Pancreatic cancer grading in pathological images using deep learning convolutional neural networks

Background: Pancreatic cancer is one of the deadliest forms cancer. The grades define how aggressively will spread and give indication for doctors to make proper prognosis treatment. current method pancreatic grading, by means manual examination cancerous tissue following a biopsy, time consuming often results in misdiagnosis thus incorrect This paper presents an automated grading system from pathology images developed comparing deep learning models on two different pathological stains. Methods: A transfer-learning technique was adopted testing 14 ImageNet pre-trained models. were fine-tuned be trained with our dataset. Results: From experiment, DenseNet appeared best at classifying validation set up 95.61% accuracy despite small sample set. Conclusions: To knowledge, this first work based images. Previous works have either focused only detection (benign or malignant), radiology (computerized tomography [CT], magnetic resonance imaging [MRI] etc.). proposed can very useful pathologists facilitating semi-automated system, which address problems found grading.