An optimal diagnosis model for pancreatic cancer prediction based on deep learning method
PeerJ Computer Science · 27.01.2026 · 12 · e3541
DOI: 10.7717/peerj-cs.3541 ↗Kurz gefasst
Fachartikel in der Zeitschrift PeerJ Computer Science, online erschienen am 27.01.2026. Verfasst von Venkatesh C. und 7 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Deanship of Research and Graduate Studies at King Khalid University, Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia, Deanship of Scientific Research at Northern Border University, Arar, KSA.
Abstract (Originalsprache)
Carcinoma of the pancreas is one of the deadliest malignant neoplasms due to its late detection, high mortality rate, and poor prognosis. Pancreatic tumors show significant variation in location, size, and shape, complicating accurate diagnosis. Moreover, they are often small and embedded within tissues of similar intensity in computed tomography (CT) images, making differentiation from healthy tissue difficult. Since biopsies and other pathological tests are not routinely performed in clinical settings, there is a growing need for non-invasive and repeatable diagnostic solutions. In current practice, CT images used for pancreatic cancer analysis require manual outlining, which is both time-consuming and subjective. Manual segmentation is inefficient for clinical use, thus emphasizing the need for a robust automated segmentation system. Both radiologists and existing algorithms struggle to accurately identify cancerous regions due to the visual similarity between tumor and normal tissue in CT scans. To address these challenges, this study proposes an automated pancreatic cancer detection system that integrates deep learning with optimization-based segmentation. This combined approach enhances threshold selection and improves segmentation accuracy in complex tumor regions. Specifically, multilevel thresholding guided by an optimization technique is applied to detect pancreatic cancer at an early stage, thereby helping reduce mortality rates. The optimization strategy is used to identify optimal parameters for maximizing the prediction rate using a deep learning classifier such as convolutional neural networks (CNN). The proposed system achieved an accuracy of 95.92%. Additionally, evaluation metrics including peak signal-to-noise ratio (PSNR), sensitivity, specificity, and mean square error (MSE) were calculated and compared with those of existing systems. This study demonstrates that the proposed particle swarm optimization (PSO)–CNN–based automated system offers a promising approach for reliable early detection and improved diagnostic accuracy in pancreatic cancer.
Förderung
- Deanship of Research and Graduate Studies at King Khalid University
- Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia
- Deanship of Scientific Research at Northern Border University, Arar, KSA
Zitieren
APA
C., V., Al-Wesabi, F. N., Alshahrani, H. M., Yahya, A. E., Kiran, A., J., C. B., L., S., & Alam, A. (2026). An optimal diagnosis model for pancreatic cancer prediction based on deep learning method. PeerJ Computer Science, 12, e3541. https://doi.org/10.7717/peerj-cs.3541
Vancouver
C. V, Al-Wesabi FN, Alshahrani HM, Yahya AE, Kiran A, J. CB, et al. An optimal diagnosis model for pancreatic cancer prediction based on deep learning method. PeerJ Computer Science. 2026;12:e3541. doi:10.7717/peerj-cs.3541
BibTeX
@article{C2026,
author = {C., Venkatesh and Al-Wesabi, Fahd N. and Alshahrani, Haya Mesfer and Yahya, Abdulsamad Ebrahim and Kiran, Ajmeera and J., Chinna Babu and L., Sivayamini and Alam, Afroj},
title = {An optimal diagnosis model for pancreatic cancer prediction based on deep learning method},
journal = {PeerJ Computer Science},
year = {2026},
volume = {12},
pages = {e3541},
doi = {10.7717/peerj-cs.3541}
}
EUVIMED