Pancreatic Cancer Classification using Deep Learning

In order to enhance patient outcomes, pancreatic cancer, a particularly deadly illness, requires early identification and precise categorization. To tackle this important medical issue, our project makes use of deep learning and machine learning methods based on Python. Two strong algorithms were used in the Machine Learning phase: Random Forest Classifier and Naive Bayes. A great test score of 99.2% and an amazing Accuracy Train score of 100% were attained by the Random Forest Classifier. With a test score of 99.2% and an accuracy train score of 99.3%, Naive Bayes also showed strong performance. This phase uses the dataset “Urinary biomarkers for pancreatic cancer,” which consists of 590 records in three different classes: PDAC, Benign, and Control. This dataset’s main characteristics are TFF1, LYVE1, REG1B, and creatinine. The potential involvement of LYVE1 in tumor metastasis, the relationship between REG1B and pancreatic regeneration, the significance of TFF1 for urinary tract repair, and creatinine as a kidney function indicator are all examined. This project’s Deep Learning section made use of the Convolutional Neural Network (CNN) architecture. With a validation accuracy of 100% and a training accuracy of 98.7%, the CNN model demonstrated exceptional performance. 1411 photos in two classes—normal and pancreatic tumor—made up the dataset in this phase. The results of machine learning are enhanced by this deep learning method, which also adds to the project’s resilience. Combining machine learning and deep learning methods with an emphasis on image analysis and urine biomarkers offers a complete approach to pancreatic cancer diagnosis and classification. The remarkable precision of this project paves the way for potentially revolutionary applications in the fields of early cancer detection and medical diagnosis.