Pathology has played an essential role in diagnosis of gastrointestinal disorders. However, errors can occur due to complex systems, time constraints and variable inputs. This can be further complicated when the biopsy images share histological features. In light of this, developing assistive computational methods can help. The goal of applying computational methods in identifying diseases is to achieve a fast, reproducible and reasonably accurate methods that can be standardized.
Deep learning in detecting diseases in histopathology images has been an active area of research. Convolutional Neural Networks (CNN), a form of deep learning, is particularly suited for distinguishing features in biopsy images. Past work in this area involves using CNN to detect cancer metastases in high resolution biopsy images.
Applying CNN to high resolution biopsy images from patients of gastro-intestinal diseases, specifically, Celiac Disease and Environmental Enteropathy helps detect distinguishing features in tissues effected by them. These diseases have significantly overlapping features which makes them difficult to be diagnosed. CNNs learn from different areas of an image and look for similar patterns in new images and classify them based on which features look the same. Our hypothesis is this deep learning technique will find differences in histologically similar looking tissues that are sometimes indistinguishable under a microscopic lens.
We demonstrate in our work a viable deep learning architecture to classify duodenal biopsy images into if each of them is either has Celiac Disease, Environmental Enteropathy or Normal tissues. We build a model using a Resnet50 CNN model, pre-trained on the ImageNet dataset. We use transfer learning and differential learning rates to further train the model. We run color normalization on our dataset to remove color differences that occur in images caused due to using different reagents while staining or different scanning machines.
You need this ticket from Eventbrite to sign up:
Applied Machine Learning Conference.