 0.990 to 0.992), respectively. The area under the curve was 0.971, indicating that CNN-based methods performed slightly better than the overall pooled estimates.\u003c/p\u003e\u003cp\u003eA major strength of this meta-analysis lies in the generally high methodological quality of the included studies, as assessed using the QUADAS-2 tool. While some risk of bias was identified, particularly in the index test domain, the overall results remained robust. Although unpublished studies were not included, the funnel plot analysis did not indicate the presence of publication bias. Another strength is the comprehensive search strategy, which involved seven major databases to maximize the retrieval of relevant literature.\u003c/p\u003e\u003cp\u003eWhile the findings of this study provide valuable insights into the application of artificial intelligence in detecting pancreatic cancer using computed tomography, several limitations should be noted. First, only ten original research articles met the inclusion criteria, reflecting a limited body of evidence. In many cases, insufficient data were available from the selected studies, which restricted deeper analysis. The small number of studies and their predominantly retrospective designs introduce a potential risk of selection bias.\u003c/p\u003e\u003cp\u003eSecond, the included studies reported a variety of diagnostic performance measures. A key limitation is that many did not clearly specify the diagnostic threshold or justify its selection, despite the necessity of reporting true positive, true negative, false positive, and false negative values at a defined cut-off.\u003c/p\u003e\u003cp\u003eThird, substantial heterogeneity was observed in this meta-analysis, a common issue in diagnostic accuracy reviews [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. This heterogeneity likely stemmed from differences in study populations, algorithm types, and interpretation of outcomes. While many studies used convolutional neural networks, others employed models such as XGBoost, k-nearest neighbors, support vector machine, random forest, 3D nnU-Net, and fully end-to-end architectures. Although meta-regression revealed that algorithm type did not significantly contribute to heterogeneity, variation in sample size appeared to be a notable source.\u003c/p\u003e\u003cp\u003eFinally, broader challenges in the current literature include the absence of large, annotated datasets, limited interpretability of artificial intelligence models, lack of methodological standardization across studies, and a shortage of prospective research.\u003cdiv class=\"BlockQuote\"\u003e\u003cp\u003eArtificial intelligence has the potential to complement human expertise and reduce the risk of error in clinical practice by delivering consistent and rapid performance. In this context, artificial intelligence based computed tomography can support both experienced clinicians and trainees by serving as an educational and diagnostic aid. However, to accurately determine its clinical effectiveness, future research should include larger patient cohorts drawn from multiple institutions and diverse populations.\u003c/p\u003e\u003c/div\u003e\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eOur systematic review and meta-analysis identified a marked increase in publications over the past four years focusing on the use of artificial intelligence models for pancreatic cancer diagnosis in computed tomography images. These studies have demonstrated promising diagnostic performance, with reported sensitivity, specificity, and accuracy exceeding 90 percent. However, most of these models remain in the developmental or testing phase. Our findings suggest that artificial intelligence techniques may serve as valuable tools to support radiologists in clinical practice. Before routine implementation can be considered, prospective clinical trials and multicenter studies are needed to validate performance. Additionally, integrating explainable artificial intelligence approaches will be essential to enhance transparency, and standardizing model evaluation and reporting practices across studies will be critical for broader clinical adoption.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eFunding:\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;No funding was received for this study.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability:\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;All data analyzed during this study are included in this published article and its supplementary files. Further details are available upon request from the corresponding author.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics, Consent to Participate, and Consent to Publish declarations:\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;Not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interests:\u003c/strong\u003e\u003cbr\u003e\u0026nbsp;The authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial number:\u003c/strong\u003e not applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor Contributions:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMustafa S. Alhasan (Conceptualization, Methodology, Literature Search, Data Extraction, Statistical Analysis, Manuscript Drafting, Critical Revision)\u003cbr\u003e\u0026nbsp;Ayman S. Alhasan (Study Design, Data Validation, Risk of Bias Assessment, Manuscript Editing, Final Approval of the Manuscript)\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eRawla P, Sunkara T, Gaduputi V. Epidemiology of Pancreatic Cancer: Global Trends, Etiology and Risk Factors. World J Oncol. 2019;10:10\u0026ndash;27.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBakasa W, Viriri S. Pancreatic Cancer Survival Prediction: A Survey of the State-of-the-Art. Comput Math Methods Med. 2021;2021:1\u0026ndash;17.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGordon-Dseagu VL, Devesa SS, Goggins M, Stolzenberg-Solomon R. Pancreatic cancer incidence trends: evidence from the Surveillance, Epidemiology and End Results (SEER) population-based data. Int J Epidemiol. 2018;47:427\u0026ndash;39.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMaisonneuve P, Lowenfels AB. Epidemiology of Pancreatic Cancer: An Update. Dig Dis. 2010;28:645\u0026ndash;56.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKeane MG, Afghani E. A Review of the Diagnosis and Management of Premalignant Pancreatic Cystic Lesions. JCM. 2021;10:1284.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHameed BS, Krishnan UM. Artificial Intelligence-Driven Diagnosis of Pancreatic Cancer. Cancers. 2022;14:5382.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSinghi AD, Koay EJ, Chari ST, Maitra A. Early Detection of Pancreatic Cancer: Opportunities and Challenges. Gastroenterology. 2019;156:2024\u0026ndash;40.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eQureshi TA, Gaddam S, Wachsman AM, et al. Predicting pancreatic ductal adenocarcinoma using artificial intelligence analysis of pre-diagnostic computed tomography images. CBM. 2022;33:211\u0026ndash;7.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGorris M, Hoogenboom SA, Wallace MB, Van Hooft JE. Artificial intelligence for the management of pancreatic diseases. Dig Endoscopy. 2021;33:231\u0026ndash;41.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCai J, Lu L, Zhang Z, Xing F, Yang L, Yin Q. Pancreas Segmentation in MRI Using Graph-Based Decision Fusion on Convolutional Neural Networks. In: Ourselin S, Joskowicz L, Sabuncu MR, Unal G, Wells W, editors. Medical Image Computing and Computer-Assisted Intervention \u0026ndash; MICCAI 2016. Cham: Springer International Publishing; 2016. pp. 442\u0026ndash;50.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePage MJ, McKenzie JE, Bossuyt PM et al. (2021) The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ n71.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWhiting PF. QUADAS-2: A Revised Tool for the Quality Assessment of Diagnostic Accuracy Studies. Ann Intern Med. 2011;155:529.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCao K, Xia Y, Yao J, et al. Large-scale pancreatic cancer detection via non-contrast CT and deep learning. Nat Med. 2023;29:3033\u0026ndash;43.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChang D, Chen P-T, Wang P, et al. Detection of pancreatic cancer with two- and three-dimensional radiomic analysis in a nationwide population-based real-world dataset. BMC Cancer. 2023;23:58.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen P-T, Wu T, Wang P, Chang D, Liu K-L, Wu M-S, Roth HR, Lee P-C, Liao W-C, Wang W. Pancreatic Cancer Detection on CT Scans with Deep Learning: A Nationwide Population-based Study. Radiology. 2023;306:172\u0026ndash;82.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKorfiatis P, Suman G, Patnam NG, et al. Automated Artificial Intelligence Model Trained on a Large Data Set Can Detect Pancreas Cancer on Diagnostic Computed Tomography Scans As Well As Visually Occult Preinvasive Cancer on Prediagnostic Computed Tomography Scans. Gastroenterology. 2023;165:1533\u0026ndash;e15464.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu K-L, Wu T, Chen P-T, Tsai YM, Roth H, Wu M-S, Liao W-C, Wang W. Deep learning to distinguish pancreatic cancer tissue from non-cancerous pancreatic tissue: a retrospective study with cross-racial external validation. Lancet Digit Health. 2020;2:e303\u0026ndash;13.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMa H, Liu Z-X, Zhang J-J, Wu F-T, Xu C-F, Shen Z, Yu C-H, Li Y-M. Construction of a convolutional neural network classifier developed by computed tomography images for pancreatic cancer diagnosis. WJG. 2020;26:5156\u0026ndash;68.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMukherjee S, Patra A, Khasawneh H, et al. Radiomics-based Machine-learning Models Can Detect Pancreatic Cancer on Prediagnostic Computed Tomography Scans at a Substantial Lead Time Before Clinical Diagnosis. Gastroenterology. 2022;163:1435\u0026ndash;e14463.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePark HJ, Shin K, You M-W, Kyung S-G, Kim SY, Park SH, Byun JH, Kim N, Kim HJ. Deep Learning\u0026ndash;based Detection of Solid and Cystic Pancreatic Neoplasms at Contrast-enhanced CT. Radiology. 2023;306:140\u0026ndash;9.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSi K, Xue Y, Yu X, Zhu X, Li Q, Gong W, Liang T, Duan S. Fully end-to-end deep-learning-based diagnosis of pancreatic tumors. Theranostics. 2021;11:1982\u0026ndash;90.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Z, Li S, Wang Z, Lu Y. (2020) A Novel and Efficient Tumor Detection Framework for Pancreatic Cancer via CT Images. In: 2020 42nd Annual International Conference of the IEEE Engineering in Medicine \u0026amp; Biology Society (EMBC). IEEE, Montreal, QC, Canada, pp 1160\u0026ndash;1164.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGoyal H, Mann R, Gandhi Z, Perisetti A, Zhang Z, Sharma N, Saligram S, Inamdar S, Tharian B. Application of artificial intelligence in pancreaticobiliary diseases. Clin Med Insights Gastroenterol. 2021;14:263177452199305.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLee J-G, Jun S, Cho Y-W, Lee H, Kim GB, Seo JB, Kim N. Deep Learning in Medical Imaging: General Overview. Korean J Radiol. 2017;18:570.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSarker IH. Machine Learning: Algorithms, Real-World Applications and Research Directions. SN COMPUT SCI. 2021;2:160.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLijmer JG, Bossuyt PMM, Heisterkamp SH. Exploring sources of heterogeneity in systematic reviews of diagnostic tests. Stat Med. 2002;21:1525\u0026ndash;37.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-cancer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bcan","sideBox":"Learn more about [BMC Cancer](http://bmccancer.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bcan/default.aspx","title":"BMC Cancer","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Artificial intelligence, Pancreatic cancer, Computed tomography, Diagnostic accuracy, Systematic review","lastPublishedDoi":"10.21203/rs.3.rs-7356052/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7356052/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eIntroduction:\u003c/h2\u003e\u003cp\u003ePancreatic cancer is among the most lethal malignancies, often diagnosed at an advanced stage due to its subtle clinical presentation. Artificial intelligence (AI) techniques applied to computed tomography (CT) have shown potential in improving diagnostic accuracy across various imaging tasks, including cancer detection and lesion classification. This systematic review and meta-analysis aimed to evaluate the diagnostic performance of AI-based CT models specifically for the detection of pancreatic cancer, while accounting for the clinical task type and algorithmic heterogeneity among studies.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e\u003cp\u003eA comprehensive literature search was conducted across PubMed/MEDLINE, the Cochrane Library, Web of Science, Embase, Scopus, CINAHL, and Google Scholar to identify relevant studies published up to April 2025. Eligible studies were screened according to predefined criteria, and diagnostic performance data including true positives, false positives, true negatives, and false negatives were extracted or calculated as needed from the selected publications.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e\u003cp\u003eTen studies comprising a total of 33,174 patients met the inclusion criteria. The pooled diagnostic sensitivity and specificity were 0.92 with a 95 percent confidence interval of 0.92 to 0.93, and 0.98 with a 95 percent confidence interval of 0.98 to 0.98, respectively. The area under the summary receiver operating characteristic curve was 0.959, and the diagnostic odds ratio was 179.57 with a 95 percent confidence interval of 57.98 to 556.16, indicating high diagnostic accuracy for pancreatic cancer detection. Significant heterogeneity was observed among the studies included. However, subgroup analyses did not reveal any statistically significant differences. No evidence of publication bias was detected.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e\u003cp\u003eThe findings of this study suggest that artificial intelligence assisted computed tomography may serve as a valuable tool in supporting the diagnosis of pancreatic cancer. However, further validation using larger and more diverse datasets is necessary to confirm its clinical utility and generalizability.\u003c/p\u003e","manuscriptTitle":"Diagnostic Performance of Artificial Intelligence -based Computed Tomography Techniques in Detecting Pancreatic Cancer: A Systematic Review","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-17 08:38:38","doi":"10.21203/rs.3.rs-7356052/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-09-25T07:28:46+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-24T10:28:32+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"159833853453775464369941826154298922121","date":"2025-09-22T09:26:49+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-22T02:21:17+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"211916725596014767651117758073922371169","date":"2025-09-21T20:19:43+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"271842373995475953503247107235883001602","date":"2025-09-20T13:45:15+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-15T07:09:26+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"14342548031893998872543146259204263551","date":"2025-09-14T18:10:13+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"147494116013119766709904988132150468642","date":"2025-09-09T14:38:43+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"227755341152311906072655332828396000672","date":"2025-09-09T13:40:42+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"223823994663819890912998256328846297086","date":"2025-09-09T12:02:06+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"31452103197475840237131644533074536098","date":"2025-09-09T11:04:58+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-09-09T11:00:54+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-08-21T14:49:23+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-19T03:08:32+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-08-19T03:08:14+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Cancer","date":"2025-08-12T12:54:15+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-cancer","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"bcan","sideBox":"Learn more about [BMC Cancer](http://bmccancer.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/bcan/default.aspx","title":"BMC Cancer","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"6861546f-2453-4d71-a71c-5022a434d075","owner":[],"postedDate":"September 17th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-10-14T07:23:11+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-17 08:38:38","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7356052","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7356052","identity":"rs-7356052","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}