Detección de cáncer de páncreas mediante una red neuronal convolucional

Abstract

Introduction: Pancreatic cancer (PC) is a highly lethal malignant neoplasm worldwide, with an overall survival rate of less than five years after diagnosis, where risk factors may be hereditary or non-hereditary. Among exocrine malignant neoplasms, ductal adenocarcinoma is the most prevalent, accounting for approximately 90% of cases. Proper diagnosis and characterization of the tumor are strong predictors that help determine the type of treatment, the likelihood of metastasis, and survival. Literature Review: Convolutional Neural Networks (CNNs), when used to detect pancreatic cancer, achieve accuracy rates of over 95% in distinguishing cancerous from normal tissue. Furthermore, adequate training of CNNs on computed tomography (CT) scans enables the detection of tumors smaller than 2 centimeters with 87% sensitivity, while for tumors larger than 2 centimeters, accuracy increases to 95%, compared to 73% achieved in evaluat ions performed by medical personnel; however, such algorithms require detailed, high-qua lity information. Conclusion: Continuous advances in artificial intelligence have transformed research into the detection of pancreatic cancer using CNNs. This system offers accuracy rates of over 95%; therefore, properly training a CNN on computed tomography scans has surpassed the accuracy of conventional medical evaluation.