Advances in computational image analysis and machine learning in the field of digital pathology seem poised to make transformational changes in disease diagnosis and prognosis, yet currently 99% of all histopathology slides generated in clinical practice are never digitized. We seek to incorporate digital imaging into the standard glass-slide clinical pathology workflow using relatively inexpensive modifications to traditional clinical microscopes. Here we will introduce PathCAM, a pathology computer-assisted microscope for "ambient" digitization of histology slides during the standard pathology workflow. PathCAM leverages high-frequency spatiotemporal sampling and human-machine interaction to deliver complete and accurate digital records of clinical glass-slide analysis. We will discuss the development of the PathCAM system and technical approach, and demonstrate the unique datasets and capabilities afforded by this hybrid digital-analog pathology imaging and analysis platform.
In traditional histopathology images, and emerging ex vivo microscopy techniques that generate large (gigapixel) image data, new strategies are needed to efficiently search the images to identify salient areas for human recognition or computer-aided diagnosis, to aid in rapid and efficient review workflows. One strategy is to learn from pathologists to develop model observers that can predict the most significant areas of an image for human review or application of computationally-expensive diagnostic algorithms. To get further understanding of the pathologists’ perception and cognition, we developed a custom web-based multiresolution viewer based on OpenSeadragon, which can record view coordinate, zoom level, coordinate dwell time, and coordinate scan path of users viewing images, under the assumption that the view coordinate represents the real-time visual area. We conducted experiments on two types of data with multiple reviewers: 1) traditional histopathology images of radical prostatectomy specimens, and 2) whole surface images of prostate margins collected on intact organs using structured illumination microscopy. Overall error rate, viewing pattern, saliency maps, dwell time and zoom level on fixation clusters were analyzed on normal tissue and cancer loci/positive margins. The result of the pilot experiments demonstrates that saliency maps correspond well with known areas of tumor in histopathology images and residual cancer in tumor margin images. These data will be used to predict saliency maps on new images based on low-level image features and to test accuracy vs. expert reviewers. This tool has promise for automated image dimensionality reduction and diagnosis of histopathology images.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.