Proposta de modelo de previsão de estadiamento em pacientes diagnosticados com PDAC

Abstract

Pancreatic cancer (PC) is difficult to diagnose early because it progresses silently, without specific symptoms, and responds poorly to most treatments. Ninety percent of pancreatic cancer cases are pancreatic ductal adenocarcinoma (PDAC), and the overall five-year survival rate after diagnosis is only 12.8%. This low survival rate leads diagnosed patients to question how much time they have left. The TNM classification system for malignant tumors has been the most common method for assessing survival and supporting medical decision-making regarding curative or palliative interventions. However, this classification can only be performed after advanced imaging exams, requiring patients to undergo new imaging tests to monitor changes in staging. Not all patients have the resources, physical availability, or emotional capacity for constant re-evaluation. Given the high mortality rate and difficulty in detecting this neoplasm, several research studies have emerged in search of biomarkers for early diagnosis. However, few of these studies focus on developing methods for early prognosis. This research proposes and evaluates a prior staging prognosis model for PDAC based on urinary biomarkers used for PDAC diagnosis, combined with age and gender. To this end, data from various health centers were collected and analyzed using Machine Learning (ML) techniques. The adopted techniques for prior staging classification were K-Nearest Neighbors (KNN), Support Vector Machines (SVM), and Random Forest. Results: The KNN classifier achieved a maximum accuracy of 0.62, SVM reached an accuracy of 0.58, and Random Forest produced the best results with an accuracy of 0.81. This indicates that the use of biomarkers for prior staging classification can assist in medical decision-making and monitoring the progression of the neoplasm.