Platform, device and process for annotation and classification of tissue specimens using convolutional neural network

Abstract

Embodiments described herein provide a platform, device and process for digital pathology that enable multi-level annotation and visualization of histopathologic slides using a modular arrangement of deep convolutional neural networks (CNNs). The CNNs can be trained using pathology images (e.g., in some cases increasing the base of data by breaking larger fields of view into smaller ones) to learn features consistent with certain pathologies. The platform can use the CNNs to visually annotate pathology slides at an interface tool of a display device. The platform can automate the process of selection, as well as provide an opportunity for the pathologist to see a depiction of predicted results. The platform can use the CNNs to identify regions of interest on pathology slides. The interface tool can enable a predicted region of interest (ROI) type to be visually presented on a surface map showing the basis of the prediction. If the ROI primarily lands in part of the hyperdimensional space not occupied by any training set, then the interface tool is capable of marking it as an ROI of unknown type.