This graphical interface provide pathologists with the ability to view a set of image patches preliminary labels, modify the labels and have the updated labels stored. This way a pathologists can quickly and effortlessly label new patches and revise old patches all on one platform.
The 2D Histological Tissue Type (HTT) encoding utility is used to convert image patches, extracted from Whole Slide Images (WSI), to encode different shapes and morphological structure of tissues and cells. The abstract level representation of a WSI image patch, therefore, can be efficiently stored and processed for pattern matching.
HistoSegNet is the first weakly-supervised semantic segmentation technique developed for histological tissue type segmentation; it was first proposed in our lab's ICCV 2019 paper submission. It is trained on the patch-level tissue type annotations of the ADP database and produces pixel-level tissue type predictions for entire whole slide images. Since the histological appearance of tissue is an important indicator of disease and organ type, this project is fundamental for our lab's related abnormality detection and organ classification projects.
The primary goal of this application is to produce a meaningful encoding of Whole Slide Images (WSIs) by the use of HistoNet's encoding of Histological Tissue Types (HTTs) in patches. The large size of WSIs, ranging from 1 GB to 25 GB results in storage and computational challenges. It would be significantly beneficial to develop a more efficient encoding of the WSI for purposes such as querying similar WSIs, obtaining WSI level information from patch level analysis, and biometrics for WSI disease recognition.
Abnormality Detector is an automated tool to detect tissue abnormalities within whole slide images. Both abnormality detections on slide-level and on image patch level are available with this tool.
Is a custom convolutional neural network (CNN) tailored to digital pathology, which assists in histological tissue type classification.
This tool will allow users to create an accurate high resolution images from a single low resolution digital input image. SISR gives users the ability to significantly reduce digital storage space of images while still maintaining visual data.
Receives Whole Slide Images (WSI) with annotated XML file exported from the Huron Viewer and extracts overlapping image patches from all annotated regions which include metadata corresponding to diagnostic relevance notes (diagnosed by pathologist). The user is then provided the extracted patches in a folder for each ROI and an overall excel file containing information about each patch extracted.