مصطفي يعقوبخاني غياثوند

Pancreatic ductal adenocarcinoma (PDAC) is one of the deadliest and most aggressive types of cancer, often diagnosed at advanced stages due to the absence of early clinical symptoms, the complex anatomical location of the pancreas, and the lack of effective biomarkers. This study aims to improve early detection of the disease by eva‎luating the performance of urinary biomarkers (LYVE1, REG1B, and TFF1) in combination with advanced machine learning algorithms. A dataset of 590 urine samples from three groups (healthy controls, benign conditions, and pancreatic cancer) was collected and preprocessed. Various machine learning models—including logistic regression, decision tree, support vector machine, naïve Bayes, k-nearest neighbors, and gradient boosting—were applied to identify pancreatic cancer. The results indicated that the gradient boosting model achieved the best performance in cancer detection. Among the biomarkers, LYVE1 was identified as the most influential in distinguishing cancer patients, while TFF1 played a complementary role in enhancing diagnostic accuracy. A comparison between binary and multiclass classification approaches revealed that both are effective in cancer detection, although the binary model demonstrated superior ability in differentiating disease types. This study highlights the potential of urinary biomarkers as a non-invasive diagnostic tool and machine learning algorithms as powerful methods for the rapid and accurate detection of pancreatic cancer. The findings can serve as a foundation for developing efficient diagnostic systems, ultimately contributing to improved treatment outcomes and reduced mortality.

الگوريتم‌هاي يادگيري ماشين , بيوماركرهاي ادراري , سرطان لوزالمعده

Machine learning algorithms , Urinary biomarkers , , Pancreatic cancer

Mostafa Yaghoubkhani Ghiasvand

Dr. Mohammad Fathian Brojeny